feat(benchmarks): expand multilingual stemming quality evaluation

* cover all Radixor dictionary languages * add PRIMARY_OUTPUT, ANY_CANDIDATE, and ALL_CANDIDATES policies * measure pairwise over-stemming and under-stemming * add balanced accuracy and complementary quality metrics * compare single-output and multi-output stemmers fairly * improve result validation, reporting, and documentation * move stemming quality tests into the standard test source set * preserve the existing JMH benchmark structure and badge output
fix: retain only the 10 latest GitHub Pages builds
2026-07-20 23:20:17 +02:00 · 2026-07-20 01:37:17 +02:00 · 2026-07-06 01:51:33 +02:00 · 2026-07-04 22:38:58 +02:00 · 2026-07-03 20:03:58 +02:00 · 2026-07-03 18:55:36 +02:00
144 changed files with 24861 additions and 674 deletions
--- a/.github/workflows/benchmarks.yml
+++ b/.github/workflows/benchmarks.yml
@@ -56,7 +56,7 @@ jobs:
          test -f gradle/verification-metadata.xml

      - name: Run JMH benchmarks
-        run: ./gradlew clean jmh --no-daemon
+        run: ./gradlew clean jmh -Pjmh.includes='.*EnglishStemmerComparisonBenchmark.*' --no-daemon

      - name: Upload JMH reports
        uses: actions/upload-artifact@v4
@@ -65,4 +65,4 @@ jobs:
          path: |
            build/reports/jmh/**
            build/results/jmh/**
-          if-no-files-found: warn
+          if-no-files-found: warn
--- a/.github/workflows/pages.yml
+++ b/.github/workflows/pages.yml
@@ -70,7 +70,7 @@ jobs:
          test -f gradle/verification-metadata.xml

      - name: Build reports for publication
-        run: ./gradlew --no-daemon clean ciRelease pmdMain javadoc jacocoCiReleaseReport pitest jmh cyclonedxBom
+        run: ./gradlew --no-daemon clean ciRelease pmdMain javadoc jacocoCiReleaseReport pitest jmh -Pjmh.includes='.*EnglishStemmerComparisonBenchmark.*' cyclonedxBom

      - name: Prepare gh-pages worktree
        shell: bash
@@ -191,11 +191,8 @@ jobs:
          COVERAGE_BADGE_LATEST_LINK='<li><a href="./builds/latest/metrics/coverage-badge.json">Coverage Badge Metadata</a></li>'
          MUTATION_BADGE_LINK='<li><a href="./metrics/pitest-badge.json">Mutation Badge Metadata</a></li>'
          MUTATION_BADGE_LATEST_LINK='<li><a href="./builds/latest/metrics/pitest-badge.json">Mutation Badge Metadata</a></li>'
-          JMH_BADGE_LINK='<li><a href="./metrics/jmh-badge.json">Benchmark Badge Metadata</a></li>'
-          JMH_BADGE_LATEST_LINK='<li><a href="./builds/latest/metrics/jmh-badge.json">Benchmark Badge Metadata</a></li>'
          COVERAGE_BADGE_REPORT_MD='- [Coverage badge metadata](https://leogalambos.github.io/Radixor/builds/latest/metrics/coverage-badge.json)'
          MUTATION_BADGE_REPORT_MD='- [Mutation badge metadata](https://leogalambos.github.io/Radixor/builds/latest/metrics/pitest-badge.json)'
-          JMH_BADGE_REPORT_MD='- [Benchmark badge metadata](https://leogalambos.github.io/Radixor/builds/latest/metrics/jmh-badge.json)'

          if [ ! -f "${RUN_METRICS_DIR}/coverage-badge.json" ]; then
            COVERAGE_BADGE_LINK='<li>Coverage Badge Metadata: not available</li>'
@@ -207,10 +204,6 @@ jobs:
            MUTATION_BADGE_REPORT_MD='- Mutation badge metadata: not currently available'
          fi

-          if [ ! -f "${RUN_METRICS_DIR}/jmh-badge.json" ]; then
-            JMH_BADGE_REPORT_MD='- Benchmark badge metadata: not currently available'
-          fi
-
          cat > "${RUN_DIR}/index.html" <<EOF
          <!doctype html>
          <html lang="en">
@@ -239,7 +232,6 @@ jobs:
              ${SBOM_XML_LINK:-<li>SBOM (XML): not available</li>}
              ${COVERAGE_BADGE_LINK}
              ${MUTATION_BADGE_LINK}
-              ${JMH_BADGE_LINK}
              <li><a href="./pitest/">Mutation Testing Report</a></li>
              $(
                [ "${HAS_JMH}" = "true" ] && { echo "${JMH_TXT_LINK:-<li>Benchmark Results (TXT): not available</li>}"; echo "${JMH_CSV_LINK:-<li>Benchmark Results (CSV): not available</li>}"; } \
@@ -271,24 +263,38 @@ jobs:
          ${SBOM_JSON_REPORT_MD}
          ${SBOM_XML_REPORT_MD}

-          ## Benchmark and badge metadata
+          ## Benchmark reports and badge metadata

          ${JMH_TXT_REPORT_MD}
          ${JMH_CSV_REPORT_MD}
          ${COVERAGE_BADGE_REPORT_MD}
          ${MUTATION_BADGE_REPORT_MD}
-          ${JMH_BADGE_REPORT_MD}

          ## Historical runs

          - [Browse historical build reports](https://leogalambos.github.io/Radixor/builds/)
          EOF

+          # Retain only the 10 most recent numbered builds to stay within
+          # GitHub Pages capacity limits. The "latest" alias is kept separately.
+          mapfile -t EXPIRED_BUILDS < <(
+            find "${SITE_DIR}/builds" -mindepth 1 -maxdepth 1 -type d -printf '%P\n' \
+              | grep -E '^[0-9]+$' \
+              | sort -r -n \
+              | tail -n +11
+          )
+
+          for build in "${EXPIRED_BUILDS[@]}"; do
+            rm -rf "${SITE_DIR}/builds/${build}"
+          done
+
          {
            echo "# Historical Build Reports"
            echo
            echo "The following build report sets are currently published on GitHub Pages."
            echo
+            echo "To stay within GitHub Pages capacity limits, only the 10 most recent build report sets are retained."
+            echo
            echo "| Build | Published | Link |"
            echo "|---:|---|---|"

--- a/.gitignore
+++ b/.gitignore
@@ -37,6 +37,7 @@ local.properties
 .settings/
 .loadpath
 .recommenders
+.classpath

 # External tool builders
 .externalToolBuilders/
--- a/.project
+++ b/.project
@@ -2,22 +2,21 @@
 <projectDescription>
 	<name>Radixor</name>
 	<comment></comment>
-	<projects>
-	</projects>
-	<buildSpec>
-		<buildCommand>
-			<name>org.eclipse.jdt.core.javabuilder</name>
-			<arguments>
-			</arguments>
-		</buildCommand>
-		<buildCommand>
-			<name>org.eclipse.buildship.core.gradleprojectbuilder</name>
-			<arguments>
-			</arguments>
-		</buildCommand>
-	</buildSpec>
+	<projects/>
 	<natures>
 		<nature>org.eclipse.jdt.core.javanature</nature>
 		<nature>org.eclipse.buildship.core.gradleprojectnature</nature>
 	</natures>
+	<buildSpec>
+		<buildCommand>
+			<name>org.eclipse.jdt.core.javabuilder</name>
+			<arguments/>
+		</buildCommand>
+		<buildCommand>
+			<name>org.eclipse.buildship.core.gradleprojectbuilder</name>
+			<arguments/>
+		</buildCommand>
+	</buildSpec>
+	<linkedResources/>
+	<filteredResources/>
 </projectDescription>
--- a/.ruleset
+++ b/.ruleset
@@ -162,7 +162,7 @@
   <rule ref="category/java/design.xml/CollapsibleIfStatements"/>
   <rule ref="category/java/design.xml/CouplingBetweenObjects">
      <properties>
-         <property name="threshold" value="60" />
+         <property name="threshold" value="70" />
      </properties>
   </rule>
   <rule ref="category/java/design.xml/CyclomaticComplexity">
--- a/README.md
+++ b/README.md
@@ -1,17 +1,14 @@
-<img src="Radixor.png" width="30%" align="right" alt="Radixor logo" />
+<img src="docs/assets/images/banner.jpg" width="100%" alt="Radixor banner" />

-# Radixor
-
-[![Quality gates](https://github.com/leogalambos/Radixor/actions/workflows/build.yml/badge.svg?branch=main)](https://github.com/leogalambos/Radixor/actions/workflows/build.yml)
-[![Coverage](https://img.shields.io/endpoint?url=https://leogalambos.github.io/Radixor/builds/latest/metrics/coverage-badge.json)](https://leogalambos.github.io/Radixor/builds/latest/coverage/)
-[![Published reports](https://img.shields.io/badge/reports-GitHub%20Pages-blue)](https://leogalambos.github.io/Radixor/builds/latest/)
-[![Mutation score](https://img.shields.io/endpoint?url=https://leogalambos.github.io/Radixor/builds/latest/metrics/pitest-badge.json)](https://leogalambos.github.io/Radixor/builds/latest/pitest/)
-[![English benchmark](https://img.shields.io/endpoint?url=https://leogalambos.github.io/Radixor/builds/latest/metrics/jmh-badge.json)](https://leogalambos.github.io/Radixor/builds/latest/jmh/jmh-results.txt)
-[![Maven Central](https://img.shields.io/maven-central/v/org.egothor/radixor)](https://central.sonatype.com/artifact/org.egothor/radixor)
 [![License](https://img.shields.io/github/license/leogalambos/Radixor)](LICENSE)
 [![Java](https://img.shields.io/badge/Java-21%2B-brightgreen)](#)
+[![Maven Central](https://img.shields.io/maven-central/v/org.egothor/radixor)](https://central.sonatype.com/artifact/org.egothor/radixor)
+[![Published reports](https://img.shields.io/badge/reports-GitHub%20Pages-blue)](https://leogalambos.github.io/Radixor/builds/latest/)
+[![Quality gates](https://github.com/leogalambos/Radixor/actions/workflows/build.yml/badge.svg?branch=main)](https://github.com/leogalambos/Radixor/actions/workflows/build.yml)
+[![Coverage](https://img.shields.io/endpoint?url=https://leogalambos.github.io/Radixor/builds/latest/metrics/coverage-badge.json)](https://leogalambos.github.io/Radixor/builds/latest/coverage/)
+[![Mutation score](https://img.shields.io/endpoint?url=https://leogalambos.github.io/Radixor/builds/latest/metrics/pitest-badge.json)](https://leogalambos.github.io/Radixor/builds/latest/pitest/)

-*Fast, deterministic, multi-language stemming for Java, built around compact patch-command tries and measured at roughly 4× to 6× the throughput of the Snowball Porter stemmer family on the current English benchmark workload.*
+*Deterministic, multi-language stemming for Java, built around compact dictionary-derived patch-command tries with an explicit quality/speed trade-off.*

 **Radixor** is a modern multi-language stemming toolkit for Java in the tradition of the original **Egothor** approach. It learns compact word-to-stem transformations from dictionary data, stores them in compiled patch-command tries, and exposes a runtime model designed for speed, determinism, and operational simplicity. Unlike a closed-form dictionary lookup stemmer, Radixor can also generalize beyond explicitly listed word forms.

@@ -52,18 +49,33 @@ Radixor is especially attractive when you want something more adaptable than sim

 ## Performance

-Radixor includes a JMH benchmark suite for both its own algorithmic core and a side-by-side English comparison against the Snowball Porter stemmer family.
+Radixor performance is best read together with stemming quality. The English dictionary coverage benchmark builds contracted compiled patch tries from deterministic slices of the `US_UK` dictionary and then measures both exact-root agreement and changed-token runtime.

-On the current English comparison workload, Radixor with bundled `US_UK` reaches approximately **31 to 32 million tokens per second**. Snowball original Porter reaches approximately **8 million tokens per second**, and Snowball English (Porter2) approximately **5 to 5.5 million tokens per second**.
+| Used rows | Actual row ratio | All exact | Changed exact | Root preserved | Speed ms/op | Error ms | ns/token |
+| ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| 100% | 100.000% | 97.478% | 97.197% | 97.552% | 23.113 | 7.065 | 109.8 |
+| 90% | 90.000% | 97.047% | 94.913% | 97.613% | 21.270 | 9.914 | 101.0 |
+| 80% | 80.000% | 96.635% | 92.768% | 97.661% | 19.170 | 6.609 | 91.1 |
+| 70% | 70.000% | 96.209% | 90.565% | 97.705% | 20.857 | 6.734 | 99.1 |
+| 60% | 60.000% | 95.750% | 88.384% | 97.703% | 14.975 | 1.215 | 71.1 |
+| 50% | 50.000% | 95.262% | 86.107% | 97.690% | 15.249 | 1.078 | 72.4 |
+| 40% | 40.000% | 94.753% | 83.855% | 97.643% | 15.323 | 2.340 | 72.8 |
+| 30% | 30.000% | 94.208% | 81.651% | 97.537% | 16.778 | 2.643 | 79.7 |
+| 20% | 20.000% | 93.633% | 79.366% | 97.416% | 18.929 | 3.241 | 89.9 |
+| 10% | 10.000% | 92.868% | 76.516% | 97.204% | 19.124 | 1.883 | 90.9 |

-That places Radixor at approximately:
+Column meanings:

- **4× the throughput of Snowball original Porter**
- **6× the throughput of Snowball English (Porter2)**
+- `Used rows` is the requested deterministic percentage of English dictionary rows used to build the stemmer.
+- `Actual row ratio` is the selected row count divided by the full parsed dictionary row count.
+- `All exact` is exact agreement over every word/root pair in the full dictionary.
+- `Changed exact` is exact agreement only where the word differs from its root.
+- `Root preserved` is the share of already-root forms that remain unchanged.
+- `Speed ms/op` is JMH average time for one changed-token benchmark operation.
+- `Error ms` is the JMH score error converted to milliseconds.
+- `ns/token` is average nanoseconds per changed token in that operation.

-on the current benchmark workload.
-
-This is a throughput comparison on the same deterministic token stream. It is **not** a claim that the compared stemmers are linguistically equivalent or interchangeable.
+The contracted trie result is materially stronger than the older uncontracted profile: full English coverage reaches 97.478% all-token exactness and 97.197% changed-token exactness at 109.8 ns/token, while even a 10% deterministic dictionary slice remains at 92.868% all-token exactness and 76.516% changed-token exactness at 90.9 ns/token. This is why Radixor benchmark results are documented with both speed and quality instead of a single Porter speed badge.

 For benchmark scope, workload design, environment, commands, report locations, and interpretation guidance, see [Benchmarking](docs/benchmarking.md).

@@ -103,6 +115,9 @@ Compared with the historical baseline, Radixor emphasizes:
 - **frequency-aware deterministic ordering**  
  Candidate results are ordered consistently and reproducibly.

+- **contracted compiled patch tries**  
+  Uniform patch-command subtrees are collapsed into accepting leaves, reducing hot lookup depth while preserving preferred stemming results.
+
 - **practical subtree reduction modes**  
  Reduction can be tuned toward stronger compression or more conservative semantic preservation.

@@ -133,8 +148,14 @@ The repository keeps the front page concise and places detailed documentation un

 ### Getting Started

+- [Fast Track](docs/fast-track.md)  
+  The shortest path from adding the dependency to getting a first stem from a bundled dictionary.
+
 - [Quick Start](docs/quick-start.md)  
-  A practical first guide to loading, compiling, and using Radixor.
+  A broader developer walkthrough covering loading options, querying, extension, persistence, and metadata.
+
+- [Integration Deep Dive](docs/integration-deep-dive.md)  
+  Dependency setup, bundled dictionary selection, production lifecycle, search-pipeline guidance, and operational checklist.

 - [Built-in Languages](docs/built-in-languages.md)  
  Overview of bundled language resources such as `US_UK`.
@@ -159,6 +180,9 @@ The repository keeps the front page concise and places detailed documentation un
 - [Extending and Persisting Compiled Tries](docs/programmatic-extending-and-persistence.md)  
  Reopening compiled tries, rebuilding them, and writing binary artifacts.

+- [Migration and Backward Compatibility](docs/migration-and-backward-compatibility.md)  
+  Migration from serialized String patch-command application to `CompiledPatchCommand`.
+
 ### Concepts and Internals

 - [Architecture and Reduction Overview](docs/architecture-and-reduction.md)  
@@ -187,7 +211,10 @@ The repository keeps the front page concise and places detailed documentation un
  Engineering standards, validation posture, auditability, and operational model.

 - [Benchmarking](docs/benchmarking.md)  
-  JMH benchmark methodology, Porter comparison, and result interpretation.
+  JMH benchmark methodology, dictionary coverage trade-offs, speed, quality, and result interpretation.
+
+- [Benchmark Results](docs/benchmarks/index.md)  
+  Structured reference for methodology, corpora, environment, English coverage, and per-language result pages.

 - [Published Reports](docs/reports.md)  
  Entry points to CI-published reports and GitHub Pages artifacts.
--- a/build.gradle
+++ b/build.gradle
@@ -18,6 +18,8 @@ version = gitVersion(prefix:'release@')

 def benchmarkReportsDirectory = layout.buildDirectory.dir('reports/jmh')
 def sbomReportsDirectory = layout.buildDirectory.dir('reports/sbom')
+def jmhIncludesProperty = providers.gradleProperty('jmh.includes')
+        .orElse(providers.systemProperty('jmh.includes'))

 def nvdApiKey = providers.gradleProperty('nvdApiKey')
        .orElse(providers.environmentVariable('NVD_API_KEY'))
@@ -28,6 +30,11 @@ apply from: 'gradle/maven-pom.gradle'

 configurations {
    mockitoAgent
+    stemmingQualityJmhRuntime {
+        canBeConsumed = false
+        canBeResolved = true
+        extendsFrom(jmhImplementation, jmhRuntimeOnly)
+    }
 }

 java {
@@ -76,6 +83,16 @@ dependencies {
    }
 }

+sourceSets.jmh.compileClasspath = sourceSets.jmh.compileClasspath - sourceSets.test.output
+sourceSets.jmh.runtimeClasspath = sourceSets.jmh.runtimeClasspath - sourceSets.test.output
+sourceSets.test.compileClasspath += sourceSets.jmh.output + configurations.jmhCompileClasspath
+sourceSets.test.runtimeClasspath += sourceSets.jmh.output + configurations.jmhCompileClasspath
+
+tasks.named('compileJmhJava', JavaCompile) {
+    classpath = classpath - sourceSets.test.output
+    setDependsOn([tasks.named('classes')])
+}
+
 dependencyCheck {
    failBuildOnCVSS = 7.0
    failOnError = true
@@ -154,7 +171,10 @@ def configureJUnitPlatformTags = { Test task, String includeTagsExpr, String exc
 }

 tasks.named('test', Test) {
-    configureJUnitPlatformTags(it, cliIncludeTags, cliExcludeTags)
+    final def requestedIncludes = splitTagExpression(cliIncludeTags)
+    final boolean slowExplicitlyIncluded = requestedIncludes.contains('slow')
+    final String defaultExcludeTags = cliExcludeTags ?: (slowExplicitlyIncluded ? null : 'slow')
+    configureJUnitPlatformTags(it, cliIncludeTags, defaultExcludeTags)
    finalizedBy(tasks.named('jacocoTestReport'))
 }

@@ -421,6 +441,7 @@ tasks.named('distTar') {

 jmh {
    jmhVersion = '1.37'
+    includeTests = false
    warmupIterations = 3
    iterations = 5
    fork = 1
@@ -430,13 +451,18 @@ jmh {
    resultsFile = benchmarkReportsDirectory.map { it.file('jmh-results.csv').asFile }.get()
    humanOutputFile = benchmarkReportsDirectory.map { it.file('jmh-results.txt').asFile }.get()
    duplicateClassesStrategy = DuplicatesStrategy.EXCLUDE
+    if (jmhIncludesProperty.isPresent()) {
+        includes = [jmhIncludesProperty.get()]
+    }
 }

 tasks.named('jmh') {
    group = 'verification'
-    description = 'Runs JMH benchmarks for the Radixor algorithmic core and Snowball comparison suite.'
+    description = 'Runs JMH benchmarks for the Radixor algorithmic core and external stemmer comparison suites.'
 }

+apply from: 'gradle/lucene-benchmarks.gradle'
+
 tasks.register('regressionArtifactGenerator', JavaExec) {
    group = 'verification'
    description = 'Generates deterministic compiled trie regression artifacts.'
@@ -458,6 +484,85 @@ tasks.register('regressionArtifactGenerator', JavaExec) {
    }
 }

+tasks.register('stemmingQuality', JavaExec) {
+    group = 'verification'
+    description = 'Evaluates pairwise over-stemming and under-stemming against bundled dictionary groups.'
+    dependsOn(tasks.named('testClasses'))
+    dependsOn(tasks.named('jmhClasses'))
+    classpath = files(sourceSets.test.runtimeClasspath, configurations.stemmingQualityJmhRuntime)
+    mainClass = 'org.egothor.stemmer.benchmark.quality.StemmingQualityApplication'
+    args layout.buildDirectory.dir('reports/stemming-quality').get().asFile.absolutePath,
+            layout.projectDirectory.dir('src/main/resources').asFile.absolutePath,
+            providers.gradleProperty('stemmingQualityLanguage').getOrElse(''),
+            providers.gradleProperty('stemmingQualityStemmer').getOrElse(''),
+            providers.gradleProperty('stemmingQualityMode').getOrElse(''),
+            providers.gradleProperty('stemmingQualityOutputPolicy').getOrElse(''),
+            providers.gradleProperty('stemmingQualityRankMetric').getOrElse('PAIRWISE_F05'),
+            providers.gradleProperty('stemmingQualityAudit').getOrElse('false'),
+            providers.gradleProperty('stemmingQualityAuditLimit').getOrElse('25')
+    maxHeapSize = '6g'
+}
+
+tasks.register('publishStemmingQualityDocumentation', JavaExec) {
+    group = 'documentation'
+    description = 'Publishes validated complete stemming-quality results on the language benchmark pages.'
+    dependsOn(tasks.named('testClasses'))
+    classpath = sourceSets.test.runtimeClasspath
+    mainClass = 'org.egothor.stemmer.benchmark.quality.StemmingQualityDocumentationPublisher'
+    args layout.buildDirectory.file('reports/stemming-quality/stemming-quality.csv').get().asFile.absolutePath,
+            layout.projectDirectory.dir('docs').asFile.absolutePath,
+            'update'
+    doFirst {
+        if (!file("$buildDir/reports/stemming-quality/stemming-quality.csv").isFile()) {
+            throw new GradleException('A complete stemming-quality CSV is required. Run stemmingQuality only when no validated complete report is available.')
+        }
+    }
+}
+
+tasks.register('verifyStemmingQualityDocumentation', JavaExec) {
+    group = 'verification'
+    description = 'Verifies published language-page quality tables against the checked-in authoritative CSV.'
+    dependsOn(tasks.named('testClasses'))
+    classpath = sourceSets.test.runtimeClasspath
+    mainClass = 'org.egothor.stemmer.benchmark.quality.StemmingQualityDocumentationPublisher'
+    args layout.projectDirectory.file('docs/benchmarks/data/stemming-quality.csv').asFile.absolutePath,
+            layout.projectDirectory.dir('docs').asFile.absolutePath,
+            'verify'
+}
+
+tasks.named('check') {
+    dependsOn(tasks.named('verifyStemmingQualityDocumentation'))
+}
+
+tasks.register('verifyStemmingQualitySourceSets') {
+    group = 'verification'
+    description = 'Verifies the production, JMH, and standard-test ownership of stemming-quality infrastructure.'
+    doLast {
+        if (sourceSets.findByName('stemmingQualityTest') != null || file('src/stemmingQualityTest').exists()) {
+            throw new GradleException('The obsolete stemmingQualityTest source set or directory still exists.')
+        }
+        if (!file('src/jmh/java/org/egothor/stemmer/benchmark/QualityStemmerMatrix.java').isFile()) {
+            throw new GradleException('The authoritative JMH stemmer matrix is not in src/jmh.')
+        }
+        if (!file('src/test/java/org/egothor/stemmer/benchmark/quality/StemmingQualityApplication.java').isFile()) {
+            throw new GradleException('The stemming-quality evaluator is not in the standard test source set.')
+        }
+    }
+}
+
+tasks.register('verifyProductionJarExcludesStemmingQuality') {
+    group = 'verification'
+    description = 'Verifies that analytical stemming-quality classes are absent from the production JAR.'
+    dependsOn(tasks.named('jar'))
+    doLast {
+        final File archive = tasks.named('jar').get().archiveFile.get().asFile
+        final def forbidden = zipTree(archive).matching { include '**/benchmark/**' }.files
+        if (!forbidden.isEmpty()) {
+            throw new GradleException("Production JAR contains analytical stemming-quality classes: ${forbidden}")
+        }
+    }
+}
+
 tasks.register('printDependencyCheckNvdConfig') {
    doLast {
        System.out.println("NVD API key present: " + (nvdApiKey != null && !nvdApiKey.isBlank()))
@@ -506,6 +611,10 @@ javadoc {
 }

 apply from: 'gradle/snowball-benchmarks.gradle'
+apply from: 'gradle/paicehusk-benchmarks.gradle'
+apply from: 'gradle/opennlp-benchmarks.gradle'
+apply from: 'gradle/hunspell-benchmarks.gradle'
+apply from: 'gradle/cistem-benchmarks.gradle'

 gradle.taskGraph.whenReady { taskGraph ->
    def banner = """
--- a/docs/architecture.md
+++ b/docs/architecture.md
@@ -53,6 +53,16 @@ Those patch-command values are inserted into a mutable trie keyed by the source

 Equivalent subtrees are merged into canonical reduced nodes.

+Before a selected semantic reduction mode is applied, Radixor also performs uniform-subtree
+contraction. If every reachable entry below a subtree resolves to the same preferred patch
+command, that subtree can be represented as an accepting leaf for that command. Runtime lookup can
+then stop at that leaf even when the input word still has remaining characters.
+
+This is a structural optimization of preferred-result lookup. It reduces trie depth in regions
+where the remaining suffix cannot change the selected command, while preserving the `get()` result
+used by the standard stemmer path. The benchmark tables in `docs/benchmarks/` are based on this
+contracted compiled representation.
+
 ### Compilation

 The reduced structure is frozen into an immutable compiled trie optimized for runtime lookup.
@@ -124,7 +134,7 @@ At runtime, lookup is conceptually simple:
 3. retrieve one or more stored patch commands,
 4. apply the chosen patch command to the original word.

-The trie itself does not create the final stem string. It selects the stored transformation command. `PatchCommandEncoder.apply(...)` then performs the actual transformation.
+The trie itself does not create the final stem string. It selects the stored transformation command. Runtime code should use `CompiledPatchCommand.apply(...)` so the serialized command is compiled once and reused.

 That separation is architecturally important:

@@ -170,6 +180,7 @@ The final compiled trie can be much smaller than the original dictionary for sev

 - patch commands are compact,
 - trie paths reuse shared structure,
+- uniform preferred-command subtrees can be contracted into accepting leaves,
 - reduction merges equivalent subtrees,
 - binary persistence stores the already reduced form,
 - GZip compression is applied on top of the binary format.
--- a/docs/assets/stylesheets/extra.css
+++ b/docs/assets/stylesheets/extra.css
@@ -67,6 +67,104 @@
  padding: 0.45rem 0.7rem;
 }

+/* Publication-quality benchmark tables retain identity columns while scrolling. */
+.quality-table {
+  max-width: 100%;
+  overflow-x: auto;
+  margin: 0.65rem 0 1rem;
+  border: 1px solid var(--md-default-fg-color--lightest);
+  border-radius: 0.2rem;
+  scrollbar-gutter: stable;
+}
+
+.quality-table:focus {
+  outline: 0.15rem solid var(--md-accent-fg-color);
+  outline-offset: 0.1rem;
+}
+
+.quality-table::before {
+  content: "Scrollable table: Rank, Stemmer, and Output policy remain visible.";
+  display: block;
+  padding: 0.35rem 0.55rem;
+  color: var(--md-default-fg-color--light);
+  font-size: 0.68rem;
+}
+
+.quality-table .md-typeset__table,
+.quality-table table {
+  margin: 0;
+}
+
+.quality-table table th:nth-child(1),
+.quality-table table td:nth-child(1),
+.quality-table table th:nth-child(2),
+.quality-table table td:nth-child(2),
+.quality-table table th:nth-child(3),
+.quality-table table td:nth-child(3) {
+  position: sticky;
+  z-index: 2;
+  background: var(--md-default-bg-color);
+  background-clip: padding-box;
+}
+
+.quality-table table th:nth-child(1),
+.quality-table table td:nth-child(1) {
+  left: 0;
+  min-width: 2.8rem;
+}
+
+.quality-table table th:nth-child(2),
+.quality-table table td:nth-child(2) {
+  left: 2.8rem;
+  min-width: 13rem;
+  white-space: normal;
+}
+
+.quality-table table th:nth-child(3),
+.quality-table table td:nth-child(3) {
+  left: 15.8rem;
+  min-width: 8.5rem;
+  box-shadow: 0.2rem 0 0.25rem rgb(0 0 0 / 8%);
+}
+
+.quality-details > summary {
+  font-weight: 600;
+}
+
+@media screen and (max-width: 44.99em) {
+  .quality-table table th:nth-child(2),
+  .quality-table table td:nth-child(2) {
+    min-width: 10rem;
+  }
+
+  .quality-table table th:nth-child(3),
+  .quality-table table td:nth-child(3) {
+    position: static;
+    min-width: 7.5rem;
+    box-shadow: none;
+  }
+}
+
+@media print {
+  .quality-table {
+    overflow: visible;
+    border: 0;
+  }
+
+  .quality-table::before {
+    display: none;
+  }
+
+  .quality-table table th,
+  .quality-table table td {
+    position: static !important;
+  }
+
+  .quality-details:not([open]) > *:not(summary) {
+    display: block;
+  }
+}
+
 /* Code blocks */
 .md-typeset pre > code {
  font-size: 0.72rem;
--- a/docs/benchmarking.md
+++ b/docs/benchmarking.md
@@ -1,159 +1,42 @@
 # Benchmarking

-Radixor includes a JMH benchmark suite for both the internal algorithmic core and a side-by-side English comparison against the Snowball Porter stemmer family.
+Radixor contains internal trie microbenchmarks, a separate stemmer comparison suite, and a dictionary coverage benchmark for Radixor itself. Published stemmer comparison results must come only from benchmark classes matching `.*StemmerComparisonBenchmark.*`; internal `FrequencyTrie*` microbenchmarks are not part of those results.

-This document explains what is benchmarked, how to run the suite, and how benchmark results should be interpreted.
+This page is the entry point for benchmark interpretation. Detailed tables and long reference material are split into focused subpages so that important points do not get buried.

-## Scope
+## Key Takeaways

-The benchmark suite currently covers two categories:
+- Speed and accuracy must be read together. A faster row is not necessarily a better stemmer.
+- Radixor is the quality-oriented baseline in same-language comparisons. Its exact-root accuracy is often close to 100%, while many faster competitors are light, minimal, possessive, or aggressive rule-based stemmers with much lower root agreement.
+- The measured Radixor cost buys dictionary-trained stemming precision. That precision improves search quality by mapping inflected forms to intended dictionary roots instead of approximate or over-reduced stems.
+- Speed benchmarks process changed dictionary tokens where the surface form differs from the expected root. Accuracy benchmarks process the complete dictionary.
+- Accuracy tables use deterministic auxiliary counters from the current JMH reports. Repeated measurement samples duplicate the same exact-root accounting and are not interpreted as timing results.
+- The historical Porter performance badge is retired. Benchmark reporting now uses speed and quality tables rather than a single Porter ratio.

- Radixor core operations,
- English stemmer comparison on the same token workload.
+## Benchmark Documentation Map

-The comparison benchmark processes the same deterministic English token stream through:
+| Page | Purpose |
+| --- | --- |
+| [Benchmark methodology](benchmarks/reference/methodology.md) | Workload design, speed pass, quality pass, normalization policy, and exact-root metrics. |
+| [Linguistic quality methodology](benchmarks/reference/linguistic-quality.md) | Pairwise gold standard, over/under-stemming, candidate policies, metrics, and ranking rules. |
+| [Tested stemmers](benchmarks/reference/tested-stemmers.md) | Upstream attribution, tested versions, language coverage, adapter behaviour, and limitations. |
+| [Reproducibility and raw data](benchmarks/reference/reproducibility.md) | Versioned quality snapshot, checksum, commands, reports, and provenance limitations. |
+| [Benchmark corpora](benchmarks/reference/corpora.md) | Dictionary row counts, complete quality tokens, already-root tokens, changed speed tokens, and timing token counts. |
+| [Benchmark environment and reports](benchmarks/reference/environment.md) | Hardware, OS, JVM, JMH settings, report files, and current badge/report policy. |
+| [English dictionary coverage benchmark](benchmarks/reference/english-coverage.md) | The quality/speed operating curve for contracted Radixor tries built from 100% down to 10% of English dictionary rows. |
+| [Candidate evaluation](benchmarks/reference/candidates.md) | Included benchmark families and evaluated candidates that were skipped. |
+| [Language benchmark pages](benchmarks/languages/index.md) | Per-language accuracy tables, speed tables, and implementation notes. |

- Radixor with bundled `US_UK` (older benchmark snapshots used the now-retired `US_UK_PROFI` resource),
- Snowball original Porter,
- Snowball English, commonly referred to as Porter2.
+## How To Read Results

-The purpose of the comparison is throughput measurement on identical input. It is not intended to demonstrate linguistic equivalence between the compared stemmers.
+Start with the [language benchmark pages](benchmarks/languages/index.md). Each language page lists accuracy first and speed second because throughput without root agreement is not enough to interpret stemmer quality.

-## How to read the published numbers
+When Radixor is slower than a narrow competitor, check the accuracy table before drawing a conclusion. Many Lucene light/minimal filters and possessive filters intentionally do less work. They can be fast precisely because they are not trying to match the dictionary root with the same precision.

-Two kinds of benchmark numbers are relevant in the project.
+The [English dictionary coverage benchmark](benchmarks/reference/english-coverage.md) shows the central operating curve explicitly: contracted tries preserve high quality even at reduced dictionary coverage, while changed-form exactness still reflects how much language knowledge was available during training. This is why Radixor performance should be discussed as a configurable quality/speed point, not as a single fixed ratio against Porter.

-### Reference measurements
+## Current Result Locations

-The detailed benchmark snapshot documented on this page comes from a controlled run on a Ryzen 5 system. Those numbers are the best reference point for understanding absolute throughput under a known local benchmark environment.
+The current measured language results are published in [Language Benchmark Pages](benchmarks/languages/index.md). Generated local report files for this benchmark update are listed in [Benchmark environment and reports](benchmarks/reference/environment.md).

-### Published badge figures
-
-The benchmark badge metadata published through GitHub Pages is generated in the GitHub-hosted container environment. That environment is convenient for continuous publication, but it is not the right place to treat absolute throughput values as stable across time. CPU scheduling, shared-host variability, and container-level noise can materially affect raw numbers from run to run.
-
-For that reason, the published badge values should be treated primarily as a compact status surface. They are useful for observing broad trends and relative positioning, but not as the authoritative source for precise absolute throughput claims.
-
-## Current snapshot
-
-A recent JMH run on JDK 21.0.10 with JMH 1.37, one thread, three warmup iterations, and five measurement iterations produced the following approximate throughput ranges:
-
-| Workload | Radixor `US_UK` *(historical runs: `US_UK_PROFI`)* | Snowball Porter | Snowball English |
-| --- | ---: | ---: | ---: |
-| About 12,000 generated tokens | 30.99 M tokens/s | 8.21 M tokens/s | 5.46 M tokens/s |
-| About 60,000 generated tokens | 32.25 M tokens/s | 8.02 M tokens/s | 5.11 M tokens/s |
-
-On that workload, Radixor measured approximately:
-
- 4 times the throughput of Snowball original Porter,
- 6 times the throughput of Snowball English.
-
-These values are workload-dependent and environment-dependent. They should be read as measured results for the documented setup, not as universal constants.
-
-## Interpreting the relative result
-
-Although the absolute numbers can move across environments, the throughput relationship between Radixor and the compared Porter-family stemmers has remained broadly stable in practical measurements. In particular, the comparison against Snowball original Porter is consistently in the rough range of about four to one in Radixor’s favor.
-
-That relative behavior is more informative than any single absolute figure. It reflects a real architectural difference rather than a cosmetic benchmark artifact.
-
-Radixor is built around a compiled patch-command trie that resolves the result through a direct lookup and patch application path. In contrast, classic rule-based stemmers such as the Porter family follow a different operational model. The result is that Radixor combines two properties that do not often appear together:
-
- dictionary-driven compiled lookup performance,
- the ability to generalize beyond explicitly listed word forms instead of behaving like a pure closed-form dictionary lookup table.
-
-Within that design space, the measured throughput profile is strong enough to place Radixor among the fastest known practical implementations of this kind, while still supporting stemming of previously unseen forms. That should still be read as a carefully bounded engineering statement, not as an absolute claim over every possible stemmer architecture or benchmark scenario.
-
-## Benchmark classes
-
-The main benchmark classes are under `src/jmh/java/org/egothor/stemmer/benchmark`.
-
-Relevant classes include:
-
- `FrequencyTrieLookupBenchmark`,
- `FrequencyTrieCompilationBenchmark`,
- `EnglishStemmerComparisonBenchmark`.
-
-The English comparison benchmark uses the bundled Radixor English resource and the official Snowball Java distribution integrated into the JMH source set.
-
-## Workload design
-
-The English comparison benchmark uses a deterministic generated corpus rather than an uncontrolled ad hoc text sample.
-
-The workload intentionally mixes:
-
- simple inflections,
- common derivational forms,
- US and UK spelling families,
- lexical forms appropriate for the current bundled `US_UK` resource (with historical continuity from earlier `US_UK_PROFI` runs).
-
-This design keeps runs reproducible across environments and avoids accidental drift caused by changing external corpora.
-
-## Running benchmarks
-
-Run the full benchmark suite:
-
-```bash
-./gradlew jmh
-```
-
-Run only the English comparison benchmark:
-
-```bash
-./gradlew jmh -Pjmh.includes=EnglishStemmerComparisonBenchmark
-```
-
-## Generated reports
-
-JMH reports are written to:
-
- `build/reports/jmh/jmh-results.txt`,
- `build/reports/jmh/jmh-results.csv`.
-
-The text report is convenient for human review. The CSV report is more useful for CI archiving, historical tracking, and external processing.
-
-## Interpreting results responsibly
-
-Benchmark numbers should always be read with care.
-
-Important factors include:
-
- CPU model and frequency behavior,
- thermal throttling,
- JVM vendor and version,
- system background load,
- operating-system scheduling noise,
- benchmark parameter changes.
-
-For meaningful comparison, keep these stable:
-
- hardware or VM class,
- JDK version,
- benchmark parameters,
- thread count,
- benchmark source revision.
-
-If a regression is suspected, repeat the run and compare against previous CSV output rather than relying on a single measurement.
-
-## Regression tracking
-
-The recommended regression workflow is:
-
-1. archive `jmh-results.csv`,
-2. compare the same benchmark names across runs,
-3. compare only like-for-like environments,
-4. investigate sustained regressions rather than one-off noise.
-
-For public reporting, the README should keep only the condensed benchmark summary, while detailed benchmark methodology and interpretation should remain in this document.
-
-## Notes on comparison fairness
-
-Radixor, Snowball Porter, and Snowball English are not the same kind of stemmer.
-
-Radixor uses a compiled patch-command trie driven by dictionary data. Snowball Porter and Snowball English are rule-based English stemmers.
-
-Because of that, the comparison should be understood as:
-
- equal input workload,
- different stemming strategies,
- measured throughput rather than semantic identity.
-
-That distinction matters whenever performance claims are discussed in documentation, release notes, or badge summaries.
+JMH TXT and CSV reports are still published as benchmark artifacts. They are no longer converted into a Shields endpoint benchmark badge.
--- a/docs/benchmarks/data/stemming-quality.csv
+++ b/docs/benchmarks/data/stemming-quality.csv
@@ -0,0 +1,309 @@
+Stemmer,Language,Dictionary mode,Output policy,Applied dictionary rows,Processed word forms,Singleton dictionary rows,Forms with one candidate,Forms with multiple candidates,Maximum candidates for one form,Total candidate assignments,Distinct output stems,True-positive pairs,False-positive pairs,False-negative pairs,True-negative pairs,Over-stemming error pairs,Over-stemming possible pairs,Over-stemming percentage,Under-stemming error pairs,Under-stemming possible pairs,Under-stemming percentage,Pairwise precision,Pairwise recall,Pairwise specificity,Pairwise accuracy,Balanced accuracy,Pairwise F0.5,Pairwise F1,Pairwise F2,Jaccard index,Fowlkes-Mallows index,Matthews correlation coefficient,Pairwise error rate,Adjusted Rand Index,Homogeneity,Completeness,V-measure,Normalized mutual information
+"CZECH_LUCENE_CZECH_STEM_FILTER","CS_CZ","ALL_WORDS","PRIMARY_OUTPUT","5113","51676","2","51676","0","1","51676","9647","177249","14480","124586","1334862335","14480","1334876815","0.001085","124586","301835","41.276194","0.924476735392","0.587238060530","0.999989152557","0.999895844650","0.793613606543","0.829234311828","0.718241200736","0.633453389361","0.560355974266","0.736809286788","0.736765291417","0.000104155350","0.718191706079","0.993800637348","0.944976928457","0.968774025802","0.968774025802"
+"CZECH_LUCENE_CZECH_STEM_FILTER","CS_CZ","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","5038","50968","2","50968","0","1","50968","9558","174387","13950","124426","1298530265","13950","1298544215","0.001074","124426","298813","41.640089","0.925930645598","0.583599107134","0.999989257201","0.999893462107","0.791794182167","0.828708724235","0.715947860002","0.630197985095","0.557569149804","0.735100195918","0.735055396892","0.000106537893","0.715897321649","0.993897397445","0.944297456221","0.968462775946","0.968462775946"
+"CZECH_RADIXOR","CS_CZ","ALL_WORDS","PRIMARY_OUTPUT","5113","51676","2","51676","0","1","51676","5162","299762","3867","2073","1334872948","3867","1334876815","0.000290","2073","301835","0.686799","0.987264062392","0.993132009210","0.999997103103","0.999995551157","0.996564556157","0.988432097845","0.990189342389","0.991952846154","0.980569312599","0.990193689085","0.990191466141","0.000004448843","0.990187117482","0.998733220675","0.998685552738","0.998709386137","0.998709386137"
+"CZECH_RADIXOR","CS_CZ","ALL_WORDS","ANY_CANDIDATE","5113","51676","2","51080","596","4","52319","5166","301835","0","0","1334876815","0","1334876815","0.000000","0","301835","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"CZECH_RADIXOR","CS_CZ","ALL_WORDS","ALL_CANDIDATES","5113","51676","2","51080","596","4","52319","5166","301835","5850","0","1334870965","5850","1334876815","0.000438","0","301835","0.000000","0.980987048442","1.000000000000","0.999995617573","0.999995618564","0.999997808787","0.984731579205","0.990402283764","0.996138677580","0.980987048442","0.990447902942","0.990445732657","0.000004381436","","","","",""
+"CZECH_RADIXOR","CS_CZ","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","5038","50968","2","50968","0","1","50968","5037","297104","3863","1709","1298540352","3863","1298544215","0.000297","1709","298813","0.571930","0.987164705765","0.994280703985","0.999997025130","0.999995710028","0.997138864558","0.988579745136","0.990709926973","0.992849308824","0.981590876052","0.990716315904","0.990714173387","0.000004289972","0.990707781520","0.998726091764","0.999029907266","0.998877976413","0.998877976413"
+"CZECH_RADIXOR","CS_CZ","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","5038","50968","2","50428","540","4","51543","5040","298813","0","0","1298544215","0","1298544215","0.000000","0","298813","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"CZECH_RADIXOR","CS_CZ","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","5038","50968","2","50428","540","4","51543","5040","298813","5782","0","1298538433","5782","1298544215","0.000445","0","298813","0.000000","0.981017416570","1.000000000000","0.999995547321","0.999995548346","0.999997773661","0.984756059381","0.990417760454","0.996144940117","0.981017416570","0.990463233325","0.990461028216","0.000004451654","","","","",""
+"DA_DK_RADIXOR","DA_DK","ALL_WORDS","PRIMARY_OUTPUT","4179","28079","32","28079","0","1","28079","4184","89188","1165","707","394110021","1165","394111186","0.000296","707","89895","0.786473","0.987106128186","0.992135268925","0.999997043981","0.999995251155","0.996066156453","0.988107873355","0.989614309174","0.991125345329","0.979442126071","0.989617503860","0.989615130363","0.000004748845","0.989611934224","0.998465862775","0.998718664384","0.998592247580","0.998592247580"
+"DA_DK_RADIXOR","DA_DK","ALL_WORDS","ANY_CANDIDATE","4179","28079","32","27756","323","3","28405","4187","89895","0","0","394111186","0","394111186","0.000000","0","89895","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"DA_DK_RADIXOR","DA_DK","ALL_WORDS","ALL_CANDIDATES","4179","28079","32","27756","323","3","28405","4187","89895","1849","0","394109337","1849","394111186","0.000469","0","89895","0.000000","0.979846093478","1.000000000000","0.999995308431","0.999995309500","0.999997654215","0.983811622975","0.989820468071","0.995903164910","0.979846093478","0.989871756076","0.989869434047","0.000004690500","","","","",""
+"DA_DK_RADIXOR","DA_DK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4173","28033","32","28033","0","1","28033","4170","89077","1165","663","392819623","1165","392820788","0.000297","663","89740","0.738801","0.987090268389","0.992611990194","0.999997034271","0.999995347541","0.996304512232","0.988189692661","0.989843428787","0.991502709249","0.979891095099","0.989847279032","0.989844954043","0.000004652459","0.989841102043","0.998463063294","0.998811876590","0.998637439483","0.998637439483"
+"DA_DK_RADIXOR","DA_DK","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","4173","28033","32","27718","315","3","28351","4173","89740","0","0","392820788","0","392820788","0.000000","0","89740","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"DA_DK_RADIXOR","DA_DK","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","4173","28033","32","27718","315","3","28351","4173","89740","1849","0","392818939","1849","392820788","0.000471","0","89740","0.000000","0.979811986156","1.000000000000","0.999995293019","0.999995294094","0.999997646509","0.983784115625","0.989803065147","0.995896117847","0.979811986156","0.989854527774","0.989852198158","0.000004705906","","","","",""
+"ENGLISH_LUCENE_KSTEM_FILTER","US_UK","ALL_WORDS","PRIMARY_OUTPUT","396939","607439","250964","607439","0","1","607439","371125","237565","1368501","76305","184489083270","1368501","184490451771","0.000742","76305","313870","24.311020","0.147917333410","0.756889795138","0.999992582267","0.999992168681","0.878441188702","0.176283968232","0.247471790726","0.415099040868","0.141208449266","0.334599940499","0.334597833111","0.000007831319","0.247469648794","0.980686838187","0.992107972963","0.986364345289","0.986364345289"
+"ENGLISH_LUCENE_KSTEM_FILTER","US_UK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","374384","583910","228735","583910","0","1","583910","347624","237551","1367069","74340","170473473135","1367069","170474840204","0.000802","74340","311891","23.835250","0.148041904002","0.761647498645","0.999991980817","0.999991544756","0.880819739731","0.176476898525","0.247899438093","0.416437018089","0.141486991947","0.335791223646","0.335788961354","0.000008455244","0.247897133948","0.979822223835","0.991990504792","0.985868818390","0.985868818390"
+"ENGLISH_LUCENE_MINIMAL_FILTER","US_UK","ALL_WORDS","PRIMARY_OUTPUT","396939","607439","250964","607439","0","1","607439","453328","137225","1122264","176645","184489329507","1122264","184490451771","0.000608","176645","313870","56.279670","0.108952916619","0.437203300730","0.999993916953","0.999992959490","0.718598608842","0.128203906480","0.174435713655","0.272816484020","0.095551668577","0.218253464509","0.218250987161","0.000007040510","0.174433464995","0.995202198233","0.981173943304","0.988138284715","0.988138284715"
+"ENGLISH_LUCENE_MINIMAL_FILTER","US_UK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","374384","583910","228735","583910","0","1","583910","430129","136932","1120871","174959","170473719333","1120871","170474840204","0.000657","174959","311891","56.096200","0.108866014789","0.439037997249","0.999993425006","0.999992398716","0.719515711128","0.128139023335","0.174469673707","0.273277328232","0.095572048952","0.218623688336","0.218621020778","0.000007601284","0.174467253215","0.994993790771","0.980519680106","0.987703711280","0.987703711280"
+"ENGLISH_LUCENE_PORTER_COPIED","US_UK","ALL_WORDS","PRIMARY_OUTPUT","396939","607439","250964","607439","0","1","607439","319968","285390","1557406","28480","184488894365","1557406","184490451771","0.000844","28480","313870","9.073820","0.154867928951","0.909261796285","0.999991558338","0.999991403982","0.954626677312","0.185678591198","0.264658505304","0.460562583837","0.152510906996","0.375253902399","0.375251973425","0.000008596018","0.264656367392","0.969648379409","0.997199419831","0.983230936080","0.983230936080"
+"ENGLISH_LUCENE_PORTER_COPIED","US_UK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","374384","583910","228735","583910","0","1","583910","298779","283761","1552702","28130","170473287502","1552702","170474840204","0.000911","28130","311891","9.019177","0.154514956196","0.909808234287","0.999990891899","0.999990726907","0.954899563093","0.185277176317","0.264165961476","0.460049474275","0.152183881415","0.374938634269","0.374936557303","0.000009273093","0.264163659878","0.968644600847","0.997108656044","0.982670549062","0.982670549062"
+"ENGLISH_LUCENE_PORTER_FILTER","US_UK","ALL_WORDS","PRIMARY_OUTPUT","396939","607439","250964","607439","0","1","607439","319968","285390","1557406","28480","184488894365","1557406","184490451771","0.000844","28480","313870","9.073820","0.154867928951","0.909261796285","0.999991558338","0.999991403982","0.954626677312","0.185678591198","0.264658505304","0.460562583837","0.152510906996","0.375253902399","0.375251973425","0.000008596018","0.264656367392","0.969648379409","0.997199419831","0.983230936080","0.983230936080"
+"ENGLISH_LUCENE_PORTER_FILTER","US_UK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","374384","583910","228735","583910","0","1","583910","298779","283761","1552702","28130","170473287502","1552702","170474840204","0.000911","28130","311891","9.019177","0.154514956196","0.909808234287","0.999990891899","0.999990726907","0.954899563093","0.185277176317","0.264165961476","0.460049474275","0.152183881415","0.374938634269","0.374936557303","0.000009273093","0.264163659878","0.968644600847","0.997108656044","0.982670549062","0.982670549062"
+"ENGLISH_LUCENE_POSSESSIVE_FILTER","US_UK","ALL_WORDS","PRIMARY_OUTPUT","396939","607439","250964","607439","0","1","607439","591899","7","1115154","313863","184489336617","1115154","184490451771","0.000604","313863","313870","99.997770","0.000006277121","0.000022302227","0.999993955492","0.999992254263","0.500008128860","0.000007330589","0.000009796848","0.000014763939","0.000004898448","0.000011831896","0.000008625150","0.000007745737","0.000007141644","0.995789196698","0.958018540631","0.976538780935","0.976538780935"
+"ENGLISH_LUCENE_POSSESSIVE_FILTER","US_UK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","374384","583910","228735","583910","0","1","583910","568400","5","1113773","311886","170473726431","1113773","170474840204","0.000653","311886","311891","99.998397","0.000004489225","0.000016031242","0.999993466643","0.999991637145","0.500004748942","0.000005244385","0.000007014251","0.000010587200","0.000003507138","0.000008483387","0.000005026087","0.000008362855","0.000004155674","0.995605378040","0.956423154691","0.975621022465","0.975621022465"
+"ENGLISH_OPENNLP_PORTER","US_UK","ALL_WORDS","PRIMARY_OUTPUT","396939","607439","250964","607439","0","1","607439","319968","285390","1557406","28480","184488894365","1557406","184490451771","0.000844","28480","313870","9.073820","0.154867928951","0.909261796285","0.999991558338","0.999991403982","0.954626677312","0.185678591198","0.264658505304","0.460562583837","0.152510906996","0.375253902399","0.375251973425","0.000008596018","0.264656367392","0.969648379409","0.997199419831","0.983230936080","0.983230936080"
+"ENGLISH_OPENNLP_PORTER","US_UK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","374384","583910","228735","583910","0","1","583910","298779","283761","1552702","28130","170473287502","1552702","170474840204","0.000911","28130","311891","9.019177","0.154514956196","0.909808234287","0.999990891899","0.999990726907","0.954899563093","0.185277176317","0.264165961476","0.460049474275","0.152183881415","0.374938634269","0.374936557303","0.000009273093","0.264163659878","0.968644600847","0.997108656044","0.982670549062","0.982670549062"
+"ENGLISH_PAICE_HUSK_LANCASTER","US_UK","ALL_WORDS","PRIMARY_OUTPUT","396939","607439","250964","607439","0","1","607439","268169","283991","3062661","29879","184487389110","3062661","184490451771","0.001660","29879","313870","9.519546","0.084858240415","0.904804536910","0.999983399352","0.999983237427","0.952393968131","0.103642734217","0.155164208820","0.308542866654","0.084107327905","0.277092260667","0.277089454298","0.000016762573","0.155161580693","0.937768073854","0.996599815184","0.966289292109","0.966289292109"
+"ENGLISH_PAICE_HUSK_LANCASTER","US_UK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","374384","583910","228735","583910","0","1","583910","249411","282398","3045870","29493","170471794334","3045870","170474840204","0.001787","29493","311891","9.456188","0.084848335531","0.905438117804","0.999982133023","0.999981960051","0.952710125414","0.103632575002","0.155156958803","0.308575577075","0.084103067491","0.277173081705","0.277170064389","0.000018039949","0.155154132315","0.936076835754","0.996486960554","0.965337716341","0.965337716341"
+"ENGLISH_RADIXOR","US_UK","ALL_WORDS","PRIMARY_OUTPUT","396939","607439","250964","607439","0","1","607439","390361","292001","1149886","21869","184489301885","1149886","184490451771","0.000623","21869","313870","6.967534","0.202513095686","0.930324656705","0.999993767233","0.999993648707","0.965159211969","0.240076409811","0.332621199859","0.541270431499","0.199487482886","0.434054059102","0.434052478080","0.000006351293","0.332619335001","0.994214506865","0.997769723414","0.995988942533","0.995988942533"
+"ENGLISH_RADIXOR","US_UK","ALL_WORDS","ANY_CANDIDATE","396939","607439","250964","578231","29208","1355","2838145","397392","313855","12","15","184490451759","12","184490451771","0.000000","15","313870","0.004779","0.999961767245","0.999952209513","0.999999999935","0.999999999854","0.999976104724","0.999959855684","0.999956988357","0.999954121045","0.999913980413","0.999956988368","0.999956988295","0.000000000146","","","","",""
+"ENGLISH_RADIXOR","US_UK","ALL_WORDS","ALL_CANDIDATES","396939","607439","250964","578231","29208","1355","2838145","397392","313855","11482166","15","184478969605","11482166","184490451771","0.006224","15","313870","0.004779","0.026606853277","0.999952209513","0.999937762817","0.999937762842","0.999944986165","0.033038791524","0.051834488023","0.120237128281","0.026606819443","0.163112175274","0.163107098882","0.000062237158","","","","",""
+"ENGLISH_RADIXOR","US_UK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","374384","583910","228735","583910","0","1","583910","367590","290572","1148489","21319","170473691715","1148489","170474840204","0.000674","21319","311891","6.835401","0.201917778329","0.931645991709","0.999993263000","0.999993137956","0.965819627354","0.239424468968","0.331901731173","0.540775136091","0.198970131062","0.433723286019","0.433721583515","0.000006862044","0.331899721995","0.993959181482","0.997731171071","0.995841604460","0.995841604460"
+"ENGLISH_RADIXOR","US_UK","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","374384","583910","228735","555084","28826","1355","2812871","374506","311891","0","0","170474840204","0","170474840204","0.000000","0","311891","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"ENGLISH_RADIXOR","US_UK","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","374384","583910","228735","555084","28826","1355","2812871","374506","311891","11470018","0","170463370186","11470018","170474840204","0.006728","0","311891","0.000000","0.026472025883","1.000000000000","0.999932717239","0.999932717362","0.999966358619","0.032872482055","0.051578660140","0.119686728696","0.026472025883","0.162702261457","0.162696787836","0.000067282638","","","","",""
+"ENGLISH_SNOWBALL_ORIGINAL_PORTER","US_UK","ALL_WORDS","PRIMARY_OUTPUT","396939","607439","250964","607439","0","1","607439","321092","285304","1555293","28566","184488896478","1555293","184490451771","0.000843","28566","313870","9.101220","0.155006228957","0.908987797496","0.999991569791","0.999991414969","0.954489683644","0.185835337999","0.264848800190","0.460750814660","0.152637303435","0.375364850057","0.375362921954","0.000008585031","0.264846663203","0.969891477221","0.997192899073","0.983352728141","0.983352728141"
+"ENGLISH_SNOWBALL_ORIGINAL_PORTER","US_UK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","374384","583910","228735","583910","0","1","583910","299877","283675","1550615","28216","170473289589","1550615","170474840204","0.000910","28216","311891","9.046750","0.154651118416","0.909532496930","0.999990904142","0.999990738644","0.954761700536","0.185431499934","0.264353286139","0.460234326480","0.152308234175","0.375046954242","0.375044878196","0.000009261356","0.264350985524","0.968893806180","0.997101844661","0.982795461434","0.982795461434"
+"ENGLISH_SNOWBALL_PORTER2","US_UK","ALL_WORDS","PRIMARY_OUTPUT","396939","607439","250964","607439","0","1","607439","318385","285334","1566711","28536","184488885060","1566711","184490451771","0.000849","28536","313870","9.091662","0.154064291094","0.909083378469","0.999991507902","0.999991353242","0.954537443185","0.184752753479","0.263476636895","0.459101696688","0.151726514306","0.374242282819","0.374240346981","0.000008646758","0.263474493989","0.969037354042","0.997181597682","0.982908049045","0.982908049045"
+"ENGLISH_SNOWBALL_PORTER2","US_UK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","374384","583910","228735","583910","0","1","583910","297220","283730","1561891","28161","170473278313","1561891","170474840204","0.000916","28161","311891","9.029116","0.153731454074","0.909708840589","0.999990837997","0.999990672823","0.954849839293","0.184374949232","0.263015918336","0.458637294569","0.151421029768","0.373966392672","0.373964308569","0.000009327177","0.263013611479","0.968019617024","0.997095706553","0.982342555079","0.982342555079"
+"FINNISH_LUCENE_FINNISH_LIGHT_STEM_FILTER","FI_FI","ALL_WORDS","PRIMARY_OUTPUT","57027","1811717","292","1811717","0","1","1811717","439975","12355389","2223150","19168306","1641124591341","2223150","1641126814491","0.000135","19168306","31523695","60.806025","0.847505295284","0.391939745642","0.999998645351","0.999986965635","0.695969195497","0.687649407375","0.535999578676","0.439151826652","0.366119825424","0.576342788507","0.576337821084","0.000013034365","0.535993941880","0.988126027331","0.886473473160","0.934543630400","0.934543630400"
+"FINNISH_LUCENE_FINNISH_LIGHT_STEM_FILTER","FI_FI","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","54762","1757055","274","1757055","0","1","1757055","431848","11988389","1806392","18825444","1543587637760","1806392","1543589444152","0.000117","18825444","30813833","61.094133","0.869052506162","0.389058673746","0.999998829746","0.999986634125","0.694528751746","0.697056446146","0.537492108587","0.437372459518","0.367513988637","0.581474346349","0.581469391800","0.000013365875","0.537486398327","0.989268269625","0.885293761089","0.934397488899","0.934397488899"
+"FINNISH_RADIXOR","FI_FI","ALL_WORDS","PRIMARY_OUTPUT","57027","1811717","292","1811717","0","1","1811717","69091","30552427","731279","971268","1641126083212","731279","1641126814491","0.000045","971268","31523695","3.081073","0.976624284859","0.969189271753","0.999999554404","0.999998962594","0.984594413078","0.975128170336","0.972892573600","0.970667204156","0.947215985975","0.972899675927","0.972899157490","0.000001037406","0.972892054895","0.996084757586","0.993746341306","0.994914175412","0.994914175412"
+"FINNISH_RADIXOR","FI_FI","ALL_WORDS","ANY_CANDIDATE","57027","1811717","292","1754389","57328","6","1876272","69769","31523695","0","0","1641126814491","0","1641126814491","0.000000","0","31523695","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"FINNISH_RADIXOR","FI_FI","ALL_WORDS","ALL_CANDIDATES","57027","1811717","292","1754389","57328","6","1876272","69769","31523695","1683575","0","1641125130916","1683575","1641126814491","0.000103","0","31523695","0.000000","0.949301011495","1.000000000000","0.999998974135","0.999998974154","0.999999487067","0.959025334376","0.973991195713","0.989431554710","0.949301011495","0.974320794962","0.974320295201","0.000001025846","","","","",""
+"FINNISH_RADIXOR","FI_FI","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","54762","1757055","274","1757055","0","1","1757055","54633","30078528","730145","735305","1543588714007","730145","1543589444152","0.000047","735305","30813833","2.386282","0.976300667023","0.976137178390","0.999999526982","0.999999050641","0.988068352686","0.976267964916","0.976218915862","0.976169871736","0.953542638154","0.976218919284","0.976218444595","0.000000949359","0.976218441173","0.996000407428","0.996068984852","0.996034694959","0.996034694959"
+"FINNISH_RADIXOR","FI_FI","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","54762","1757055","274","1712724","44331","6","1805864","54984","30813833","0","0","1543589444152","0","1543589444152","0.000000","0","30813833","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"FINNISH_RADIXOR","FI_FI","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","54762","1757055","274","1712724","44331","6","1805864","54984","30813833","1653320","0","1543587790832","1653320","1543589444152","0.000107","0","30813833","0.000000","0.949077148834","1.000000000000","0.999998928912","0.999998928933","0.999999464456","0.958842548108","0.973873352732","0.989382907677","0.949077148834","0.974205906795","0.974205385065","0.000001071067","","","","",""
+"FRENCH_LUCENE_FRENCH_LIGHT_STEM_FILTER","FR_FR","ALL_WORDS","PRIMARY_OUTPUT","59240","425210","2301","425210","0","1","425210","245918","202782","276403","5251833","90395828427","276403","90396104830","0.000306","5251833","5454615","96.282377","0.423181026117","0.037176226003","0.999996942313","0.999938848002","0.518586584158","0.137547303040","0.068348107452","0.045471618191","0.035383242558","0.125428359900","0.125414592230","0.000061151998","0.068339028277","0.974109647704","0.812375827422","0.885921707253","0.885921707253"
+"FRENCH_LUCENE_FRENCH_LIGHT_STEM_FILTER","FR_FR","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","57698","421231","2133","421231","0","1","421231","245182","200690","262689","5239869","88711863817","262689","88712126506","0.000296","5239869","5440559","96.311225","0.433101197939","0.036887753630","0.999997038860","0.999937976681","0.518442396245","0.137570562409","0.067985131280","0.045148356975","0.035188720533","0.126396717862","0.126383026150","0.000062023319","0.067976159358","0.975085555241","0.811143708698","0.885591261484","0.885591261484"
+"FRENCH_LUCENE_FRENCH_MINIMAL_STEM_FILTER","FR_FR","ALL_WORDS","PRIMARY_OUTPUT","59240","425210","2301","425210","0","1","425210","269236","183612","160438","5271003","90395944392","160438","90396104830","0.000177","5271003","5454615","96.633823","0.533678244441","0.033661770812","0.999998225167","0.999939918724","0.516829997990","0.134399775137","0.063329059361","0.041424008382","0.032699958487","0.134031916915","0.134021061615","0.000060081276","0.063322352769","0.984019125555","0.810978546011","0.889158144694","0.889158144694"
+"FRENCH_LUCENE_FRENCH_MINIMAL_STEM_FILTER","FR_FR","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","57698","421231","2133","421231","0","1","421231","268411","181686","147476","5258873","88711979030","147476","88712126506","0.000166","5258873","5440559","96.660527","0.551965293685","0.033394730211","0.999998337589","0.999939061122","0.516696533900","0.134438681544","0.062979128454","0.041121435592","0.032513396928","0.135767198057","0.135756528540","0.000060938878","0.062972571968","0.985086367216","0.809773733549","0.888868235590","0.888868235590"
+"FRENCH_RADIXOR","FR_FR","ALL_WORDS","PRIMARY_OUTPUT","59240","425210","2301","425210","0","1","425210","60225","4985455","318767","469160","90395786063","318767","90396104830","0.000353","469160","5454615","8.601157","0.939903156391","0.913988429981","0.999996473664","0.999991284144","0.956992451823","0.934603310507","0.926764667966","0.919056419273","0.863524187383","0.926855226151","0.926850879168","0.000008715856","0.926760310630","0.988772235003","0.985214034569","0.986989927876","0.986989927876"
+"FRENCH_RADIXOR","FR_FR","ALL_WORDS","ANY_CANDIDATE","59240","425210","2301","382170","43040","56","477024","60383","5454383","12","232","90396104818","12","90396104830","0.000000","232","5454615","0.004253","0.999997799939","0.999957467209","0.999999999867","0.999999997301","0.999978733538","0.999989733133","0.999977633167","0.999965533495","0.999955267335","0.999977633371","0.999977632021","0.000000002699","","","","",""
+"FRENCH_RADIXOR","FR_FR","ALL_WORDS","ALL_CANDIDATES","59240","425210","2301","382170","43040","56","477024","60383","5454383","1056255","232","90395048575","1056255","90396104830","0.001168","232","5454615","0.004253","0.837764747479","0.999957467209","0.999988315260","0.999988313399","0.999972891234","0.865852951156","0.911703747510","0.962682080453","0.837734895644","0.915275431226","0.915270082203","0.000011686601","","","","",""
+"FRENCH_RADIXOR","FR_FR","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","57698","421231","2133","421231","0","1","421231","58069","4975123","315266","465436","88711811240","315266","88712126506","0.000355","465436","5440559","8.554930","0.940407784758","0.914450702584","0.999996446190","0.999991200142","0.957223574387","0.935099145312","0.927247620620","0.919526848134","0.864363145162","0.927338427699","0.927334038919","0.000008799858","0.927243221287","0.988915897225","0.985549842615","0.987230000708","0.987230000708"
+"FRENCH_RADIXOR","FR_FR","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","57698","421231","2133","380101","41130","56","468574","58208","5440559","0","0","88712126506","0","88712126506","0.000000","0","5440559","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"FRENCH_RADIXOR","FR_FR","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","57698","421231","2133","380101","41130","56","468574","58208","5440559","938985","0","88711187521","938985","88712126506","0.001058","0","5440559","0.000000","0.852813147774","1.000000000000","0.999989415370","0.999989416019","0.999994707685","0.878679151458","0.920560336911","0.966633773699","0.852813147774","0.923478829088","0.923473941734","0.000010583981","","","","",""
+"GERMAN_CISTEM","DE_DE","ALL_WORDS","PRIMARY_OUTPUT","54092","296974","1474","296974","0","1","296974","59097","1053889","477122","329983","44094768857","477122","44095245979","0.001082","329983","1383872","23.844908","0.688361481400","0.761550923785","0.999989179741","0.999981696901","0.880770051763","0.701851885397","0.723108954973","0.745693871888","0.566304351331","0.724031989665","0.724022910459","0.000018303099","0.723099826442","0.974048119240","0.975147027686","0.974597263694","0.974597263694"
+"GERMAN_CISTEM","DE_DE","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","16007","150098","228","150098","0","1","150098","23023","725447","156784","147964","11263599558","156784","11263756342","0.001392","147964","873411","16.940936","0.822286906717","0.830590638313","0.999986080665","0.999972946470","0.915288359489","0.823934343933","0.826417914358","0.828916502414","0.704184159310","0.826428343371","0.826414817348","0.000027053530","0.826404386881","0.985935685912","0.973569821618","0.979713735095","0.979713735095"
+"GERMAN_LUCENE_GERMAN_LIGHT_STEM_FILTER","DE_DE","ALL_WORDS","PRIMARY_OUTPUT","54092","296974","1474","296974","0","1","296974","98357","709263","205740","674609","44095040239","205740","44095245979","0.000467","674609","1383872","48.747933","0.775148278202","0.512520666651","0.999995334191","0.999980035912","0.756258000421","0.703092101246","0.617052253820","0.549774428024","0.446186239158","0.630301128270","0.630292039259","0.000019964088","0.617042686770","0.980753120457","0.936533167951","0.958133203614","0.958133203614"
+"GERMAN_LUCENE_GERMAN_LIGHT_STEM_FILTER","DE_DE","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","16007","150098","228","150098","0","1","150098","50335","471565","55477","401846","11263700865","55477","11263756342","0.000493","401846","873411","46.008809","0.894738939212","0.539911908597","0.999995074734","0.999959401861","0.769953491666","0.790797426464","0.673446377708","0.586423560557","0.507666155661","0.695039717114","0.695022690564","0.000040598139","0.673427319227","0.991320177896","0.915069562866","0.951669957566","0.951669957566"
+"GERMAN_LUCENE_GERMAN_MINIMAL_STEM_FILTER","DE_DE","ALL_WORDS","PRIMARY_OUTPUT","54092","296974","1474","296974","0","1","296974","140505","271626","110840","1112246","44095135139","110840","44095245979","0.000251","1112246","1383872","80.372029","0.710196461908","0.196279713731","0.999997486350","0.999972263504","0.598138600041","0.466112921692","0.307558349534","0.229493166050","0.181724639931","0.373359288402","0.373350267608","0.000027736496","0.307548938689","0.983615403456","0.896263607272","0.937910029995","0.937910029995"
+"GERMAN_LUCENE_GERMAN_MINIMAL_STEM_FILTER","DE_DE","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","16007","150098","228","150098","0","1","150098","80363","132221","21214","741190","11263735128","21214","11263756342","0.000188","741190","873411","84.861537","0.861739498811","0.151384628772","0.999998116614","0.999932318770","0.575691372693","0.444544636019","0.257528392768","0.181269722976","0.147794886125","0.361184321539","0.361168285320","0.000067681230","0.257511188319","0.992642524078","0.854402700840","0.918349417869","0.918349417869"
+"GERMAN_LUCENE_GERMAN_STEM_FILTER","DE_DE","ALL_WORDS","PRIMARY_OUTPUT","54092","296974","1474","296974","0","1","296974","81085","619354","331871","764518","44094914108","331871","44095245979","0.000753","764518","1383872","55.244849","0.651111987174","0.447551507654","0.999992473769","0.999975136671","0.723771990712","0.596821367368","0.530473894660","0.477402037056","0.360982967729","0.539820480819","0.539808754751","0.000024863329","0.530461889452","0.975549631706","0.942889706548","0.958941664321","0.958941664321"
+"GERMAN_LUCENE_GERMAN_STEM_FILTER","DE_DE","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","16007","150098","228","150098","0","1","150098","41574","378734","78723","494677","11263677619","78723","11263756342","0.000699","494677","873411","56.637368","0.827911694432","0.433626322545","0.999993010946","0.999949097306","0.716809666745","0.700518896035","0.569153364571","0.479276535831","0.397773842757","0.599169678345","0.599148958371","0.000050902694","0.569130398173","0.988583531594","0.918717729459","0.952371013508","0.952371013508"
+"GERMAN_RADIXOR","DE_DE","ALL_WORDS","PRIMARY_OUTPUT","54092","296974","1474","296974","0","1","296974","68104","1128969","98192","254903","44095147787","98192","44095245979","0.000223","254903","1383872","18.419550","0.919984419322","0.815804496370","0.999997773184","0.999991992699","0.907901134777","0.897072808397","0.864768082211","0.834709150216","0.761754553110","0.866329859738","0.866325955582","0.000008007301","0.864764092865","0.989946248415","0.975085217969","0.982459538105","0.982459538105"
+"GERMAN_RADIXOR","DE_DE","ALL_WORDS","ANY_CANDIDATE","54092","296974","1474","248400","48574","8","361016","70717","1272705","1375","111167","44095244604","1375","44095245979","0.000003","111167","1383872","8.033041","0.998920789903","0.919669593720","0.999999968818","0.999997447832","0.959834781269","0.981996366774","0.957658377578","0.934497606897","0.918756727140","0.958476435291","0.958475209548","0.000002552168","","","","",""
+"GERMAN_RADIXOR","DE_DE","ALL_WORDS","ALL_CANDIDATES","54092","296974","1474","248400","48574","8","361016","70717","1272705","244817","111167","44095001162","244817","44095245979","0.000555","111167","1383872","8.033041","0.838673179038","0.919669593720","0.999994447996","0.999991927184","0.959832020858","0.853710645080","0.877305874349","0.902242446842","0.781429112618","0.878238135035","0.878234164088","0.000008072816","","","","",""
+"GERMAN_RADIXOR","DE_DE","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","16007","150098","228","150098","0","1","150098","17264","814297","47898","59114","11263708444","47898","11263756342","0.000425","59114","873411","6.768177","0.944446441930","0.932318232768","0.999995747600","0.999990500176","0.966156990184","0.941995622128","0.938343149309","0.934718891125","0.883847872972","0.938362743125","0.938357995965","0.000009499824","0.938338399230","0.994062310308","0.990664418294","0.992360455671","0.992360455671"
+"GERMAN_RADIXOR","DE_DE","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","16007","150098","228","135120","14978","8","167157","18366","873411","0","0","11263756342","0","11263756342","0.000000","0","873411","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"GERMAN_RADIXOR","DE_DE","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","16007","150098","228","135120","14978","8","167157","18366","873411","97544","0","11263658798","97544","11263756342","0.000866","0","873411","0.000000","0.899538083639","1.000000000000","0.999991340012","0.999991340683","0.999995670006","0.917982540684","0.947112449481","0.978151677228","0.899538083639","0.948439815507","0.948435708759","0.000008659317","","","","",""
+"HE_IL_RADIXOR","HE_IL","ALL_WORDS","PRIMARY_OUTPUT","2358","58714","0","58714","0","1","58714","2358","688361","25234","19916","1722904030","25234","1722929264","0.001465","19916","708277","2.811894","0.964638205144","0.971881057835","0.999985354013","0.999973805398","0.985933205924","0.966078126522","0.968246086849","0.970423799230","0.938446730860","0.968252859146","0.968239762122","0.000026194602","0.968232984327","0.993166390361","0.993627509527","0.993396896433","0.993396896433"
+"HE_IL_RADIXOR","HE_IL","ALL_WORDS","ANY_CANDIDATE","2358","58714","0","56674","2040","40","62376","2358","708277","0","0","1722929264","0","1722929264","0.000000","0","708277","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"HE_IL_RADIXOR","HE_IL","ALL_WORDS","ALL_CANDIDATES","2358","58714","0","56674","2040","40","62376","2358","708277","86628","0","1722842636","86628","1722929264","0.005028","0","708277","0.000000","0.891020939609","1.000000000000","0.999949720513","0.999949741174","0.999974860256","0.910874182109","0.942370251906","0.976122467036","0.891020939609","0.943939055029","0.943915324345","0.000050258826","","","","",""
+"HE_IL_RADIXOR","HE_IL","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","2358","58714","0","58714","0","1","58714","2358","688361","25234","19916","1722904030","25234","1722929264","0.001465","19916","708277","2.811894","0.964638205144","0.971881057835","0.999985354013","0.999973805398","0.985933205924","0.966078126522","0.968246086849","0.970423799230","0.938446730860","0.968252859146","0.968239762122","0.000026194602","0.968232984327","0.993166390361","0.993627509527","0.993396896433","0.993396896433"
+"HE_IL_RADIXOR","HE_IL","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","2358","58714","0","56674","2040","40","62376","2358","708277","0","0","1722929264","0","1722929264","0.000000","0","708277","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"HE_IL_RADIXOR","HE_IL","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","2358","58714","0","56674","2040","40","62376","2358","708277","86628","0","1722842636","86628","1722929264","0.005028","0","708277","0.000000","0.891020939609","1.000000000000","0.999949720513","0.999949741174","0.999974860256","0.910874182109","0.942370251906","0.976122467036","0.891020939609","0.943939055029","0.943915324345","0.000050258826","","","","",""
+"HUNGARIAN_LUCENE_HUNGARIAN_LIGHT_STEM_FILTER","HU_HU","ALL_WORDS","PRIMARY_OUTPUT","19406","916344","1","916344","0","1","916344","94328","14036270","4132555","8125833","419816410338","4132555","419820542893","0.000984","8125833","22162103","36.665442","0.772546931351","0.633345580968","0.999990156377","0.999970802427","0.816667868673","0.740017627855","0.696054898613","0.657022705053","0.533806904809","0.699492090778","0.699477892426","0.000029197573","0.696040442258","0.982615378770","0.926771756762","0.953876941828","0.953876941828"
+"HUNGARIAN_LUCENE_HUNGARIAN_LIGHT_STEM_FILTER","HU_HU","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","18360","878513","1","878513","0","1","878513","91516","13492703","3639046","7918708","385867055871","3639046","385870694917","0.000943","7918708","21411411","36.983588","0.787584676848","0.630164121365","0.999990569261","0.999970049260","0.815077345313","0.750107959995","0.700134758022","0.656404225003","0.538621031944","0.704491026122","0.704476576404","0.000029950740","0.700119966552","0.983686881315","0.925487153321","0.953699929950","0.953699929950"
+"HUNGARIAN_RADIXOR","HU_HU","ALL_WORDS","PRIMARY_OUTPUT","19406","916344","1","916344","0","1","916344","20535","21962266","272900","199837","419820269993","272900","419820542893","0.000065","199837","22162103","0.901706","0.987726648859","0.990982940563","0.999999349960","0.999998874014","0.995491145262","0.988376194087","0.989352115329","0.990329965723","0.978928596533","0.989353455019","0.989352892139","0.000001125986","0.989351552308","0.998036093538","0.997808712909","0.997922390271","0.997922390271"
+"HUNGARIAN_RADIXOR","HU_HU","ALL_WORDS","ANY_CANDIDATE","19406","916344","1","904024","12320","5","929326","20567","22162103","0","0","419820542893","0","419820542893","0.000000","0","22162103","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"HUNGARIAN_RADIXOR","HU_HU","ALL_WORDS","ALL_CANDIDATES","19406","916344","1","904024","12320","5","929326","20567","22162103","460158","0","419820082735","460158","419820542893","0.000110","0","22162103","0.000000","0.979659062372","1.000000000000","0.999998903917","0.999998903975","0.999999451959","0.983660778882","0.989725029923","0.995864516790","0.979659062372","0.989777279176","0.989776736737","0.000001096025","","","","",""
+"HUNGARIAN_RADIXOR","HU_HU","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","18360","878513","1","878513","0","1","878513","18363","21247134","272775","164277","385870422142","272775","385870694917","0.000071","164277","21411411","0.767240","0.987324528185","0.992327595785","0.999999293092","0.999998867424","0.996163444439","0.988321101757","0.989819739994","0.991322930046","0.979844666523","0.989822900984","0.989822335019","0.000001132576","0.989819173678","0.997945135090","0.998273386381","0.998109233747","0.998109233747"
+"HUNGARIAN_RADIXOR","HU_HU","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","18360","878513","1","867360","11153","5","890245","18375","21411411","0","0","385870694917","0","385870694917","0.000000","0","21411411","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"HUNGARIAN_RADIXOR","HU_HU","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","18360","878513","1","867360","11153","5","890245","18375","21411411","458462","0","385870236455","458462","385870694917","0.000119","0","21411411","0.000000","0.979036823854","1.000000000000","0.999998811877","0.999998811943","0.999999405938","0.983158850285","0.989407384494","0.995735852714","0.979036823854","0.989462896653","0.989462308851","0.000001188057","","","","",""
+"HUNSPELL_CZECH_LUCENE_FILTER","CS_CZ","ALL_WORDS","PRIMARY_OUTPUT","5113","51676","2","51676","0","1","51676","10920","213552","11408","88283","1334865407","11408","1334876815","0.000855","88283","301835","29.248762","0.949288762447","0.707512382593","0.999991453893","0.999925335085","0.853751918243","0.888559718726","0.810759403563","0.745486280807","0.681745481942","0.819532521678","0.819499025505","0.000074664915","0.810722859062","0.995776551361","0.952852006662","0.973841506371","0.973841506371"
+"HUNSPELL_CZECH_LUCENE_FILTER","CS_CZ","ALL_WORDS","ANY_CANDIDATE","5113","51676","2","48359","3317","5","55596","11359","224312","10102","77523","1334866713","10102","1334876815","0.000757","77523","301835","25.683900","0.956905304291","0.743160998559","0.999992432261","0.999934372078","0.871576715410","0.904855299474","0.836596431881","0.777913569166","0.719093919606","0.843288029954","0.843258147533","0.000065627922","","","","",""
+"HUNSPELL_CZECH_LUCENE_FILTER","CS_CZ","ALL_WORDS","ALL_CANDIDATES","5113","51676","2","48359","3317","5","55596","11359","224312","13917","77523","1334862898","13917","1334876815","0.001043","77523","301835","25.683900","0.941581419558","0.743160998559","0.999989574319","0.999931514783","0.871575286439","0.893850652440","0.830686733424","0.775860578084","0.710405634802","0.836508570179","0.836476906392","0.000068485217","","","","",""
+"HUNSPELL_CZECH_LUCENE_FILTER","CS_CZ","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","5038","50968","2","50968","0","1","50968","10816","210827","11239","87986","1298532976","11239","1298544215","0.000866","87986","298813","29.445171","0.949388920411","0.705548286052","0.999991344923","0.999923605087","0.852769815487","0.888008949714","0.809504702628","0.743753342581","0.679973036781","0.818437368155","0.818403143837","0.000076394913","0.809467327086","0.995812473772","0.952393750423","0.973619286126","0.973619286126"
+"HUNSPELL_CZECH_LUCENE_FILTER","CS_CZ","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","5038","50968","2","47731","3237","5","54804","11240","221382","10028","77431","1298534187","10028","1298544215","0.000772","77431","298813","25.912862","0.956665658355","0.740871381098","0.999992277506","0.999932663918","0.870431829302","0.904003665310","0.835052421340","0.775874033233","0.716815448726","0.841882537861","0.841851913927","0.000067336082","","","","",""
+"HUNSPELL_CZECH_LUCENE_FILTER","CS_CZ","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","5038","50968","2","47731","3237","5","54804","11240","221382","13601","77431","1298530614","13601","1298544215","0.001047","77431","298813","25.912862","0.942119217135","0.740871381098","0.999989525963","0.999929913009","0.870430453530","0.893573737936","0.829462940899","0.773935751817","0.708617411512","0.835457458856","0.835425115185","0.000070086991","","","","",""
+"HUNSPELL_DUTCH_LUCENE_FILTER","NL_NL","ALL_WORDS","PRIMARY_OUTPUT","4992","26477","85","26477","0","1","26477","15909","18482","1333","46084","350436627","1333","350437960","0.000380","46084","64566","71.375027","0.932727731517","0.286249728960","0.999996196188","0.999864717095","0.643122962574","0.642512480358","0.438060700869","0.332315636923","0.280459491039","0.516713712165","0.516673857221","0.000135282905","0.438012080403","0.996931617211","0.889026094124","0.939891935467","0.939891935467"
+"HUNSPELL_DUTCH_LUCENE_FILTER","NL_NL","ALL_WORDS","ANY_CANDIDATE","4992","26477","85","25223","1254","3","27763","16027","21374","1164","43192","350436796","1164","350437960","0.000332","43192","64566","66.895889","0.948353891206","0.331041105226","0.999996678442","0.999873450270","0.665518891834","0.690740573172","0.490769654666","0.380588457347","0.325178761600","0.560307166017","0.560267948638","0.000126549730","","","","",""
+"HUNSPELL_DUTCH_LUCENE_FILTER","NL_NL","ALL_WORDS","ALL_CANDIDATES","4992","26477","85","25223","1254","3","27763","16027","21374","1738","43192","350436222","1738","350437960","0.000496","43192","64566","66.895889","0.924800969193","0.331041105226","0.999995040492","0.999871812621","0.665518072859","0.680639942935","0.487556741714","0.379812066416","0.322363658301","0.553305643343","0.553264631551","0.000128187379","","","","",""
+"HUNSPELL_DUTCH_LUCENE_FILTER","NL_NL","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4796","25678","84","25678","0","1","25678","15258","18333","1310","44814","329602546","1310","329603856","0.000397","44814","63147","70.967742","0.933309575930","0.290322580645","0.999996025532","0.999860089122","0.645159303089","0.646808120294","0.442879574828","0.336717714000","0.284422172921","0.520538994337","0.520497519470","0.000139910878","0.442828931093","0.996884174988","0.889060613994","0.939890140871","0.939890140871"
+"HUNSPELL_DUTCH_LUCENE_FILTER","NL_NL","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","4796","25678","84","24492","1186","3","26896","15323","21212","1141","41935","329602715","1141","329603856","0.000346","41935","63147","66.408539","0.948955397486","0.335914611937","0.999996538269","0.999869334815","0.667955575103","0.695206444720","0.496187134503","0.385755489360","0.329952712792","0.564595416287","0.564554662442","0.000130665185","","","","",""
+"HUNSPELL_DUTCH_LUCENE_FILTER","NL_NL","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","4796","25678","84","24492","1186","3","26896","15323","21212","1712","41935","329602144","1712","329603856","0.000519","41935","63147","66.408539","0.925318443553","0.335914611937","0.999994805886","0.999867602764","0.667954708912","0.684951854459","0.492895400309","0.384956009176","0.327047903915","0.557519493726","0.557476858392","0.000132397236","","","","",""
+"HUNSPELL_ENGLISH_LUCENE_FILTER","US_UK","ALL_WORDS","PRIMARY_OUTPUT","396939","607439","250964","607439","0","1","607439","557518","46002","1981986","267868","184488469785","1981986","184490451771","0.001074","267868","313870","85.343614","0.022683566175","0.146563864020","0.999989256972","0.999987805059","0.573276560496","0.027298226808","0.039286754363","0.070050933952","0.020036970960","0.057659267324","0.057655308782","0.000012194941","0.039283923590","0.993096189204","0.963677172906","0.978165531652","0.978165531652"
+"HUNSPELL_ENGLISH_LUCENE_FILTER","US_UK","ALL_WORDS","ANY_CANDIDATE","396939","607439","250964","600602","6837","4","614296","557638","51229","1978852","262641","184488472919","1978852","184490451771","0.001073","262641","313870","83.678274","0.025234953679","0.163217255552","0.999989273960","0.999987850378","0.581603264756","0.030369825498","0.043711664621","0.077960810954","0.022344183028","0.064177721084","0.064173802046","0.000012149622","","","","",""
+"HUNSPELL_ENGLISH_LUCENE_FILTER","US_UK","ALL_WORDS","ALL_CANDIDATES","396939","607439","250964","600602","6837","4","614296","557638","51229","2008917","262641","184488442854","2008917","184490451771","0.001089","262641","313870","83.678274","0.024866684206","0.163217255552","0.999989110997","0.999987687416","0.581603183275","0.029942881217","0.043158091605","0.077253888104","0.022054971033","0.063707707153","0.063703758400","0.000012312584","","","","",""
+"HUNSPELL_ENGLISH_LUCENE_FILTER","US_UK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","374384","583910","228735","583910","0","1","583910","535362","45926","1978041","265965","170472862163","1978041","170474840204","0.001160","265965","311891","85.274984","0.022691081426","0.147250161114","0.999988396874","0.999986836756","0.573619278994","0.027311677226","0.039322595808","0.070190378755","0.020055617372","0.057803679777","0.057799415161","0.000013163244","0.039319549964","0.993066314983","0.962316521867","0.977449637185","0.977449637185"
+"HUNSPELL_ENGLISH_LUCENE_FILTER","US_UK","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","374384","583910","228735","577124","6786","4","590716","535485","51150","1974950","260741","170472865254","1974950","170474840204","0.001158","260741","311891","83.600040","0.025245545630","0.163999602425","0.999988415006","0.999986885532","0.581994008716","0.030387494919","0.043755514884","0.078123472659","0.022367099418","0.064344847861","0.064340626006","0.000013114468","","","","",""
+"HUNSPELL_ENGLISH_LUCENE_FILTER","US_UK","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","374384","583910","228735","577124","6786","4","590716","535485","51150","2004598","260741","170472835606","2004598","170474840204","0.001176","260741","311891","83.600040","0.024881454342","0.163999602425","0.999988241092","0.999986711618","0.581993921758","0.029965261387","0.043207600483","0.077422296168","0.022080830084","0.063879172034","0.063874918547","0.000013288382","","","","",""
+"HUNSPELL_FRENCH_LUCENE_FILTER","FR_FR","ALL_WORDS","PRIMARY_OUTPUT","59240","425210","2301","425210","0","1","425210","154336","3422734","776728","2031881","90395328102","776728","90396104830","0.000859","2031881","5454615","37.250677","0.815041069547","0.627493232795","0.999991407506","0.999968931852","0.813742320150","0.769068574566","0.709075347131","0.657764674673","0.549277098051","0.715145268872","0.715130511338","0.000031068148","0.709060074832","0.978337247291","0.913705954219","0.944917713687","0.944917713687"
+"HUNSPELL_FRENCH_LUCENE_FILTER","FR_FR","ALL_WORDS","ANY_CANDIDATE","59240","425210","2301","411699","13511","4","439015","154718","3610612","745831","1844003","90395358999","745831","90396104830","0.000825","1844003","5454615","33.806291","0.828798173189","0.661937093635","0.999991749302","0.999971351888","0.830964421468","0.789018996925","0.736029080656","0.689708764155","0.582314885091","0.740683639600","0.740669908387","0.000028648112","","","","",""
+"HUNSPELL_FRENCH_LUCENE_FILTER","FR_FR","ALL_WORDS","ALL_CANDIDATES","59240","425210","2301","411699","13511","4","439015","154718","3610612","1043199","1844003","90395061631","1043199","90396104830","0.001154","1844003","5454615","33.806291","0.775839843947","0.661937093635","0.999988459691","0.999968062476","0.830962776663","0.750027659074","0.714376699201","0.681961135862","0.555665643861","0.716629033342","0.716613365354","0.000031937524","","","","",""
+"HUNSPELL_FRENCH_LUCENE_FILTER","FR_FR","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","57698","421231","2133","421231","0","1","421231","153822","3412548","763305","2028011","88711363201","763305","88712126506","0.000860","2028011","5440559","37.275784","0.817209801207","0.627242163903","0.999991395708","0.999968537054","0.813616779806","0.770536594362","0.709734150326","0.657825640123","0.550068151075","0.715952822518","0.715937898033","0.000031462946","0.709718690125","0.979328164393","0.913161860024","0.945088340370","0.945088340370"
+"HUNSPELL_FRENCH_LUCENE_FILTER","FR_FR","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","57698","421231","2133","407794","13437","4","434961","154205","3600083","733584","1840476","88711392922","733584","88712126506","0.000827","1840476","5440559","33.828803","0.830724418835","0.661711967465","0.999991730736","0.999970985904","0.830851849101","0.790350629656","0.736648201095","0.689779349655","0.583090317150","0.741417756470","0.741403865778","0.000029014096","","","","",""
+"HUNSPELL_FRENCH_LUCENE_FILTER","FR_FR","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","57698","421231","2133","407794","13437","4","434961","154205","3600083","1027635","1840476","88711098871","1027635","88712126506","0.001158","1840476","5440559","33.828803","0.777939148410","0.661711967465","0.999988416071","0.999967671442","0.830850191768","0.751538185756","0.715133880405","0.682093458746","0.556582409247","0.717475884237","0.717460037184","0.000032328558","","","","",""
+"HUNSPELL_GERMAN_LUCENE_FILTER","DE_DE","ALL_WORDS","PRIMARY_OUTPUT","54092","296974","1474","296974","0","1","296974","182774","391862","203883","992010","44095042096","203883","44095245979","0.000462","992010","1383872","71.683653","0.657768004767","0.283163471766","0.999995376304","0.999972880172","0.641579424035","0.520145203475","0.395896782054","0.319562150060","0.246802560848","0.431573715426","0.431562811676","0.000027119828","0.395885371175","0.980462638581","0.886872825924","0.931322397630","0.931322397630"
+"HUNSPELL_GERMAN_LUCENE_FILTER","DE_DE","ALL_WORDS","ANY_CANDIDATE","54092","296974","1474","289083","7891","3","305052","183111","408175","158403","975697","44095087576","158403","44095245979","0.000359","975697","1383872","70.504859","0.720421548313","0.294951411691","0.999996407708","0.999974281481","0.647473909700","0.559115650060","0.418544438463","0.334456395588","0.264657729653","0.460965674088","0.460955788036","0.000025718519","","","","",""
+"HUNSPELL_GERMAN_LUCENE_FILTER","DE_DE","ALL_WORDS","ALL_CANDIDATES","54092","296974","1474","289083","7891","3","305052","183111","408175","242551","975697","44095003428","242551","44095245979","0.000550","975697","1383872","70.504859","0.627260936247","0.294951411691","0.999994499384","0.999972373218","0.647472955538","0.511911128190","0.401234052132","0.329906951166","0.250964847398","0.430129630047","0.430118032816","0.000027626782","","","","",""
+"HUNSPELL_GERMAN_LUCENE_FILTER","DE_DE","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","16007","150098","228","150098","0","1","150098","86983","278093","84679","595318","11263671663","84679","11263756342","0.000752","595318","873411","68.160122","0.766577905682","0.318398783620","0.999992482170","0.999939634323","0.659195632895","0.598178360154","0.449922058465","0.360558871242","0.290257700216","0.494041974653","0.494019111671","0.000060365677","0.449897024669","0.988041339480","0.865580709092","0.922765807515","0.922765807515"
+"HUNSPELL_GERMAN_LUCENE_FILTER","DE_DE","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","16007","150098","228","145109","4989","3","155207","87393","288864","60996","584547","11263695346","60996","11263756342","0.000542","584547","873411","66.926911","0.825655976676","0.330730893016","0.999994584755","0.999942692923","0.665362738886","0.635466205220","0.472281285177","0.375782098835","0.309141519702","0.522560942370","0.522540242219","0.000057307077","","","","",""
+"HUNSPELL_GERMAN_LUCENE_FILTER","DE_DE","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","16007","150098","228","145109","4989","3","155207","87393","288864","96545","584547","11263659797","96545","11263756342","0.000857","584547","873411","66.926911","0.749499881944","0.330730893016","0.999991428703","0.999939537116","0.665361160860","0.598050472724","0.458944090497","0.372338300095","0.297811447117","0.497878263505","0.497854575726","0.000060462884","","","","",""
+"HUNSPELL_POLISH_LUCENE_FILTER","PL_PL","ALL_WORDS","PRIMARY_OUTPUT","9990","122341","1","122341","0","1","122341","18419","971262","52652","149705","7482425351","52652","7482478003","0.000704","149705","1120967","13.354987","0.948577712581","0.866450127435","0.999992963294","0.999972959935","0.933221545364","0.930929837176","0.905655838249","0.881717903868","0.827578626454","0.906584403102","0.906571161039","0.000027040065","0.905642343969","0.994545991966","0.970520439141","0.982386343372","0.982386343372"
+"HUNSPELL_POLISH_LUCENE_FILTER","PL_PL","ALL_WORDS","ANY_CANDIDATE","9990","122341","1","110894","11447","6","135231","19068","1040224","42213","80743","7482435790","42213","7482478003","0.000564","80743","1120967","7.202977","0.961001887408","0.927970225707","0.999994358420","0.999983569937","0.963982292063","0.954208759768","0.944197251162","0.934393641743","0.894293230626","0.944341642819","0.944333470354","0.000016430063","","","","",""
+"HUNSPELL_POLISH_LUCENE_FILTER","PL_PL","ALL_WORDS","ALL_CANDIDATES","9990","122341","1","110894","11447","6","135231","19068","1040224","82745","80743","7482395258","82745","7482478003","0.001106","80743","1120967","7.202977","0.926315864463","0.927970225707","0.999988941498","0.999978153827","0.963979583602","0.926646264647","0.927142307089","0.927638880888","0.864180136112","0.927142676087","0.927131751477","0.000021846173","","","","",""
+"HUNSPELL_POLISH_LUCENE_FILTER","PL_PL","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","9846","120925","1","120925","0","1","120925","18149","965984","51950","148667","7310200749","51950","7310252699","0.000711","148667","1114651","13.337538","0.948965257080","0.866624620621","0.999992893543","0.999972560946","0.933308757082","0.931268723294","0.905927782480","0.881929423296","0.828032892137","0.906860880124","0.906847444801","0.000027439054","0.905914089179","0.994583905165","0.970514203019","0.982401644006","0.982401644006"
+"HUNSPELL_POLISH_LUCENE_FILTER","PL_PL","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","9846","120925","1","109660","11265","6","133595","18789","1034283","41671","80368","7310211028","41671","7310252699","0.000570","80368","1114651","7.210149","0.961270649117","0.927898508143","0.999994299650","0.999983308321","0.963946403896","0.954405554191","0.944289819479","0.934386268967","0.894459328803","0.944437187555","0.944428885928","0.000016691679","","","","",""
+"HUNSPELL_POLISH_LUCENE_FILTER","PL_PL","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","9846","120925","1","109660","11265","6","133595","18789","1034283","81865","80368","7310170834","81865","7310252699","0.001120","80368","1114651","7.210149","0.926653992123","0.927898508143","0.999988801345","0.999977810854","0.963943654744","0.926902628188","0.927275832560","0.927649337585","0.864412176686","0.927276041347","0.927264945147","0.000022189146","","","","",""
+"HUNSPELL_SPANISH_LUCENE_FILTER","ES_ES","ALL_WORDS","PRIMARY_OUTPUT","65059","871332","3589","871332","0","1","871332","495840","9662476","536192","32310860","379566781918","536192","379567318110","0.000141","32310860","41973336","76.979490","0.947425291224","0.230205099733","0.999998587360","0.999913471422","0.615101843546","0.583708381625","0.370408466579","0.271277635967","0.227301418167","0.467014061518","0.466991649518","0.000086528578","0.370381248953","0.993314263125","0.790558492734","0.880413722434","0.880413722434"
+"HUNSPELL_SPANISH_LUCENE_FILTER","ES_ES","ALL_WORDS","ANY_CANDIDATE","65059","871332","3589","853455","17877","5","890999","496361","10079118","416345","31894218","379566901765","416345","379567318110","0.000110","31894218","41973336","75.986855","0.960330954432","0.240131449166","0.999998903106","0.999914884689","0.620065176136","0.600267728541","0.384194728757","0.282504215637","0.237772914592","0.480214185303","0.480192080762","0.000085115311","","","","",""
+"HUNSPELL_SPANISH_LUCENE_FILTER","ES_ES","ALL_WORDS","ALL_CANDIDATES","65059","871332","3589","853455","17877","5","890999","496361","10079118","888077","31894218","379566430033","888077","379567318110","0.000234","31894218","41973336","75.986855","0.919024235459","0.240131449166","0.999997660291","0.999913642011","0.620064554728","0.587073016700","0.380771322449","0.281759130783","0.235155989841","0.469772946730","0.469749183448","0.000086357989","","","","",""
+"HUNSPELL_SPANISH_LUCENE_FILTER","ES_ES","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","64918","869371","3525","869371","0","1","869371","495045","9628515","531181","32234855","377860138584","531181","377860669765","0.000141","32234855","41863370","77.000144","0.947716841134","0.229998564377","0.999998594241","0.999913295008","0.614998579309","0.583531128169","0.370163304101","0.271052948234","0.227116805648","0.466876335765","0.466853897372","0.000086704992","0.370136051001","0.993362468962","0.790499503024","0.880396073652","0.880396073652"
+"HUNSPELL_SPANISH_LUCENE_FILTER","ES_ES","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","64918","869371","3525","851564","17807","5","888962","495572","10041262","412198","31822108","377860257567","412198","377860669765","0.000109","31822108","41863370","76.014205","0.960568271175","0.239857947413","0.999998909127","0.999914702064","0.619928428270","0.599999808789","0.383863548307","0.282205460900","0.237519268813","0.479999931119","0.479977799265","0.000085297936","","","","",""
+"HUNSPELL_SPANISH_LUCENE_FILTER","ES_ES","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","64918","869371","3525","851564","17807","5","888962","495572","10041262","878949","31822108","377859790816","878949","377860669765","0.000233","31822108","41863370","76.014205","0.919511720057","0.239857947413","0.999997673881","0.999913466955","0.619927810647","0.586904802235","0.380469146267","0.281467012980","0.234925531298","0.469629847641","0.469606065497","0.000086533045","","","","",""
+"HUNSPELL_UKRAINIAN_LUCENE_FILTER","UK_UA","ALL_WORDS","PRIMARY_OUTPUT","1493","14245","4","14245","0","1","14245","3137","50416","794","14924","101386756","794","101387550","0.000783","14924","65340","22.840526","0.984495215778","0.771594735231","0.999992168664","0.999845070949","0.885793451947","0.933007624547","0.865139425139","0.806475349522","0.762331024889","0.871568313648","0.871498740996","0.000154929051","0.865063055969","0.998114340300","0.949803904722","0.973360047526","0.973360047526"
+"HUNSPELL_UKRAINIAN_LUCENE_FILTER","UK_UA","ALL_WORDS","ANY_CANDIDATE","1493","14245","4","12923","1322","6","15740","3311","55875","326","9465","101387224","326","101387550","0.000322","9465","65340","14.485767","0.994199391470","0.855142332415","0.999996784615","0.999903492153","0.927569558515","0.962883947281","0.919442821764","0.879752236578","0.850896963421","0.922053136488","0.922008115880","0.000096507847","","","","",""
+"HUNSPELL_UKRAINIAN_LUCENE_FILTER","UK_UA","ALL_WORDS","ALL_CANDIDATES","1493","14245","4","12923","1322","6","15740","3311","55875","1271","9465","101386279","1271","101387550","0.001254","9465","65340","14.485767","0.977758723270","0.855142332415","0.999987463944","0.999894177485","0.927564898180","0.950500809733","0.912349166435","0.877142031861","0.838825419225","0.914397547654","0.914347195556","0.000105822515","","","","",""
+"HUNSPELL_UKRAINIAN_LUCENE_FILTER","UK_UA","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","1491","14236","4","14236","0","1","14236","3134","50404","794","14920","101258612","794","101259406","0.000784","14920","65324","22.839998","0.984491581702","0.771600024493","0.999992158753","0.999844914465","0.885796091623","0.933006560145","0.865141346698","0.806479484406","0.762334008893","0.871569692311","0.871500048785","0.000155085535","0.865064900251","0.998112856419","0.949788155938","0.973351072054","0.973351072054"
+"HUNSPELL_UKRAINIAN_LUCENE_FILTER","UK_UA","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","1491","14236","4","12915","1321","6","15730","3308","55859","326","9465","101259080","326","101259406","0.000322","9465","65324","14.489315","0.994197739610","0.855106851999","0.999996780546","0.999903370085","0.927551816273","0.962873710629","0.919421606630","0.879721936116","0.850860624524","0.922033242016","0.921988165267","0.000096629915","","","","",""
+"HUNSPELL_UKRAINIAN_LUCENE_FILTER","UK_UA","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","1491","14236","4","12915","1321","6","15730","3308","55859","1271","9465","101258135","1271","101259406","0.001255","9465","65324","14.489315","0.977752494311","0.855106851999","0.999987448080","0.999894043636","0.927547150039","0.950487333415","0.912326261290","0.877111165546","0.838786695698","0.914375665383","0.914325250217","0.000105956364","","","","",""
+"ITALIAN_LUCENE_ITALIAN_LIGHT_STEM_FILTER","IT_IT","ALL_WORDS","PRIMARY_OUTPUT","10009","327551","0","327551","0","1","327551","244870","109684","10589","6034130","53638510622","10589","53638521211","0.000020","6034130","6143814","98.214725","0.911958627456","0.017852754006","0.999999802586","0.999887319289","0.508926278296","0.082781551919","0.035019947839","0.022207258650","0.017822037328","0.127596916262","0.127588341500","0.000112680711","0.035015703871","0.997481424185","0.737537113266","0.848036553212","0.848036553212"
+"ITALIAN_LUCENE_ITALIAN_LIGHT_STEM_FILTER","IT_IT","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","10007","327469","0","327469","0","1","327469","244808","109658","10588","6032516","53611656484","10588","53611667072","0.000020","6032516","6142174","98.214671","0.911947174958","0.017853287777","0.999999802506","0.999887292971","0.508926545141","0.082783771729","0.035020966336","0.022207918023","0.017822564890","0.127598022525","0.127589445488","0.000112707029","0.035016721203","0.997481197125","0.737534145120","0.848034509070","0.848034509070"
+"ITALIAN_RADIXOR","IT_IT","ALL_WORDS","PRIMARY_OUTPUT","10009","327551","0","327551","0","1","327551","10010","6100906","124172","42908","53638397039","124172","53638521211","0.000231","42908","6143814","0.698394","0.980052940702","0.993016064614","0.999997685022","0.999996885431","0.996506874818","0.982618418699","0.986491918597","0.990396078172","0.973343909830","0.986513210398","0.986511657877","0.000003114569","0.986490361200","0.995780270704","0.997112550353","0.996445965204","0.996445965204"
+"ITALIAN_RADIXOR","IT_IT","ALL_WORDS","ANY_CANDIDATE","10009","327551","0","321297","6254","4","334175","10012","6143734","0","80","53638521211","0","53638521211","0.000000","80","6143814","0.001302","1.000000000000","0.999986978772","1.000000000000","0.999999998509","0.999993489386","0.999997395727","0.999993489344","0.999989582991","0.999986978772","0.999993489365","0.999993488619","0.000000001491","","","","",""
+"ITALIAN_RADIXOR","IT_IT","ALL_WORDS","ALL_CANDIDATES","10009","327551","0","321297","6254","4","334175","10012","6143734","170950","80","53638350261","170950","53638521211","0.000319","80","6143814","0.001302","0.972928178195","0.999986978772","0.999996812925","0.999996811799","0.999991895849","0.978222150749","0.986272020913","0.994455476443","0.972915852437","0.986364795335","0.986363222748","0.000003188201","","","","",""
+"ITALIAN_RADIXOR","IT_IT","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","10007","327469","0","327469","0","1","327469","10007","6099346","124171","42828","53611542901","124171","53611667072","0.000232","42828","6142174","0.697278","0.980048098206","0.993027224563","0.999997683881","0.999996885382","0.996512454222","0.982616709845","0.986494972258","0.990403969991","0.973349855458","0.986516316590","0.986514764059","0.000003114618","0.986493414837","0.995779575755","0.997114909855","0.996446795437","0.996446795437"
+"ITALIAN_RADIXOR","IT_IT","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","10007","327469","0","321217","6252","4","334089","10007","6142174","0","0","53611667072","0","53611667072","0.000000","0","6142174","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"ITALIAN_RADIXOR","IT_IT","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","10007","327469","0","321217","6252","4","334089","10007","6142174","170949","0","53611496123","170949","53611667072","0.000319","0","6142174","0.000000","0.972921642743","1.000000000000","0.999996811347","0.999996811712","0.999998405674","0.978219357390","0.986274996092","0.994464412864","0.972921642743","0.986367904356","0.986366331762","0.000003188288","","","","",""
+"NL_NL_RADIXOR","NL_NL","ALL_WORDS","PRIMARY_OUTPUT","4992","26477","85","26477","0","1","26477","5015","63102","1214","1464","350436746","1214","350437960","0.000346","1464","64566","2.267447","0.981124448038","0.977325527367","0.999996535763","0.999992359542","0.988661031565","0.980362303079","0.979221303208","0.978082956166","0.959288537549","0.979223145453","0.979219325206","0.000007640458","0.979217482290","0.997464133435","0.997003025118","0.997233525974","0.997233525974"
+"NL_NL_RADIXOR","NL_NL","ALL_WORDS","ANY_CANDIDATE","4992","26477","85","25905","572","3","27061","5016","64566","0","0","350437960","0","350437960","0.000000","0","64566","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"NL_NL_RADIXOR","NL_NL","ALL_WORDS","ALL_CANDIDATES","4992","26477","85","25905","572","3","27061","5016","64566","2651","0","350435309","2651","350437960","0.000756","0","64566","0.000000","0.960560572474","1.000000000000","0.999992435180","0.999992436574","0.999996217590","0.968197604323","0.979883596519","0.991855131329","0.960560572474","0.980081921308","0.980078214229","0.000007563426","","","","",""
+"NL_NL_RADIXOR","NL_NL","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4796","25678","84","25678","0","1","25678","4797","61763","1214","1384","329602642","1214","329603856","0.000368","1384","63147","2.191711","0.980723121139","0.978082885964","0.999996316791","0.999992119320","0.989039601378","0.980193934392","0.979401224192","0.978609795129","0.959633939808","0.979402113872","0.979398173122","0.000007880680","0.979397283106","0.997373193672","0.997139403762","0.997256285015","0.997256285015"
+"NL_NL_RADIXOR","NL_NL","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","4796","25678","84","25129","549","3","26239","4797","63147","0","0","329603856","0","329603856","0.000000","0","63147","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"NL_NL_RADIXOR","NL_NL","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","4796","25678","84","25129","549","3","26239","4797","63147","2651","0","329601205","2651","329603856","0.000804","0","63147","0.000000","0.959710021581","1.000000000000","0.999991957012","0.999991958552","0.999995978506","0.967506182222","0.979440846873","0.991673628866","0.959710021581","0.979647906945","0.979643967288","0.000008041448","","","","",""
+"NN_NO_RADIXOR","NN_NO","ALL_WORDS","PRIMARY_OUTPUT","4688","18250","23","18250","0","1","18250","4680","26716","6230","3936","166485243","6230","166491473","0.003742","3936","30652","12.840924","0.810902689249","0.871590760799","0.999962580666","0.999938951055","0.935776670732","0.822354650447","0.840152206044","0.858737158800","0.724364188493","0.840699287413","0.840668985911","0.000061048945","0.840121715471","0.983845117159","0.986801676045","0.985321178741","0.985321178741"
+"NN_NO_RADIXOR","NN_NO","ALL_WORDS","ANY_CANDIDATE","4688","18250","23","15846","2404","5","21513","4693","30652","0","0","166491473","0","166491473","0.000000","0","30652","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"NN_NO_RADIXOR","NN_NO","ALL_WORDS","ALL_CANDIDATES","4688","18250","23","15846","2404","5","21513","4693","30652","13214","0","166478259","13214","166491473","0.007937","0","30652","0.000000","0.698764418912","1.000000000000","0.999920632572","0.999920647181","0.999960316286","0.743561877778","0.822673716418","0.920624241623","0.698764418912","0.835921299473","0.835888126353","0.000079352819","","","","",""
+"NN_NO_RADIXOR","NN_NO","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4681","18219","23","18219","0","1","18219","4668","26671","6230","3924","165920046","6230","165926276","0.003755","3924","30595","12.825625","0.810644053372","0.871743748979","0.999962453204","0.999938815429","0.935853101091","0.822169063928","0.840084414766","0.858797921188","0.724263408011","0.840638974932","0.840608609146","0.000061184571","0.840053856992","0.983814668550","0.986841730061","0.985325874420","0.985325874420"
+"NN_NO_RADIXOR","NN_NO","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","4681","18219","23","15820","2399","5","21477","4681","30595","0","0","165926276","0","165926276","0.000000","0","30595","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"NN_NO_RADIXOR","NN_NO","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","4681","18219","23","15820","2399","5","21477","4681","30595","13214","0","165913062","13214","165926276","0.007964","0","30595","0.000000","0.698372480541","1.000000000000","0.999920362222","0.999920376903","0.999960181111","0.743206805583","0.822402021397","0.920488118949","0.698372480541","0.835686831618","0.835653554835","0.000079623097","","","","",""
+"NORWEGIAN_BOKMAL_LUCENE_NORWEGIAN_LIGHT_STEM_FILTER","NB_NO","ALL_WORDS","PRIMARY_OUTPUT","17929","75310","252","75310","0","1","75310","25999","99529","25171","42651","2835593044","25171","2835618215","0.000888","42651","142180","29.997890","0.798147554130","0.700021100014","0.999991123276","0.999976083311","0.850006111645","0.776381478361","0.745870803357","0.717667503101","0.594732030284","0.747475838282","0.747464055956","0.000023916689","0.745858895755","0.987774378220","0.965622291196","0.976572729149","0.976572729149"
+"NORWEGIAN_BOKMAL_LUCENE_NORWEGIAN_LIGHT_STEM_FILTER","NB_NO","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","17914","75251","252","75251","0","1","75251","25985","99450","25118","42641","2831151666","25118","2831176784","0.000887","42641","142091","30.009642","0.798359129150","0.699903582915","0.999991128071","0.999976068044","0.849947355493","0.776512696705","0.745896444523","0.717602881668","0.594764635875","0.747512150366","0.747500361706","0.000023931956","0.745884529658","0.987789184092","0.965602788874","0.976569991255","0.976569991255"
+"NORWEGIAN_BOKMAL_LUCENE_NORWEGIAN_MINIMAL_STEM_FILTER","NB_NO","ALL_WORDS","PRIMARY_OUTPUT","17929","75310","252","75310","0","1","75310","27457","94526","14772","47654","2835603443","14772","2835618215","0.000521","47654","142180","33.516669","0.864846566268","0.664833309889","0.999994790554","0.999977986151","0.832414050221","0.815762584315","0.751763573752","0.697075888841","0.602260563739","0.758273568838","0.758263230800","0.000022013849","0.751752754540","0.992088894987","0.962515968891","0.977078714611","0.977078714611"
+"NORWEGIAN_BOKMAL_LUCENE_NORWEGIAN_MINIMAL_STEM_FILTER","NB_NO","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","17914","75251","252","75251","0","1","75251","27443","94447","14719","47644","2831162065","14719","2831176784","0.000520","47644","142091","33.530625","0.865168642251","0.664693752595","0.999994801102","0.999977973869","0.832344276848","0.815949754214","0.751795969864","0.696994967013","0.602302149098","0.758335144541","0.758324803793","0.000022026131","0.751785145440","0.992107474043","0.962494026490","0.977076419047","0.977076419047"
+"NORWEGIAN_BOKMAL_RADIXOR","NB_NO","ALL_WORDS","PRIMARY_OUTPUT","17929","75310","252","75310","0","1","75310","17886","135010","11482","7170","2835606733","11482","2835618215","0.000405","7170","142180","5.042903","0.921620293258","0.949570966381","0.999995950795","0.999993422575","0.974783458588","0.927078011613","0.935386875069","0.943846020481","0.878616704195","0.935491246621","0.935487968734","0.000006577425","0.935383586923","0.993354053349","0.994615320153","0.993984286646","0.993984286646"
+"NORWEGIAN_BOKMAL_RADIXOR","NB_NO","ALL_WORDS","ANY_CANDIDATE","17929","75310","252","71073","4237","9","79825","17962","142180","0","0","2835618215","0","2835618215","0.000000","0","142180","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"NORWEGIAN_BOKMAL_RADIXOR","NB_NO","ALL_WORDS","ALL_CANDIDATES","17929","75310","252","71073","4237","9","79825","17962","142180","20161","0","2835598054","20161","2835618215","0.000711","0","142180","0.000000","0.875810793330","1.000000000000","0.999992890087","0.999992890443","0.999996445043","0.898118108406","0.933794385280","0.972422273928","0.875810793330","0.935847633608","0.935844306704","0.000007109557","","","","",""
+"NORWEGIAN_BOKMAL_RADIXOR","NB_NO","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","17914","75251","252","75251","0","1","75251","17838","134987","11482","7104","2831165302","11482","2831176784","0.000406","7104","142091","4.999613","0.921607985307","0.950003870759","0.999995944443","0.999993435568","0.974999907601","0.927150543912","0.935590518436","0.944185565020","0.878976122105","0.935698217036","0.935694946007","0.000006564432","0.935587236809","0.993348241255","0.994693619298","0.994020475044","0.994020475044"
+"NORWEGIAN_BOKMAL_RADIXOR","NB_NO","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","17914","75251","252","71047","4204","9","79733","17914","142091","0","0","2831176784","0","2831176784","0.000000","0","142091","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"NORWEGIAN_BOKMAL_RADIXOR","NB_NO","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","17914","75251","252","71047","4204","9","79733","17914","142091","20161","0","2831156623","20161","2831176784","0.000712","0","142091","0.000000","0.875742671893","1.000000000000","0.999992878933","0.999992879290","0.999996439466","0.898060798964","0.933755663840","0.972405477022","0.875742671893","0.935811237319","0.935807905326","0.000007120710","","","","",""
+"PERSIAN_LUCENE_PERSIAN_STEM_FILTER","FA_IR","ALL_WORDS","PRIMARY_OUTPUT","69","3701","0","3701","0","1","3701","3190","430","179","98018","6748223","179","6748402","0.002652","98018","98448","99.563221","0.706075533662","0.004367788071","0.999973475202","0.985658076342","0.502170631636","0.021311605408","0.008681870034","0.005451304637","0.004359860890","0.055533668104","0.054800599292","0.014341923658","0.008506575635","0.985685778881","0.520346904570","0.681125382676","0.681125382676"
+"PERSIAN_LUCENE_PERSIAN_STEM_FILTER","FA_IR","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","69","3701","0","3701","0","1","3701","3190","430","179","98018","6748223","179","6748402","0.002652","98018","98448","99.563221","0.706075533662","0.004367788071","0.999973475202","0.985658076342","0.502170631636","0.021311605408","0.008681870034","0.005451304637","0.004359860890","0.055533668104","0.054800599292","0.014341923658","0.008506575635","0.985685778881","0.520346904570","0.681125382676","0.681125382676"
+"PERSIAN_RADIXOR","FA_IR","ALL_WORDS","PRIMARY_OUTPUT","69","3701","0","3701","0","1","3701","69","93636","8621","4812","6739781","8621","6748402","0.127749","4812","98448","4.887860","0.915692813206","0.951121404193","0.998722512381","0.998038075904","0.974921958287","0.922565796215","0.933070924989","0.943818050233","0.874538848780","0.933239001706","0.932248664283","0.001961924096","0.932075735269","0.980342969277","0.984586447427","0.982460126226","0.982460126226"
+"PERSIAN_RADIXOR","FA_IR","ALL_WORDS","ANY_CANDIDATE","69","3701","0","3387","314","2","4015","69","98448","0","0","6748402","0","6748402","0.000000","0","98448","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"PERSIAN_RADIXOR","FA_IR","ALL_WORDS","ALL_CANDIDATES","69","3701","0","3387","314","2","4015","69","98448","13433","0","6734969","13433","6748402","0.199055","0","98448","0.000000","0.879934930864","1.000000000000","0.998009454683","0.998038075904","0.999004727341","0.901584696651","0.936133391021","0.973435401930","0.879934930864","0.938048469358","0.937114390300","0.001961924096","","","","",""
+"PERSIAN_RADIXOR","FA_IR","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","69","3701","0","3701","0","1","3701","69","93636","8621","4812","6739781","8621","6748402","0.127749","4812","98448","4.887860","0.915692813206","0.951121404193","0.998722512381","0.998038075904","0.974921958287","0.922565796215","0.933070924989","0.943818050233","0.874538848780","0.933239001706","0.932248664283","0.001961924096","0.932075735269","0.980342969277","0.984586447427","0.982460126226","0.982460126226"
+"PERSIAN_RADIXOR","FA_IR","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","69","3701","0","3387","314","2","4015","69","98448","0","0","6748402","0","6748402","0.000000","0","98448","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"PERSIAN_RADIXOR","FA_IR","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","69","3701","0","3387","314","2","4015","69","98448","13433","0","6734969","13433","6748402","0.199055","0","98448","0.000000","0.879934930864","1.000000000000","0.998009454683","0.998038075904","0.999004727341","0.901584696651","0.936133391021","0.973435401930","0.879934930864","0.938048469358","0.937114390300","0.001961924096","","","","",""
+"POLISH_LUCENE_MORFOLOGIK_FILTER","PL_PL","ALL_WORDS","PRIMARY_OUTPUT","9990","122341","1","122341","0","1","122341","15519","1004747","99228","116220","7482378775","99228","7482478003","0.001326","116220","1120967","10.367834","0.910117529835","0.896321657997","0.999986738618","0.999971210643","0.948154198307","0.907324485129","0.903166914014","0.899047271013","0.823431500703","0.903193253581","0.903178865015","0.000028789357","0.903152518035","0.990022261217","0.977053921984","0.983495343428","0.983495343428"
+"POLISH_LUCENE_MORFOLOGIK_FILTER","PL_PL","ALL_WORDS","ANY_CANDIDATE","9990","122341","1","109468","12873","5","136636","16295","1093112","85532","27855","7482392471","85532","7482478003","0.001143","27855","1120967","2.484908","0.927431862377","0.975150918805","0.999988569028","0.999984848600","0.987569743916","0.936598359399","0.950692965028","0.965218263555","0.906019814355","0.950992130738","0.950984648194","0.000015151400","","","","",""
+"POLISH_LUCENE_MORFOLOGIK_FILTER","PL_PL","ALL_WORDS","ALL_CANDIDATES","9990","122341","1","109468","12873","5","136636","16295","1093112","143096","27855","7482334907","143096","7482478003","0.001912","27855","1120967","2.484908","0.884246016852","0.975150918805","0.999980875854","0.999977156579","0.987565897330","0.901045352805","0.927476322293","0.955504786999","0.864760696263","0.928586730350","0.928575670129","0.000022843421","","","","",""
+"POLISH_LUCENE_MORFOLOGIK_FILTER","PL_PL","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","9846","120925","1","120925","0","1","120925","15277","999138","99224","115513","7310153475","99224","7310252699","0.001357","115513","1114651","10.363154","0.909661841906","0.896368459724","0.999986426735","0.999970629707","0.948177443229","0.906971715650","0.902966227492","0.898995962905","0.823097930182","0.902990688822","0.902976009256","0.000029370293","0.902951540918","0.989889334726","0.977011745487","0.983408384376","0.983408384376"
+"POLISH_LUCENE_MORFOLOGIK_FILTER","PL_PL","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","9846","120925","1","108162","12763","5","135105","16044","1087157","85532","27494","7310167167","85532","7310252699","0.001170","27494","1114651","2.466602","0.927063356099","0.975333983462","0.999988299720","0.999984541059","0.987661141591","0.936331423447","0.950586270515","0.965281863330","0.905826028197","0.950892420848","0.950884788442","0.000015458941","","","","",""
+"POLISH_LUCENE_MORFOLOGIK_FILTER","PL_PL","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","9846","120925","1","108162","12763","5","135105","16044","1087157","143085","27494","7310109614","143085","7310252699","0.001957","27494","1114651","2.466602","0.883693614752","0.975333983462","0.999980426805","0.999976669344","0.987657205134","0.900617649988","0.927255102898","0.955516285728","0.864376148890","0.928383764096","0.928372472930","0.000023330656","","","","",""
+"POLISH_LUCENE_STEMPEL_DIRECT","PL_PL","ALL_WORDS","PRIMARY_OUTPUT","9990","122341","1","122341","0","1","122341","31432","797573","66669","323394","7482411334","66669","7482478003","0.000891","323394","1120967","28.849556","0.922858412343","0.711504442147","0.999991089984","0.999947877619","0.855747766065","0.871105640425","0.803515398127","0.745658746735","0.671563509358","0.810319603524","0.810295555502","0.000052122381","0.803489769425","0.991766794523","0.931068514076","0.960459620808","0.960459620808"
+"POLISH_LUCENE_STEMPEL_DIRECT","PL_PL","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","9846","120925","1","120925","0","1","120925","30830","794493","66274","320158","7310186425","66274","7310252699","0.000907","320158","1114651","28.722712","0.923005877316","0.712772876892","0.999990934103","0.999947146412","0.856381905497","0.871590591697","0.804379630033","0.746792242917","0.672771767894","0.811106376847","0.811081975971","0.000052853588","0.804353636298","0.991711086900","0.931317880193","0.960566151650","0.960566151650"
+"POLISH_LUCENE_STEMPEL_FILTER","PL_PL","ALL_WORDS","PRIMARY_OUTPUT","9990","122341","1","122341","0","1","122341","31432","797573","66669","323394","7482411334","66669","7482478003","0.000891","323394","1120967","28.849556","0.922858412343","0.711504442147","0.999991089984","0.999947877619","0.855747766065","0.871105640425","0.803515398127","0.745658746735","0.671563509358","0.810319603524","0.810295555502","0.000052122381","0.803489769425","0.991766794523","0.931068514076","0.960459620808","0.960459620808"
+"POLISH_LUCENE_STEMPEL_FILTER","PL_PL","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","9846","120925","1","120925","0","1","120925","30830","794493","66274","320158","7310186425","66274","7310252699","0.000907","320158","1114651","28.722712","0.923005877316","0.712772876892","0.999990934103","0.999947146412","0.856381905497","0.871590591697","0.804379630033","0.746792242917","0.672771767894","0.811106376847","0.811081975971","0.000052853588","0.804353636298","0.991711086900","0.931317880193","0.960566151650","0.960566151650"
+"POLISH_RADIXOR","PL_PL","ALL_WORDS","PRIMARY_OUTPUT","9990","122341","1","122341","0","1","122341","10074","1099420","13669","21547","7482464334","13669","7482478003","0.000183","21547","1120967","1.922180","0.987719760055","0.980778203105","0.999998173199","0.999995294243","0.990388188152","0.986323599045","0.984236742499","0.982158698021","0.968962733423","0.984242862020","0.984240510632","0.000004705757","0.984234389298","0.996967243455","0.996469409869","0.996718264498","0.996718264498"
+"POLISH_RADIXOR","PL_PL","ALL_WORDS","ANY_CANDIDATE","9990","122341","1","119475","2866","4","125778","10079","1120967","0","0","7482478003","0","7482478003","0.000000","0","1120967","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"POLISH_RADIXOR","PL_PL","ALL_WORDS","ALL_CANDIDATES","9990","122341","1","119475","2866","4","125778","10079","1120967","38073","0","7482439930","38073","7482478003","0.000509","0","1120967","0.000000","0.967151263114","1.000000000000","0.999994911712","0.999994912475","0.999997455856","0.973547222425","0.983301367057","0.993252946885","0.967151263114","0.983438489746","0.983435987734","0.000005087525","","","","",""
+"POLISH_RADIXOR","PL_PL","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","9846","120925","1","120925","0","1","120925","9844","1093651","13669","21000","7310239030","13669","7310252699","0.000187","21000","1114651","1.883998","0.987655781527","0.981160022285","0.999998130160","0.999995258206","0.990579076223","0.986349757961","0.984397186102","0.982452329568","0.969273787578","0.984402543989","0.984400174373","0.000004741794","0.984394814870","0.996926141446","0.996646530259","0.996786316244","0.996786316244"
+"POLISH_RADIXOR","PL_PL","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","9846","120925","1","118145","2780","4","124274","9847","1114651","0","0","7310252699","0","7310252699","0.000000","0","1114651","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"POLISH_RADIXOR","PL_PL","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","9846","120925","1","118145","2780","4","124274","9847","1114651","38073","0","7310214626","38073","7310252699","0.000521","0","1114651","0.000000","0.966971278467","1.000000000000","0.999994791835","0.999994792629","0.999997395918","0.973401318686","0.983208335630","0.993214975136","0.966971278467","0.983346977657","0.983344416937","0.000005207371","","","","",""
+"PORTUGUESE_LUCENE_PORTUGUESE_LIGHT_STEM_FILTER","PT_PT","ALL_WORDS","PRIMARY_OUTPUT","4001","211489","0","211489","0","1","211489","112814","149830","2511","5339230","22358201245","2511","22358203756","0.000011","5339230","5489060","97.270389","0.983517240927","0.027296112631","0.999999887692","0.999761142266","0.513648000162","0.122843213263","0.053118010934","0.033885033146","0.027283631587","0.163848092400","0.163827867245","0.000238857734","0.053105458848","0.999226179883","0.720580123442","0.837329788424","0.837329788424"
+"PORTUGUESE_LUCENE_PORTUGUESE_LIGHT_STEM_FILTER","PT_PT","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4001","211489","0","211489","0","1","211489","112814","149830","2511","5339230","22358201245","2511","22358203756","0.000011","5339230","5489060","97.270389","0.983517240927","0.027296112631","0.999999887692","0.999761142266","0.513648000162","0.122843213263","0.053118010934","0.033885033146","0.027283631587","0.163848092400","0.163827867245","0.000238857734","0.053105458848","0.999226179883","0.720580123442","0.837329788424","0.837329788424"
+"PORTUGUESE_LUCENE_PORTUGUESE_MINIMAL_STEM_FILTER","PT_PT","ALL_WORDS","PRIMARY_OUTPUT","4001","211489","0","211489","0","1","211489","167745","43406","598","5445654","22358203158","598","22358203756","0.000003","5445654","5489060","99.209227","0.986410326334","0.007907729192","0.999999973254","0.999756469021","0.503953851223","0.038310165654","0.015689679353","0.009864890589","0.007906867787","0.088319113068","0.088308061848","0.000243530979","0.015685836581","0.999664059174","0.692382565086","0.818121623354","0.818121623354"
+"PORTUGUESE_LUCENE_PORTUGUESE_MINIMAL_STEM_FILTER","PT_PT","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4001","211489","0","211489","0","1","211489","167745","43406","598","5445654","22358203158","598","22358203756","0.000003","5445654","5489060","99.209227","0.986410326334","0.007907729192","0.999999973254","0.999756469021","0.503953851223","0.038310165654","0.015689679353","0.009864890589","0.007906867787","0.088319113068","0.088308061848","0.000243530979","0.015685836581","0.999664059174","0.692382565086","0.818121623354","0.818121623354"
+"PORTUGUESE_LUCENE_PORTUGUESE_STEM_FILTER","PT_PT","ALL_WORDS","PRIMARY_OUTPUT","4001","211489","0","211489","0","1","211489","27586","3803488","99075","1685572","22358104681","99075","22358203756","0.000443","1685572","5489060","30.707844","0.974612837768","0.692921556696","0.999995568741","0.999920198913","0.846458562719","0.901329863268","0.809974591186","0.735433886866","0.680636384053","0.821784792219","0.821750382444","0.000079801087","0.809935821347","0.996728545383","0.918475110538","0.956003146785","0.956003146785"
+"PORTUGUESE_LUCENE_PORTUGUESE_STEM_FILTER","PT_PT","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4001","211489","0","211489","0","1","211489","27586","3803488","99075","1685572","22358104681","99075","22358203756","0.000443","1685572","5489060","30.707844","0.974612837768","0.692921556696","0.999995568741","0.999920198913","0.846458562719","0.901329863268","0.809974591186","0.735433886866","0.680636384053","0.821784792219","0.821750382444","0.000079801087","0.809935821347","0.996728545383","0.918475110538","0.956003146785","0.956003146785"
+"PORTUGUESE_RADIXOR","PT_PT","ALL_WORDS","PRIMARY_OUTPUT","4001","211489","0","211489","0","1","211489","4001","5472616","20678","16444","22358183078","20678","22358203756","0.000092","16444","5489060","0.299578","0.996235774018","0.997004222945","0.999999075149","0.999998340077","0.998501649047","0.996389369023","0.996619850353","0.996850438335","0.993262474550","0.996619924417","0.996619094289","0.000001659923","0.996619020188","0.999299376330","0.999346803887","0.999323089546","0.999323089546"
+"PORTUGUESE_RADIXOR","PT_PT","ALL_WORDS","ANY_CANDIDATE","4001","211489","0","210699","790","3","212297","4001","5489060","0","0","22358203756","0","22358203756","0.000000","0","5489060","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"PORTUGUESE_RADIXOR","PT_PT","ALL_WORDS","ALL_CANDIDATES","4001","211489","0","210699","790","3","212297","4001","5489060","38310","0","22358165446","38310","22358203756","0.000171","0","5489060","0.000000","0.993069036450","1.000000000000","0.999998286535","0.999998286956","0.999999143267","0.994447532369","0.996522466897","0.998606078314","0.993069036450","0.996528492543","0.996527638784","0.000001713044","","","","",""
+"PORTUGUESE_RADIXOR","PT_PT","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4001","211489","0","211489","0","1","211489","4001","5472616","20678","16444","22358183078","20678","22358203756","0.000092","16444","5489060","0.299578","0.996235774018","0.997004222945","0.999999075149","0.999998340077","0.998501649047","0.996389369023","0.996619850353","0.996850438335","0.993262474550","0.996619924417","0.996619094289","0.000001659923","0.996619020188","0.999299376330","0.999346803887","0.999323089546","0.999323089546"
+"PORTUGUESE_RADIXOR","PT_PT","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","4001","211489","0","210699","790","3","212297","4001","5489060","0","0","22358203756","0","22358203756","0.000000","0","5489060","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"PORTUGUESE_RADIXOR","PT_PT","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","4001","211489","0","210699","790","3","212297","4001","5489060","38310","0","22358165446","38310","22358203756","0.000171","0","5489060","0.000000","0.993069036450","1.000000000000","0.999998286535","0.999998286956","0.999999143267","0.994447532369","0.996522466897","0.998606078314","0.993069036450","0.996528492543","0.996527638784","0.000001713044","","","","",""
+"RUSSIAN_LUCENE_RUSSIAN_LIGHT_STEM_FILTER","RU_RU","ALL_WORDS","PRIMARY_OUTPUT","37410","768882","10","768882","0","1","768882","232250","3081067","321183","10008438","295575969833","321183","295576291016","0.000109","10008438","13089505","76.461547","0.905596884415","0.235384531348","0.999998913367","0.999965054154","0.617691722357","0.577011147253","0.373649378129","0.276277984307","0.229747124085","0.461696326851","0.461686629842","0.000034945846","0.373637933830","0.994310930069","0.870888421754","0.928516167212","0.928516167212"
+"RUSSIAN_LUCENE_RUSSIAN_LIGHT_STEM_FILTER","RU_RU","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","37297","768133","10","768133","0","1","768133","232143","3078888","318921","10008238","295000362731","318921","295000681652","0.000108","10008238","13087126","76.473918","0.906139220892","0.235260820443","0.999998918914","0.999964994315","0.617629869679","0.577038425373","0.373539598427","0.276151716079","0.229664120975","0.461713175622","0.461703471649","0.000035005685","0.373528142103","0.994349544377","0.870767393403","0.928464208705","0.928464208705"
+"RUSSIAN_RADIXOR","RU_RU","ALL_WORDS","PRIMARY_OUTPUT","37410","768882","10","768882","0","1","768882","37561","12823203","155850","266302","295576135166","155850","295576291016","0.000053","266302","13089505","2.034470","0.987992190185","0.979655304001","0.999999472725","0.999998571830","0.989827388363","0.986313480705","0.983806085477","0.981311406466","0.968128298562","0.983814916245","0.983814202914","0.000001428170","0.983805371373","0.997699288696","0.997273959852","0.997486578934","0.997486578934"
+"RUSSIAN_RADIXOR","RU_RU","ALL_WORDS","ANY_CANDIDATE","37410","768882","10","749720","19162","4","788492","37593","13089492","0","13","295576291016","0","295576291016","0.000000","13","13089505","0.000099","1.000000000000","0.999999006838","1.000000000000","0.999999999956","0.999999503419","0.999999801367","0.999999503419","0.999999205470","0.999999006838","0.999999503419","0.999999503397","0.000000000044","","","","",""
+"RUSSIAN_RADIXOR","RU_RU","ALL_WORDS","ALL_CANDIDATES","37410","768882","10","749720","19162","4","788492","37593","13089492","434710","13","295575856306","434710","295576291016","0.000147","13","13089505","0.000099","0.967856883534","0.999999006838","0.999998529280","0.999998529301","0.999998768059","0.974118939969","0.983665447282","0.993400920813","0.967855953192","0.983796687479","0.983795964011","0.000001470699","","","","",""
+"RUSSIAN_RADIXOR","RU_RU","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","37297","768133","10","768133","0","1","768133","37282","12821513","155850","265613","295000525802","155850","295000681652","0.000053","265613","13087126","2.029575","0.987990626447","0.979704252867","0.999999471696","0.999998571379","0.989851862281","0.986322156837","0.983829991833","0.981350389119","0.968174600634","0.983838715706","0.983838002141","0.000001428621","0.983829277503","0.997696524283","0.997321167437","0.997508810549","0.997508810549"
+"RUSSIAN_RADIXOR","RU_RU","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","37297","768133","10","749142","18991","4","787549","37306","13087126","0","0","295000681652","0","295000681652","0.000000","0","13087126","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"RUSSIAN_RADIXOR","RU_RU","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","37297","768133","10","749142","18991","4","787549","37306","13087126","434710","0","295000246942","434710","295000681652","0.000147","0","13087126","0.000000","0.967851259252","1.000000000000","0.999998526410","0.999998526476","0.999999263205","0.974114570610","0.983663023007","0.993400519870","0.967851259252","0.983794317554","0.983793592699","0.000001473524","","","","",""
+"SNOWBALL_DANISH_DIRECT","DA_DK","ALL_WORDS","PRIMARY_OUTPUT","4179","28079","32","28079","0","1","28079","5553","78732","6341","11163","394104845","6341","394111186","0.001609","11163","89895","12.417821","0.925464013259","0.875821792091","0.999983910632","0.999955596266","0.937902851361","0.915090414169","0.899958849618","0.885319563795","0.818113803566","0.900300811178","0.900278764621","0.000044403734","0.899936659693","0.994195946109","0.978579164615","0.986325742978","0.986325742978"
+"SNOWBALL_DANISH_DIRECT","DA_DK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4173","28033","32","28033","0","1","28033","5539","78627","6341","11113","392814447","6341","392820788","0.001614","11113","89740","12.383552","0.925371904717","0.876164475150","0.999983857779","0.999955577673","0.938074166465","0.915093153823","0.900096160451","0.885582797210","0.818340774971","0.900432112497","0.900410054086","0.000044422327","0.900073960926","0.994185354918","0.978644331817","0.986353630937","0.986353630937"
+"SNOWBALL_DANISH_LUCENE_FILTER","DA_DK","ALL_WORDS","PRIMARY_OUTPUT","4179","28079","32","28079","0","1","28079","5546","78744","6507","11151","394104679","6507","394111186","0.001651","11151","89895","12.404472","0.923672449590","0.875955281161","0.999983489431","0.999955205602","0.937969385296","0.913717599716","0.899181254496","0.885100184115","0.816829526358","0.899497504322","0.899475250557","0.000044794398","0.899158868074","0.994052746860","0.978603476314","0.986267614487","0.986267614487"
+"SNOWBALL_DANISH_LUCENE_FILTER","DA_DK","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4173","28033","32","28033","0","1","28033","5539","78627","6341","11113","392814447","6341","392820788","0.001614","11113","89740","12.383552","0.925371904717","0.876164475150","0.999983857779","0.999955577673","0.938074166465","0.915093153823","0.900096160451","0.885582797210","0.818340774971","0.900432112497","0.900410054086","0.000044422327","0.900073960926","0.994185354918","0.978644331817","0.986353630937","0.986353630937"
+"SNOWBALL_DUTCH_DIRECT","NL_NL","ALL_WORDS","PRIMARY_OUTPUT","4992","26477","85","26477","0","1","26477","12051","29325","4382","35241","350433578","4382","350437960","0.001250","35241","64566","54.581359","0.869997329931","0.454186413902","0.999987495647","0.999886953739","0.727086954774","0.735353119953","0.596806854375","0.502190286022","0.425320531415","0.628602392126","0.628557411604","0.000113046261","0.596755898222","0.992814719235","0.917346281080","0.953589661124","0.953589661124"
+"SNOWBALL_DUTCH_DIRECT","NL_NL","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4796","25678","84","25678","0","1","25678","11466","29111","4382","34036","329599474","4382","329603856","0.001329","34036","63147","53.899631","0.869166691547","0.461003689803","0.999986705253","0.999883464224","0.730495197528","0.738411822300","0.602462748344","0.508789468717","0.431088865524","0.633000040962","0.632953313739","0.000116535776","0.602409959779","0.992557044900","0.918058993802","0.953855618789","0.953855618789"
+"SNOWBALL_DUTCH_LUCENE_FILTER","NL_NL","ALL_WORDS","PRIMARY_OUTPUT","4992","26477","85","26477","0","1","26477","14573","15302","1588","49264","350436372","1588","350437960","0.000453","49264","64566","76.300220","0.905979869745","0.236997800700","0.999995468527","0.999854916880","0.618496634614","0.579068464950","0.375712040856","0.278062466837","0.231308764398","0.463373754768","0.463333378452","0.000145083120","0.375664346452","0.995827666179","0.888410302915","0.939057139355","0.939057139355"
+"SNOWBALL_DUTCH_LUCENE_FILTER","NL_NL","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4796","25678","84","25678","0","1","25678","14116","14972","1544","48175","329602312","1544","329603856","0.000468","48175","63147","76.290243","0.906514894648","0.237097565997","0.999995315589","0.999849184178","0.618546440793","0.579362438183","0.375883408860","0.278182412747","0.231438685443","0.463608105042","0.463566154833","0.000150815822","0.375833834664","0.995816517119","0.887491664267","0.938538755173","0.938538755173"
+"SNOWBALL_FINNISH_DIRECT","FI_FI","ALL_WORDS","PRIMARY_OUTPUT","57027","1811717","292","1811717","0","1","1811717","381483","15114332","1544812","16409363","1641125269679","1544812","1641126814491","0.000094","16409363","31523695","52.054060","0.907269425128","0.479459403474","0.999999058688","0.999989060059","0.739729231081","0.769880311353","0.627374073993","0.529384116965","0.457061215373","0.659544431681","0.659540149918","0.000010939941","0.627369124557","0.991871857177","0.904138579582","0.945975396220","0.945975396220"
+"SNOWBALL_FINNISH_DIRECT","FI_FI","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","54762","1757055","274","1757055","0","1","1757055","372232","14692070","1513705","16121763","1543587930447","1513705","1543589444152","0.000098","16121763","30813833","52.319888","0.906594717007","0.476801117213","0.999999019360","0.999988575255","0.738400068286","0.768116957494","0.624933751043","0.526744348874","0.454475376380","0.657468914800","0.657464451566","0.000011424745","0.624928589970","0.991731678095","0.902933406396","0.945251664438","0.945251664438"
+"SNOWBALL_FINNISH_LUCENE_FILTER","FI_FI","ALL_WORDS","PRIMARY_OUTPUT","57027","1811717","292","1811717","0","1","1811717","377778","15153638","1922153","16370057","1641124892338","1922153","1641126814491","0.000117","16370057","31523695","51.929372","0.887434028678","0.480706275073","0.999998828760","0.999988854086","0.740352551917","0.758996033322","0.623613097472","0.529216231176","0.453079796332","0.653142485450","0.653138019077","0.000011145914","0.623608016975","0.990717710840","0.904385055188","0.945584912497","0.945584912497"
+"SNOWBALL_FINNISH_LUCENE_FILTER","FI_FI","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","54762","1757055","274","1757055","0","1","1757055","372232","14692070","1513705","16121763","1543587930447","1513705","1543589444152","0.000098","16121763","30813833","52.319888","0.906594717007","0.476801117213","0.999999019360","0.999988575255","0.738400068286","0.768116957494","0.624933751043","0.526744348874","0.454475376380","0.657468914800","0.657464451566","0.000011424745","0.624928589970","0.991731678095","0.902933406396","0.945251664438","0.945251664438"
+"SNOWBALL_FRENCH_DIRECT","FR_FR","ALL_WORDS","PRIMARY_OUTPUT","59240","425210","2301","425210","0","1","425210","85627","3766640","1654723","1687975","90394450107","1654723","90396104830","0.001831","1687975","5454615","30.945814","0.694777309691","0.690541862258","0.999981694753","0.999963023890","0.845261778506","0.693926068790","0.692653111288","0.691384815533","0.529815856272","0.692656348624","0.692637859933","0.000036976110","0.692634622294","0.959459328254","0.944947915186","0.952148333884","0.952148333884"
+"SNOWBALL_FRENCH_DIRECT","FR_FR","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","57698","421231","2133","421231","0","1","421231","84526","3758589","1646111","1681970","88710480395","1646111","88712126506","0.001856","1681970","5440559","30.915389","0.695429718578","0.690846106071","0.999981444352","0.999962486787","0.845413775212","0.694508136723","0.693130334647","0.691757988461","0.530374491828","0.693134123475","0.693115366288","0.000037513213","0.693111577099","0.959520798119","0.944537159644","0.951970023370","0.951970023370"
+"SNOWBALL_FRENCH_LUCENE_FILTER","FR_FR","ALL_WORDS","PRIMARY_OUTPUT","59240","425210","2301","425210","0","1","425210","85202","3763777","1661388","1690838","90394443442","1661388","90396104830","0.001838","1690838","5454615","30.998301","0.693762678186","0.690016985617","0.999981621022","0.999962918494","0.844999303320","0.693010289898","0.691884762376","0.690762884895","0.528917286853","0.691887297134","0.691868755638","0.000037081506","0.691866220643","0.958697792387","0.944714715363","0.951654891797","0.951654891797"
+"SNOWBALL_FRENCH_LUCENE_FILTER","FR_FR","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","57698","421231","2133","421231","0","1","421231","84810","3755856","1641925","1684703","88710484581","1641925","88712126506","0.001851","1684703","5440559","30.965623","0.695814817237","0.690343767984","0.999981491538","0.999962503165","0.845162629761","0.694713680979","0.693068495729","0.691431084156","0.530302080457","0.693073894149","0.693055145391","0.000037496835","0.693049746458","0.959566165512","0.944384738154","0.951914926182","0.951914926182"
+"SNOWBALL_GERMAN_DIRECT","DE_DE","ALL_WORDS","PRIMARY_OUTPUT","54092","296974","1474","296974","0","1","296974","81649","771138","190680","612734","44095055299","190680","44095245979","0.000432","612734","1383872","44.276783","0.801750435114","0.557232171762","0.999995675724","0.999981780603","0.778613923743","0.737064397386","0.657493530688","0.593429030432","0.489750735447","0.668401927114","0.668393541401","0.000018219397","0.657484715679","0.983724573695","0.949324273697","0.966218331938","0.966218331938"
+"SNOWBALL_GERMAN_DIRECT","DE_DE","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","16007","150098","228","150098","0","1","150098","37843","516936","87697","356475","11263668645","87697","11263756342","0.000779","356475","873411","40.814118","0.854958297017","0.591858815609","0.999992214231","0.999960569321","0.795925514920","0.785153327381","0.699486618802","0.630674793334","0.537854226580","0.711347035607","0.711329191110","0.000039430679","0.699467554207","0.988417636496","0.932451723900","0.959619376607","0.959619376607"
+"SNOWBALL_GERMAN_LUCENE_FILTER","DE_DE","ALL_WORDS","PRIMARY_OUTPUT","54092","296974","1474","296974","0","1","296974","86669","751056","295701","632816","44094950278","295701","44095245979","0.000671","632816","1383872","45.727929","0.717507501741","0.542720714054","0.999993294039","0.999978943584","0.771357004047","0.674088567377","0.617993120299","0.570516594262","0.447170798768","0.624024185176","0.624014089276","0.000021056416","0.617982794334","0.975844522648","0.942925157079","0.959102449371","0.959102449371"
+"SNOWBALL_GERMAN_LUCENE_FILTER","DE_DE","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","16007","150098","228","150098","0","1","150098","46077","481501","77653","391910","11263678689","77653","11263756342","0.000689","391910","873411","44.871200","0.861124126806","0.551287996144","0.999993105941","0.999958315274","0.775640551042","0.774110642769","0.672222202832","0.594035281304","0.506276128631","0.689004640259","0.688986412764","0.000041684726","0.672202362239","0.989021274644","0.919542244735","0.953017108147","0.953017108147"
+"SNOWBALL_HUNGARIAN_DIRECT","HU_HU","ALL_WORDS","PRIMARY_OUTPUT","19406","916344","1","916344","0","1","916344","116105","14287912","1506056","7874191","419819036837","1506056","419820542893","0.000359","7874191","22162103","35.529981","0.904643595580","0.644700189328","0.999996412620","0.999977657711","0.822348300974","0.837136808086","0.752865700984","0.684009307333","0.603676525918","0.763690969794","0.763680928295","0.000022342289","0.752854843818","0.991947513126","0.924304257644","0.956931989551","0.956931989551"
+"SNOWBALL_HUNGARIAN_DIRECT","HU_HU","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","18360","878513","1","878513","0","1","878513","111379","13776526","1496670","7634885","385869198247","1496670","385870694917","0.000388","7634885","21411411","35.658019","0.902006757459","0.643419810119","0.999996121317","0.999976336507","0.821707965718","0.834898516372","0.751079383241","0.682554714263","0.601382804609","0.761819543337","0.761808891723","0.000023663493","0.751067882221","0.991609896137","0.923288102987","0.956230170303","0.956230170303"
+"SNOWBALL_HUNGARIAN_LUCENE_FILTER","HU_HU","ALL_WORDS","PRIMARY_OUTPUT","19406","916344","1","916344","0","1","916344","114867","14299358","1792049","7862745","419818750844","1792049","419820542893","0.000427","7862745","22162103","35.478334","0.888633169244","0.645216656560","0.999995731393","0.999977003783","0.822606193976","0.826287586346","0.747610245439","0.682613266689","0.596946950992","0.757205997314","0.757195513225","0.000022996217","0.747599036407","0.990687085622","0.924490230693","0.956444632598","0.956444632598"
+"SNOWBALL_HUNGARIAN_LUCENE_FILTER","HU_HU","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","18360","878513","1","878513","0","1","878513","111379","13776526","1496670","7634885","385869198247","1496670","385870694917","0.000388","7634885","21411411","35.658019","0.902006757459","0.643419810119","0.999996121317","0.999976336507","0.821707965718","0.834898516372","0.751079383241","0.682554714263","0.601382804609","0.761819543337","0.761808891723","0.000023663493","0.751067882221","0.991609896137","0.923288102987","0.956230170303","0.956230170303"
+"SNOWBALL_ITALIAN_DIRECT","IT_IT","ALL_WORDS","PRIMARY_OUTPUT","10009","327551","0","327551","0","1","327551","46828","4499650","504775","1644164","53638016436","504775","53638521211","0.000941","1644164","6143814","26.761292","0.899134266174","0.732387080729","0.999990589319","0.999959941236","0.866188835024","0.859975076366","0.807239600802","0.760598128154","0.676782697802","0.811488952720","0.811469907061","0.000040058764","0.807219778599","0.987993752409","0.933407841859","0.959925420024","0.959925420024"
+"SNOWBALL_ITALIAN_DIRECT","IT_IT","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","10007","327469","0","327469","0","1","327469","46814","4498652","504774","1643522","53611162298","504774","53611667072","0.000942","1643522","6142174","26.757985","0.899114326863","0.732420149608","0.999990584624","0.999959933163","0.866205367116","0.859969602244","0.807251650876","0.760623806435","0.676799637969","0.811498274672","0.811479224597","0.000040066837","0.807231824542","0.987990915845","0.933413003102","0.959926810498","0.959926810498"
+"SNOWBALL_ITALIAN_LUCENE_FILTER","IT_IT","ALL_WORDS","PRIMARY_OUTPUT","10009","327551","0","327551","0","1","327551","46828","4499650","504775","1644164","53638016436","504775","53638521211","0.000941","1644164","6143814","26.761292","0.899134266174","0.732387080729","0.999990589319","0.999959941236","0.866188835024","0.859975076366","0.807239600802","0.760598128154","0.676782697802","0.811488952720","0.811469907061","0.000040058764","0.807219778599","0.987993752409","0.933407841859","0.959925420024","0.959925420024"
+"SNOWBALL_ITALIAN_LUCENE_FILTER","IT_IT","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","10007","327469","0","327469","0","1","327469","46814","4498652","504774","1643522","53611162298","504774","53611667072","0.000942","1643522","6142174","26.757985","0.899114326863","0.732420149608","0.999990584624","0.999959933163","0.866205367116","0.859969602244","0.807251650876","0.760623806435","0.676799637969","0.811498274672","0.811479224597","0.000040066837","0.807231824542","0.987990915845","0.933413003102","0.959926810498","0.959926810498"
+"SNOWBALL_NORWEGIAN_BOKMAL_DIRECT","NB_NO","ALL_WORDS","PRIMARY_OUTPUT","17929","75310","252","75310","0","1","75310","24394","106626","23997","35554","2835594218","23997","2835618215","0.000846","35554","142180","25.006330","0.816288096277","0.749936699958","0.999991537295","0.999978999989","0.874964118626","0.802094867845","0.781706946038","0.762329786671","0.641641141674","0.782409356499","0.782398932962","0.000021000011","0.781696464373","0.988328173631","0.971119668400","0.979648355684","0.979648355684"
+"SNOWBALL_NORWEGIAN_BOKMAL_DIRECT","NB_NO","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","17914","75251","252","75251","0","1","75251","24367","106567","23997","35524","2831152787","23997","2831176784","0.000848","35524","142091","25.000880","0.816205079501","0.749991202821","0.999991524019","0.999978977642","0.874991363420","0.802043209347","0.781698483431","0.762360357576","0.641629738452","0.782397999310","0.782387564360","0.000021022358","0.781687990535","0.988317966243","0.971127035574","0.979647089734","0.979647089734"
+"SNOWBALL_NORWEGIAN_BOKMAL_LUCENE_FILTER","NB_NO","ALL_WORDS","PRIMARY_OUTPUT","17929","75310","252","75310","0","1","75310","24396","106589","24046","35591","2835594169","24046","2835618215","0.000848","35591","142180","25.032353","0.815929880966","0.749676466451","0.999991520015","0.999978969662","0.874833993233","0.801758635215","0.781401315910","0.762052176648","0.641229410562","0.782101930719","0.782091491842","0.000021030338","0.781390819068","0.988295184140","0.971086244692","0.979615142710","0.979615142710"
+"SNOWBALL_NORWEGIAN_BOKMAL_LUCENE_FILTER","NB_NO","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","17914","75251","252","75251","0","1","75251","24381","106512","23993","35579","2831152791","23993","2831176784","0.000847","35579","142091","25.039587","0.816152637830","0.749604126933","0.999991525432","0.999978959629","0.874797826182","0.801914137847","0.781464144742","0.762031224736","0.641314033862","0.782170943927","0.782160500766","0.000021040371","0.781453643056","0.988310445474","0.971073741466","0.979616277822","0.979616277822"
+"SNOWBALL_NORWEGIAN_NYNORSK_DIRECT","NN_NO","ALL_WORDS","PRIMARY_OUTPUT","4688","18250","23","18250","0","1","18250","6138","22004","8274","8648","166483199","8274","166491473","0.004970","8648","30652","28.213493","0.726732280864","0.717865065901","0.999950303761","0.999898379870","0.858907684831","0.724941356316","0.722271459051","0.719621155632","0.565277706417","0.722285066089","0.722234252664","0.000101620130","0.722220641604","0.980542486408","0.964998187466","0.972708239744","0.972708239744"
+"SNOWBALL_NORWEGIAN_NYNORSK_DIRECT","NN_NO","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4681","18219","23","18219","0","1","18219","6120","21971","8274","8624","165918002","8274","165926276","0.004987","8624","30595","28.187612","0.726434121342","0.718123876450","0.999950134480","0.999898178365","0.859037005465","0.724756721095","0.722255095332","0.719770679771","0.565257660346","0.722267047015","0.722216132089","0.000101821635","0.722204176866","0.980505813025","0.965064509051","0.972723884952","0.972723884952"
+"SNOWBALL_NORWEGIAN_NYNORSK_LUCENE_FILTER","NN_NO","ALL_WORDS","PRIMARY_OUTPUT","4688","18250","23","18250","0","1","18250","6144","21978","8295","8674","166483178","8295","166491473","0.004982","8674","30652","28.298317","0.725993459518","0.717016834138","0.999950177629","0.999898097625","0.858483505883","0.724180198229","0.721477226098","0.718794356395","0.564305338023","0.721491186328","0.721440231913","0.000101902375","0.721426267560","0.980461058483","0.964862123312","0.972599049418","0.972599049418"
+"SNOWBALL_NORWEGIAN_NYNORSK_LUCENE_FILTER","NN_NO","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4681","18219","23","18219","0","1","18219","6130","21948","8274","8647","165918002","8274","165926276","0.004987","8647","30595","28.262788","0.726225928132","0.717372119627","0.999950134480","0.999898039774","0.858661127054","0.724437725685","0.721771872996","0.719125568472","0.564665929147","0.721785448310","0.721734464790","0.000101960226","0.721720885459","0.980505813025","0.964944952705","0.972663150405","0.972663150405"
+"SNOWBALL_PORTUGUESE_DIRECT","PT_PT","ALL_WORDS","PRIMARY_OUTPUT","4001","211489","0","211489","0","1","211489","11315","4817239","167230","671821","22358036526","167230","22358203756","0.000748","671821","5489060","12.239272","0.966449786326","0.877607277020","0.999992520419","0.999962481554","0.938799898719","0.947270839082","0.919888415834","0.894044585092","0.851660540743","0.920957852105","0.920939611009","0.000037518446","0.919869695779","0.996663145176","0.967923515462","0.982083116554","0.982083116554"
+"SNOWBALL_PORTUGUESE_DIRECT","PT_PT","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4001","211489","0","211489","0","1","211489","11315","4817239","167230","671821","22358036526","167230","22358203756","0.000748","671821","5489060","12.239272","0.966449786326","0.877607277020","0.999992520419","0.999962481554","0.938799898719","0.947270839082","0.919888415834","0.894044585092","0.851660540743","0.920957852105","0.920939611009","0.000037518446","0.919869695779","0.996663145176","0.967923515462","0.982083116554","0.982083116554"
+"SNOWBALL_PORTUGUESE_LUCENE_FILTER","PT_PT","ALL_WORDS","PRIMARY_OUTPUT","4001","211489","0","211489","0","1","211489","11315","4817239","167230","671821","22358036526","167230","22358203756","0.000748","671821","5489060","12.239272","0.966449786326","0.877607277020","0.999992520419","0.999962481554","0.938799898719","0.947270839082","0.919888415834","0.894044585092","0.851660540743","0.920957852105","0.920939611009","0.000037518446","0.919869695779","0.996663145176","0.967923515462","0.982083116554","0.982083116554"
+"SNOWBALL_PORTUGUESE_LUCENE_FILTER","PT_PT","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","4001","211489","0","211489","0","1","211489","11315","4817239","167230","671821","22358036526","167230","22358203756","0.000748","671821","5489060","12.239272","0.966449786326","0.877607277020","0.999992520419","0.999962481554","0.938799898719","0.947270839082","0.919888415834","0.894044585092","0.851660540743","0.920957852105","0.920939611009","0.000037518446","0.919869695779","0.996663145176","0.967923515462","0.982083116554","0.982083116554"
+"SNOWBALL_RUSSIAN_DIRECT","RU_RU","ALL_WORDS","PRIMARY_OUTPUT","37410","768882","10","768882","0","1","768882","64358","8766656","3782908","4322849","295572508108","3782908","295576291016","0.001280","4322849","13089505","33.025305","0.698562595481","0.669746946122","0.999987201585","0.999972577645","0.834867073854","0.692602792505","0.683851352013","0.675318311031","0.519585195076","0.684003044583","0.683989349009","0.000027422355","0.683837646322","0.974179960240","0.953661001039","0.963811283954","0.963811283954"
+"SNOWBALL_RUSSIAN_DIRECT","RU_RU","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","37297","768133","10","768133","0","1","768133","64159","8764719","3782908","4322407","294996898744","3782908","295000681652","0.001282","4322407","13087126","33.027931","0.698516062041","0.669720685810","0.999987176613","0.999972525638","0.834853931211","0.692560581516","0.683815365804","0.675288254087","0.519543647630","0.683966853085","0.683953131513","0.000027474362","0.683801634112","0.974148936252","0.953633561396","0.963782086975","0.963782086975"
+"SNOWBALL_RUSSIAN_LUCENE_FILTER","RU_RU","ALL_WORDS","PRIMARY_OUTPUT","37410","768882","10","768882","0","1","768882","64266","8766889","3785790","4322616","295572505226","3785790","295576291016","0.001281","4322616","13089505","33.023525","0.698407806015","0.669764746642","0.999987191835","0.999972568683","0.834875969239","0.692484865100","0.683786451263","0.675303850911","0.519510266339","0.683936347366","0.683922647122","0.000027431317","0.683772741022","0.974131099393","0.953673855106","0.963793934411","0.963793934411"
+"SNOWBALL_RUSSIAN_LUCENE_FILTER","RU_RU","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","37297","768133","10","768133","0","1","768133","64159","8764719","3782908","4322407","294996898744","3782908","295000681652","0.001282","4322407","13087126","33.027931","0.698516062041","0.669720685810","0.999987176613","0.999972525638","0.834853931211","0.692560581516","0.683815365804","0.675288254087","0.519543647630","0.683966853085","0.683953131513","0.000027474362","0.683801634112","0.974148936252","0.953633561396","0.963782086975","0.963782086975"
+"SNOWBALL_SPANISH_DIRECT","ES_ES","ALL_WORDS","PRIMARY_OUTPUT","65059","871332","3589","871332","0","1","871332","195021","12811687","2228819","29161649","379565089291","2228819","379567318110","0.000587","29161649","41973336","69.476605","0.851812232913","0.305233946618","0.999994128001","0.999917308483","0.652614037309","0.627191552465","0.449423738186","0.350172671706","0.289843040458","0.509903921959","0.509876023351","0.000082691517","0.449391616998","0.981405614580","0.852462513401","0.912400934512","0.912400934512"
+"SNOWBALL_SPANISH_DIRECT","ES_ES","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","64918","869371","3525","869371","0","1","869371","194444","12787018","2201196","29076352","377858468569","2201196","377860669765","0.000583","29076352","41863370","69.455354","0.853138205793","0.305446455935","0.999994174583","0.999917233823","0.652720315259","0.627945981812","0.449838583213","0.350441220963","0.290188220622","0.510478247707","0.510450359519","0.000082766177","0.449806446076","0.981468761133","0.852555702466","0.912481600659","0.912481600659"
+"SNOWBALL_SPANISH_LUCENE_FILTER","ES_ES","ALL_WORDS","PRIMARY_OUTPUT","65059","871332","3589","871332","0","1","871332","194971","12811693","2230481","29161643","379565087629","2230481","379567318110","0.000588","29161643","41973336","69.476591","0.851718175843","0.305234089566","0.999994123622","0.999917304121","0.652614106594","0.627150877515","0.449410800675","0.350169642836","0.289832278511","0.509875888791","0.509847985527","0.000082695879","0.449378676109","0.981386049614","0.852460744495","0.912391466047","0.912391466047"
+"SNOWBALL_SPANISH_LUCENE_FILTER","ES_ES","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","64918","869371","3525","869371","0","1","869371","194444","12787018","2201196","29076352","377858468569","2201196","377860669765","0.000583","29076352","41863370","69.455354","0.853138205793","0.305446455935","0.999994174583","0.999917233823","0.652720315259","0.627945981812","0.449838583213","0.350441220963","0.290188220622","0.510478247707","0.510450359519","0.000082766177","0.449806446076","0.981468761133","0.852555702466","0.912481600659","0.912481600659"
+"SNOWBALL_SWEDISH_DIRECT","SV_SE","ALL_WORDS","PRIMARY_OUTPUT","12371","98108","68","98108","0","1","98108","25915","237017","67105","148325","4812088331","67105","4812155436","0.001394","148325","385342","38.491781","0.779348419384","0.615082186733","0.999986055105","0.999955235704","0.807534120919","0.739831942216","0.687539886056","0.642151948805","0.523855832838","0.692360693585","0.692339154006","0.000044764296","0.687517812951","0.984860422704","0.942685282143","0.963311451566","0.963311451566"
+"SNOWBALL_SWEDISH_DIRECT","SV_SE","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","12342","97881","68","97881","0","1","97881","25840","236588","67105","147975","4789844472","67105","4789911577","0.001401","147975","384563","38.478741","0.779036724587","0.615212591955","0.999985990347","0.999955100897","0.807599291151","0.739644914918","0.687500000000","0.642223301999","0.523809523810","0.692295603454","0.692273994517","0.000044899103","0.687477858823","0.984821307273","0.942694565973","0.963297587020","0.963297587020"
+"SNOWBALL_SWEDISH_LUCENE_FILTER","SV_SE","ALL_WORDS","PRIMARY_OUTPUT","12371","98108","68","98108","0","1","98108","26781","230676","64262","154666","4812091174","64262","4812155436","0.001335","154666","385342","40.137333","0.782116919488","0.598626674487","0.999986645901","0.999954508853","0.799306660194","0.736939762085","0.678179573117","0.628097931391","0.513064830384","0.684248529829","0.684226838572","0.000045491147","0.678157227687","0.985207247898","0.939659207875","0.961894327137","0.961894327137"
+"SNOWBALL_SWEDISH_LUCENE_FILTER","SV_SE","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","12342","97881","68","97881","0","1","97881","26706","230247","64262","154316","4789847315","64262","4789911577","0.001342","154316","384563","40.127625","0.781799537535","0.598723746174","0.999986583886","0.999954370671","0.799355165030","0.736743719918","0.678122496584","0.628142458291","0.512999498691","0.684165146635","0.684143384784","0.000045629329","0.678100081584","0.985169028543","0.939660518299","0.961876797342","0.961876797342"
+"SNOWBALL_YIDDISH_DIRECT","YI","ALL_WORDS","PRIMARY_OUTPUT","802","3578","0","3578","0","1","3578","1087","4962","1151","1382","6391758","1151","6392909","0.018004","1382","6344","21.784363","0.811712743334","0.782156368222","0.999819956768","0.999604172550","0.890988162495","0.805624107027","0.796660512162","0.787894185271","0.662041360907","0.796797522188","0.796599716782","0.000395827450","0.796462473806","0.982918530193","0.962014249878","0.972354049666","0.972354049666"
+"SNOWBALL_YIDDISH_DIRECT","YI","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","802","3578","0","3578","0","1","3578","1087","4962","1151","1382","6391758","1151","6392909","0.018004","1382","6344","21.784363","0.811712743334","0.782156368222","0.999819956768","0.999604172550","0.890988162495","0.805624107027","0.796660512162","0.787894185271","0.662041360907","0.796797522188","0.796599716782","0.000395827450","0.796462473806","0.982918530193","0.962014249878","0.972354049666","0.972354049666"
+"SNOWBALL_YIDDISH_LUCENE_FILTER","YI","ALL_WORDS","PRIMARY_OUTPUT","802","3578","0","3578","0","1","3578","1087","4962","1151","1382","6391758","1151","6392909","0.018004","1382","6344","21.784363","0.811712743334","0.782156368222","0.999819956768","0.999604172550","0.890988162495","0.805624107027","0.796660512162","0.787894185271","0.662041360907","0.796797522188","0.796599716782","0.000395827450","0.796462473806","0.982918530193","0.962014249878","0.972354049666","0.972354049666"
+"SNOWBALL_YIDDISH_LUCENE_FILTER","YI","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","802","3578","0","3578","0","1","3578","1087","4962","1151","1382","6391758","1151","6392909","0.018004","1382","6344","21.784363","0.811712743334","0.782156368222","0.999819956768","0.999604172550","0.890988162495","0.805624107027","0.796660512162","0.787894185271","0.662041360907","0.796797522188","0.796599716782","0.000395827450","0.796462473806","0.982918530193","0.962014249878","0.972354049666","0.972354049666"
+"SPANISH_LUCENE_SPANISH_LIGHT_STEM_FILTER","ES_ES","ALL_WORDS","PRIMARY_OUTPUT","65059","871332","3589","871332","0","1","871332","405552","1244317","147956","40729019","379567170154","147956","379567318110","0.000039","40729019","41973336","97.035458","0.893730611741","0.029645415842","0.999999610198","0.999892318297","0.514822513020","0.130863846499","0.057387272020","0.036752000024","0.029541282827","0.162772895888","0.162762055080","0.000107681703","0.057380579619","0.993823553768","0.756690454887","0.859195405761","0.859195405761"
+"SPANISH_LUCENE_SPANISH_LIGHT_STEM_FILTER","ES_ES","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","64918","869371","3525","869371","0","1","869371","404617","1241848","146613","40621522","377860523152","146613","377860669765","0.000039","40621522","41863370","97.033569","0.894406108634","0.029664310351","0.999999611992","0.999892119974","0.514831961171","0.130949068412","0.057424066047","0.036775459718","0.029560783207","0.162886280533","0.162875426941","0.000107880026","0.057417362063","0.993866319748","0.756725425267","0.859233931168","0.859233931168"
+"SPANISH_LUCENE_SPANISH_MINIMAL_STEM_FILTER","ES_ES","ALL_WORDS","PRIMARY_OUTPUT","65059","871332","3589","871332","0","1","871332","718633","148463","47859","41824873","379567270251","47859","379567318110","0.000013","41824873","41973336","99.646292","0.756221921130","0.003537078873","0.999999873912","0.999889695187","0.501768476392","0.017360591398","0.007041223811","0.004416184633","0.003533050405","0.051718628951","0.051713939443","0.000110304813","0.007040201537","0.995635307295","0.710609719902","0.829315972283","0.829315972283"
+"SPANISH_LUCENE_SPANISH_MINIMAL_STEM_FILTER","ES_ES","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","64918","869371","3525","869371","0","1","869371","717093","148226","47148","41715144","377860622617","47148","377860669765","0.000012","41715144","41863370","99.645929","0.758678227400","0.003540708739","0.999999875224","0.999889489251","0.501770291981","0.017379114288","0.007048522419","0.004420728101","0.003536725554","0.051829129163","0.051824445330","0.000110510749","0.007047500484","0.995675746140","0.710626433954","0.829341382913","0.829341382913"
+"SPANISH_LUCENE_SPANISH_PLURAL_STEM_FILTER","ES_ES","ALL_WORDS","PRIMARY_OUTPUT","65059","871332","3589","871332","0","1","871332","578805","325245","58578","41648091","379567259532","58578","379567318110","0.000015","41648091","41973336","99.225115","0.847382777999","0.007748847983","0.999999845672","0.999890132644","0.503874346827","0.037377069210","0.015357262275","0.009663967067","0.007738048760","0.081032341260","0.081026288458","0.000109867356","0.015355289170","0.995442321769","0.723731297627","0.838115191065","0.838115191065"
+"SPANISH_LUCENE_SPANISH_PLURAL_STEM_FILTER","ES_ES","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","64918","869371","3525","869371","0","1","869371","577533","324656","57716","41538714","377860612049","57716","377860669765","0.000015","41538714","41863370","99.224487","0.849057985417","0.007755132948","0.999999847256","0.999889928152","0.503877490102","0.037408921072","0.015369880354","0.009671831022","0.007744455857","0.081145286724","0.081139234944","0.000110071848","0.015367905834","0.995484117647","0.723752971494","0.838144538380","0.838144538380"
+"SPANISH_RADIXOR","ES_ES","ALL_WORDS","PRIMARY_OUTPUT","65059","871332","3589","871332","0","1","871332","64995","41074684","288483","898652","379567029627","288483","379567318110","0.000076","898652","41973336","2.141007","0.993025606574","0.978589931475","0.999999239969","0.999996872745","0.989294585722","0.990104500109","0.985754921826","0.981443392220","0.971909988067","0.985781345071","0.985779787115","0.000003127255","0.985753358111","0.995417814373","0.993266303762","0.994340895233","0.994340895233"
+"SPANISH_RADIXOR","ES_ES","ALL_WORDS","ANY_CANDIDATE","65059","871332","3589","828695","42637","21","916797","65118","41972710","2","626","379567318108","2","379567318110","0.000000","626","41973336","0.001491","0.999999952350","0.999985085770","0.999999999995","0.999999998346","0.999992542882","0.999996978999","0.999992519005","0.999988059050","0.999985038121","0.999992519032","0.999992518205","0.000000001654","","","","",""
+"SPANISH_RADIXOR","ES_ES","ALL_WORDS","ALL_CANDIDATES","65059","871332","3589","828695","42637","21","916797","65118","41972710","1349800","626","379565968310","1349800","379567318110","0.000356","626","41973336","0.001491","0.968842987168","0.999985085770","0.999996443846","0.999996442590","0.999990764808","0.974915259157","0.984167740127","0.993597526064","0.968828987818","0.984290880594","0.984289129583","0.000003557410","","","","",""
+"SPANISH_RADIXOR","ES_ES","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","64918","869371","3525","869371","0","1","869371","64814","40978337","276044","885033","377860393721","276044","377860669765","0.000073","885033","41863370","2.114099","0.993308734895","0.978859012067","0.999999269456","0.999996927576","0.989429140761","0.990384762162","0.986030938205","0.981715226337","0.972446769193","0.986057405488","0.986055874970","0.000003072424","0.986029401906","0.995463637710","0.993323040564","0.994392187139","0.994392187139"
+"SPANISH_RADIXOR","ES_ES","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","64918","869371","3525","826968","42403","21","914127","64933","41863370","0","0","377860669765","0","377860669765","0.000000","0","41863370","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"SPANISH_RADIXOR","ES_ES","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","64918","869371","3525","826968","42403","21","914127","64933","41863370","1255381","0","377859414384","1255381","377860669765","0.000332","0","41863370","0.000000","0.970885497124","1.000000000000","0.999996677662","0.999996678030","0.999998338831","0.976571978660","0.985227704543","0.994038240493","0.970885497124","0.985335220686","0.985333583876","0.000003321970","","","","",""
+"SWEDISH_LUCENE_SWEDISH_LIGHT_STEM_FILTER","SV_SE","ALL_WORDS","PRIMARY_OUTPUT","12371","98108","68","98108","0","1","98108","22392","218635","45941","166707","4812109495","45941","4812155436","0.000955","166707","385342","43.262089","0.826359911708","0.567379107390","0.999990453135","0.999955813777","0.783684780262","0.757232036109","0.672807954234","0.605320541501","0.506940918144","0.684733049508","0.684712936280","0.000044186223","0.672786622564","0.986795482859","0.942302523776","0.964035907715","0.964035907715"
+"SWEDISH_LUCENE_SWEDISH_LIGHT_STEM_FILTER","SV_SE","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","12342","97881","68","97881","0","1","97881","22338","218126","45941","166437","4789865636","45941","4789911577","0.000959","166437","384563","43.279515","0.826025213298","0.567204853301","0.999990408800","0.999955664954","0.783597631051","0.756945124029","0.672574503184","0.605125951621","0.506675896159","0.684489232882","0.684469049184","0.000044335046","0.672553099274","0.986761366955","0.942264620928","0.963999791833","0.963999791833"
+"SWEDISH_LUCENE_SWEDISH_MINIMAL_STEM_FILTER","SV_SE","ALL_WORDS","PRIMARY_OUTPUT","12371","98108","68","98108","0","1","98108","23360","228181","40227","157161","4812115209","40227","4812155436","0.000836","157161","385342","40.784809","0.850127417961","0.592151906618","0.999991640544","0.999958984659","0.796071773581","0.781991317186","0.698068068834","0.630412272016","0.536178622033","0.709510092538","0.709491456160","0.000041015341","0.698048215965","0.988492665376","0.944581755622","0.966038479572","0.966038479572"
+"SWEDISH_LUCENE_SWEDISH_MINIMAL_STEM_FILTER","SV_SE","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","12342","97881","68","97881","0","1","97881","23312","227624","40227","156939","4789871350","40227","4789911577","0.000840","156939","384563","40.809698","0.849815755775","0.591903017191","0.999991601724","0.999958840540","0.795947309457","0.781693541131","0.697790053555","0.630152322431","0.535850655618","0.709230928471","0.709212225121","0.000041159460","0.697770131116","0.988462934404","0.944527865197","0.965996098328","0.965996098328"
+"SWEDISH_RADIXOR","SV_SE","ALL_WORDS","PRIMARY_OUTPUT","12371","98108","68","98108","0","1","98108","12330","365796","24473","19546","4812130963","24473","4812155436","0.000509","19546","385342","5.072377","0.937291970410","0.949276227351","0.999994914337","0.999990853272","0.974635570844","0.939664553041","0.943246034417","0.946854921499","0.892588119029","0.943265066457","0.943260495884","0.000009146728","0.943241460869","0.992630770222","0.993394969179","0.993012722673","0.993012722673"
+"SWEDISH_RADIXOR","SV_SE","ALL_WORDS","ANY_CANDIDATE","12371","98108","68","92341","5767","5","104148","12371","385342","0","0","4812155436","0","4812155436","0.000000","0","385342","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"SWEDISH_RADIXOR","SV_SE","ALL_WORDS","ALL_CANDIDATES","12371","98108","68","92341","5767","5","104148","12371","385342","47848","0","4812107588","47848","4812155436","0.000994","0","385342","0.000000","0.889545003347","1.000000000000","0.999990056847","0.999990057643","0.999995028423","0.909639856815","0.941544130223","0.975767741439","0.889545003347","0.943156934634","0.943152245645","0.000009942357","","","","",""
+"SWEDISH_RADIXOR","SV_SE","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","12342","97881","68","97881","0","1","97881","12301","365017","24473","19546","4789887104","24473","4789911577","0.000511","19546","384563","5.082652","0.937166551131","0.949173477428","0.999994890720","0.999990810798","0.974584184074","0.939543572972","0.943131801052","0.946747541943","0.892383555482","0.943150907472","0.943146315681","0.000009189202","0.943127206266","0.992611730682","0.993377890892","0.992994663001","0.992994663001"
+"SWEDISH_RADIXOR","SV_SE","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","12342","97881","68","92114","5767","5","103921","12342","384563","0","0","4789911577","0","4789911577","0.000000","0","384563","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"SWEDISH_RADIXOR","SV_SE","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","12342","97881","68","92114","5767","5","103921","12342","384563","47848","0","4789863729","47848","4789911577","0.000999","0","384563","0.000000","0.889346015712","1.000000000000","0.999990010672","0.999990011473","0.999995005336","0.909473386475","0.941432652692","0.975719846569","0.889346015712","0.943051438529","0.943046728292","0.000009988527","","","","",""
+"UKRAINIAN_LUCENE_MORFOLOGIK_FILTER","UK_UA","ALL_WORDS","PRIMARY_OUTPUT","1493","14245","4","14245","0","1","14245","2358","56032","828","9308","101386722","828","101387550","0.000817","9308","65340","14.245485","0.985437917693","0.857545148454","0.999991833317","0.999900091560","0.928768490886","0.956895962839","0.917054009820","0.880397209478","0.846814169992","0.919270093835","0.919222898475","0.000099908440","0.917004266345","0.997989675681","0.970999849348","0.984309781661","0.984309781661"
+"UKRAINIAN_LUCENE_MORFOLOGIK_FILTER","UK_UA","ALL_WORDS","ANY_CANDIDATE","1493","14245","4","12038","2207","6","16937","2912","60394","122","4946","101387428","122","101387550","0.000120","4946","65340","7.569636","0.997984004230","0.924303642485","0.999998796696","0.999950045780","0.962151219591","0.982322936592","0.959731756929","0.938156308641","0.922581039382","0.960437530635","0.960413432420","0.000049954220","","","","",""
+"UKRAINIAN_LUCENE_MORFOLOGIK_FILTER","UK_UA","ALL_WORDS","ALL_CANDIDATES","1493","14245","4","12038","2207","6","16937","2912","60394","1368","4946","101386182","1368","101387550","0.001349","4946","65340","7.569636","0.977850458211","0.924303642485","0.999986507219","0.999937764217","0.962145074852","0.966650447520","0.950323362339","0.934538657225","0.905348683816","0.950700131656","0.950669478973","0.000062235783","","","","",""
+"UKRAINIAN_LUCENE_MORFOLOGIK_FILTER","UK_UA","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","1491","14236","4","14236","0","1","14236","2356","56016","828","9308","101258578","828","101259406","0.000818","9308","65324","14.248974","0.985433818873","0.857510256567","0.999991822982","0.999899965191","0.928751039775","0.956884181756","0.917032283413","0.880367133966","0.846777119361","0.919249480202","0.919202225823","0.000100034809","0.916982477117","0.997988093697","0.970977645923","0.984297603943","0.984297603943"
+"UKRAINIAN_LUCENE_MORFOLOGIK_FILTER","UK_UA","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","1491","14236","4","12029","2207","6","16928","2910","60378","122","4946","101259284","122","101259406","0.000120","4946","65324","7.571490","0.997983471074","0.924285101953","0.999998795174","0.999949982596","0.962141948563","0.982318335047","0.959721515768","0.938140934008","0.922562112276","0.960427641371","0.960403512884","0.000050017404","","","","",""
+"UKRAINIAN_LUCENE_MORFOLOGIK_FILTER","UK_UA","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","1491","14236","4","12029","2207","6","16928","2910","60378","1368","4946","101258038","1368","101259406","0.001351","4946","65324","7.571490","0.977844718686","0.924285101953","0.999986490144","0.999937685499","0.962135796049","0.966641904786","0.950310852286","0.934522445998","0.905325976129","0.950687806541","0.950657115187","0.000062314501","","","","",""
+"UKRAINIAN_MORFOLOGIK_DIRECT","UK_UA","ALL_WORDS","PRIMARY_OUTPUT","1493","14245","4","14245","0","1","14245","2365","56016","828","9324","101386722","828","101387550","0.000817","9324","65340","14.269972","0.985433818873","0.857300275482","0.999991833317","0.999899933851","0.928646054399","0.956831877998","0.916912197996","0.880190066750","0.846572361262","0.919136923635","0.919089660097","0.000100066149","0.916862376978","0.997989675681","0.970875945953","0.984246115917","0.984246115917"
+"UKRAINIAN_MORFOLOGIK_DIRECT","UK_UA","ALL_WORDS","ANY_CANDIDATE","1493","14245","4","12038","2207","6","16937","2919","60378","122","4962","101387428","122","101387550","0.000120","4962","65340","7.594123","0.997983471074","0.924058769513","0.999998796696","0.999949888071","0.962028783105","0.982267195939","0.959599491418","0.937954390108","0.922336622774","0.960310042786","0.960285871670","0.000050111929","","","","",""
+"UKRAINIAN_MORFOLOGIK_DIRECT","UK_UA","ALL_WORDS","ALL_CANDIDATES","1493","14245","4","12038","2207","6","16937","2919","60378","1368","4962","101386182","1368","101387550","0.001349","4962","65340","7.594123","0.977844718686","0.924058769513","0.999986507219","0.999937606509","0.962022638366","0.966592384831","0.950191209102","0.934337338211","0.905108832524","0.950571400540","0.950540673331","0.000062393491","","","","",""
+"UKRAINIAN_MORFOLOGIK_DIRECT","UK_UA","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","1491","14236","4","14236","0","1","14236","2356","56016","828","9308","101258578","828","101259406","0.000818","9308","65324","14.248974","0.985433818873","0.857510256567","0.999991822982","0.999899965191","0.928751039775","0.956884181756","0.917032283413","0.880367133966","0.846777119361","0.919249480202","0.919202225823","0.000100034809","0.916982477117","0.997988093697","0.970977645923","0.984297603943","0.984297603943"
+"UKRAINIAN_MORFOLOGIK_DIRECT","UK_UA","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","1491","14236","4","12029","2207","6","16928","2910","60378","122","4946","101259284","122","101259406","0.000120","4946","65324","7.571490","0.997983471074","0.924285101953","0.999998795174","0.999949982596","0.962141948563","0.982318335047","0.959721515768","0.938140934008","0.922562112276","0.960427641371","0.960403512884","0.000050017404","","","","",""
+"UKRAINIAN_MORFOLOGIK_DIRECT","UK_UA","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","1491","14236","4","12029","2207","6","16928","2910","60378","1368","4946","101258038","1368","101259406","0.001351","4946","65324","7.571490","0.977844718686","0.924285101953","0.999986490144","0.999937685499","0.962135796049","0.966641904786","0.950310852286","0.934522445998","0.905325976129","0.950687806541","0.950657115187","0.000062314501","","","","",""
+"UKRAINIAN_RADIXOR","UK_UA","ALL_WORDS","PRIMARY_OUTPUT","1493","14245","4","14245","0","1","14245","1493","64732","880","608","101386670","880","101387550","0.000868","608","65340","0.930517","0.986587819301","0.990694827058","0.999991320433","0.999985333094","0.995343073746","0.987406494442","0.988637057853","0.989870692292","0.977529447297","0.988639190514","0.988631855097","0.000014666906","0.988629719696","0.997993591453","0.998265712624","0.998129633491","0.998129633491"
+"UKRAINIAN_RADIXOR","UK_UA","ALL_WORDS","ANY_CANDIDATE","1493","14245","4","14055","190","2","14435","1493","65340","0","0","101387550","0","101387550","0.000000","0","65340","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"UKRAINIAN_RADIXOR","UK_UA","ALL_WORDS","ALL_CANDIDATES","1493","14245","4","14055","190","2","14435","1493","65340","1490","0","101386060","1490","101387550","0.001470","0","65340","0.000000","0.977704623672","1.000000000000","0.999985303916","0.999985313380","0.999992651958","0.982083809295","0.988726639933","0.995459946982","0.977704623672","0.988789473888","0.988782208195","0.000014686620","","","","",""
+"UKRAINIAN_RADIXOR","UK_UA","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","1491","14236","4","14236","0","1","14236","1491","64716","880","608","101258526","880","101259406","0.000869","608","65324","0.930745","0.986584547838","0.990692547915","0.999991309449","0.999985314543","0.995341928682","0.987403420118","0.988634280477","0.989868213355","0.977524016676","0.988636414174","0.988629069474","0.000014685457","0.988626933033","0.997992012551","0.998264347489","0.998128161444","0.998128161444"
+"UKRAINIAN_RADIXOR","UK_UA","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","1491","14236","4","14046","190","2","14426","1491","65324","0","0","101259406","0","101259406","0.000000","0","65324","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"UKRAINIAN_RADIXOR","UK_UA","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","1491","14236","4","14046","190","2","14426","1491","65324","1490","0","101257916","1490","101259406","0.001471","0","65324","0.000000","0.977699284581","1.000000000000","0.999985285318","0.999985294804","0.999992642659","0.982079499669","0.988723909852","0.995458840023","0.977699284581","0.988786774073","0.988779499204","0.000014705196","","","","",""
+"YI_RADIXOR","YI","ALL_WORDS","PRIMARY_OUTPUT","802","3578","0","3578","0","1","3578","802","6195","195","149","6392714","195","6392909","0.003050","149","6344","2.348676","0.969483568075","0.976513240858","0.999969497454","0.999946243726","0.988241369156","0.970881394183","0.972985707555","0.975099162627","0.947392567671","0.972992055990","0.972965163911","0.000053756274","0.972958803000","0.995691103897","0.996142223728","0.995916612726","0.995916612726"
+"YI_RADIXOR","YI","ALL_WORDS","ANY_CANDIDATE","802","3578","0","3489","89","3","3676","802","6344","0","0","6392909","0","6392909","0.000000","0","6344","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"YI_RADIXOR","YI","ALL_WORDS","ALL_CANDIDATES","802","3578","0","3489","89","3","3676","802","6344","389","0","6392520","389","6392909","0.006085","0","6344","0.000000","0.942224862617","1.000000000000","0.999939151332","0.999939211655","0.999969575666","0.953239572064","0.970253116158","0.987885016662","0.942224862617","0.970682678643","0.970653145819","0.000060788345","","","","",""
+"YI_RADIXOR","YI","LOWERCASE_GROUPS_ONLY","PRIMARY_OUTPUT","802","3578","0","3578","0","1","3578","802","6195","195","149","6392714","195","6392909","0.003050","149","6344","2.348676","0.969483568075","0.976513240858","0.999969497454","0.999946243726","0.988241369156","0.970881394183","0.972985707555","0.975099162627","0.947392567671","0.972992055990","0.972965163911","0.000053756274","0.972958803000","0.995691103897","0.996142223728","0.995916612726","0.995916612726"
+"YI_RADIXOR","YI","LOWERCASE_GROUPS_ONLY","ANY_CANDIDATE","802","3578","0","3489","89","3","3676","802","6344","0","0","6392909","0","6392909","0.000000","0","6344","0.000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","1.000000000000","0.000000000000","","","","",""
+"YI_RADIXOR","YI","LOWERCASE_GROUPS_ONLY","ALL_CANDIDATES","802","3578","0","3489","89","3","3676","802","6344","389","0","6392520","389","6392909","0.006085","0","6344","0.000000","0.942224862617","1.000000000000","0.999939151332","0.999939211655","0.999969575666","0.953239572064","0.970253116158","0.987885016662","0.942224862617","0.970682678643","0.970653145819","0.000060788345","","","","",""
--- a/docs/benchmarks/data/stemming-quality.sha256
+++ b/docs/benchmarks/data/stemming-quality.sha256
@@ -0,0 +1 @@
+5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28  stemming-quality.csv
--- a/docs/benchmarks/index.md
+++ b/docs/benchmarks/index.md
@@ -0,0 +1,327 @@
+# Benchmark Results
+
+This section contains the published Radixor benchmark reference set. It is intentionally split into
+two layers:
+
+- **benchmark reference pages**, which explain methodology, corpora, environment, candidate
+  selection, and the English dictionary coverage experiment;
+- **language result pages**, which contain the actual same-language accuracy and throughput tables.
+- **pairwise quality pages and generated sections**, which publish over-stemming, under-stemming,
+  candidate-policy, classification, and partition measurements from one checked result snapshot.
+
+This structure keeps methodology separate from per-language result pages, while preserving all
+measured data and the command-class analysis for each Radixor language resource.
+
+## Read This First
+
+Start with [Benchmarking](../benchmarking.md) for the high-level interpretation model. The most
+important rule is that speed and exact-root quality must be read together. Many competing stemmers
+are intentionally light, minimal, or aggressive; they can be fast because they are not trying to
+match dictionary roots with the same precision.
+
+Radixor rows in the refreshed tables use contracted compiled patch tries. Contraction collapses
+uniform preferred-command subtrees into accepting leaves, reducing hot lookup depth while preserving
+the preferred result measured by the accuracy pass.
+
+## Reference Pages
+
+| Page | Purpose |
+| --- | --- |
+| [Methodology](reference/methodology.md) | Workload design, normalization, speed metrics, quality metrics, and interpretation rules. |
+| [Linguistic quality methodology](reference/linguistic-quality.md) | Gold-standard groups, output policies, pairwise formulas, ranking rules, aggregation, and limitations. |
+| [Tested stemmers](reference/tested-stemmers.md) | Versions, upstream attribution, evaluated coverage, adapters, preprocessing, and output capability. |
+| [Reproducibility and raw data](reference/reproducibility.md) | Commands, versioned CSV snapshot, checksum, generated artifacts, and unavailable provenance. |
+| [Corpora](reference/corpora.md) | Dictionary row counts, complete quality tokens, already-root tokens, changed speed tokens, and timing token counts. |
+| [Environment and reports](reference/environment.md) | Hardware, JVM, JMH settings, report files, and badge/report policy. |
+| [English dictionary coverage](reference/english-coverage.md) | Quality/speed operating curve for contracted Radixor tries built from 100% down to 10% of English dictionary rows. |
+| [Candidate evaluation](reference/candidates.md) | Included benchmark families and evaluated candidates that were skipped. |
+
+## Language Results
+
+Each language page contains:
+
+- the dictionary corpus size,
+- the Radixor patch-command distribution,
+- exact-root quality metrics,
+- throughput metrics,
+- interpretation notes for the compared stemmers.
+
+Open [Language Benchmark Pages](languages/index.md) for the complete language list.
+
+## Key Published Result
+
+The English dictionary coverage benchmark shows the current contracted-trie operating curve. With
+the full English dictionary, Radixor reaches `97.478%` all-token exactness and `97.197%`
+changed-token exactness at `135.8 ns/token`. Even with a deterministic 10% dictionary slice, it
+keeps `92.868%` all-token exactness and `76.516%` changed-token exactness at `86.0 ns/token`.
+
+Those figures should not be reduced to a single speed badge. The professional interpretation is a
+quality/speed envelope: the amount and quality of dictionary knowledge affect stemming precision,
+while contracted tries reduce lookup cost in uniform regions of the compiled graph.
+
+## Quality versus performance
+
+Each language page keeps exact-root accuracy, JMH latency, and pairwise linguistic-quality results in separate tables. No undocumented scalar combines them. The current repository checkout does not contain the dated machine-readable JMH CSV files named by the performance provenance page, so this revision preserves the existing performance tables but does not regenerate a cross-language Pareto frontier from rounded Markdown values. A defensible Pareto analysis requires the original unrounded JMH snapshot on the same hardware and JVM. Readers can still inspect the quality and speed dimensions side by side on every language page.
+
+<!-- STEMMING-QUALITY-OVERVIEW:START -->
+
+## Pairwise Quality Findings
+
+The validated snapshot is a broad multilingual comparison covering the complete 20-language Radixor dictionary universe; 19 languages have existing benchmark pages. The direct ranking below uses only deterministic `PRIMARY_OUTPUT` rows over identical per-language inputs. Candidate-aware rows are intentionally excluded from this claim.
+
+!!! success "Evidence-based primary-output result"
+    Radixor achieved the highest balanced accuracy among the evaluated deterministic stemmers for every documented language in both `ALL_WORDS` and `LOWERCASE_GROUPS_ONLY`: **38 wins in 38 language-mode comparisons, with no exact first-place ties**. This statement is limited to the evaluated implementations, versions, dictionaries, adapters, and balanced-accuracy metric; it is not a universal claim about every stemming use case.
+
+### Per-language winner matrix
+
+| Language | Dictionary mode | Winner | Balanced accuracy | Runner-up | Difference | Exact tie | Deterministic stemmers |
+|---|---|---|---:|---|---:|---|---:|
+|Czech (`CS_CZ`)|ALL_WORDS|Radixor|0.996565|HUNSPELL CZECH LUCENE FILTER|0.142812638|no|3|
+|Czech (`CS_CZ`)|LOWERCASE_GROUPS_ONLY|Radixor|0.997139|HUNSPELL CZECH LUCENE FILTER|0.144369049|no|3|
+|Danish (`DA_DK`)|ALL_WORDS|Radixor|0.996066|SNOWBALL DANISH LUCENE FILTER|0.058096771|no|3|
+|Danish (`DA_DK`)|LOWERCASE_GROUPS_ONLY|Radixor|0.996305|SNOWBALL DANISH DIRECT|0.058230346|no|3|
+|Dutch (`NL_NL`)|ALL_WORDS|Radixor|0.988661|SNOWBALL DUTCH DIRECT|0.261574077|no|4|
+|Dutch (`NL_NL`)|LOWERCASE_GROUPS_ONLY|Radixor|0.989040|SNOWBALL DUTCH DIRECT|0.258544404|no|4|
+|English (`US_UK`)|ALL_WORDS|Radixor|0.965159|ENGLISH LUCENE PORTER COPIED|0.010532535|no|11|
+|English (`US_UK`)|LOWERCASE_GROUPS_ONLY|Radixor|0.965820|ENGLISH LUCENE PORTER COPIED|0.010920064|no|11|
+|Finnish (`FI_FI`)|ALL_WORDS|Radixor|0.984594|SNOWBALL FINNISH LUCENE FILTER|0.244241861|no|4|
+|Finnish (`FI_FI`)|LOWERCASE_GROUPS_ONLY|Radixor|0.988068|SNOWBALL FINNISH DIRECT|0.249668284|no|4|
+|French (`FR_FR`)|ALL_WORDS|Radixor|0.956992|SNOWBALL FRENCH DIRECT|0.111730673|no|6|
+|French (`FR_FR`)|LOWERCASE_GROUPS_ONLY|Radixor|0.957224|SNOWBALL FRENCH DIRECT|0.111809799|no|6|
+|German (`DE_DE`)|ALL_WORDS|Radixor|0.907901|GERMAN CISTEM|0.027131083|no|8|
+|German (`DE_DE`)|LOWERCASE_GROUPS_ONLY|Radixor|0.966157|GERMAN CISTEM|0.050868631|no|8|
+|Hungarian (`HU_HU`)|ALL_WORDS|Radixor|0.995491|SNOWBALL HUNGARIAN LUCENE FILTER|0.172884951|no|4|
+|Hungarian (`HU_HU`)|LOWERCASE_GROUPS_ONLY|Radixor|0.996163|SNOWBALL HUNGARIAN DIRECT|0.174455479|no|4|
+|Italian (`IT_IT`)|ALL_WORDS|Radixor|0.996507|SNOWBALL ITALIAN DIRECT|0.130318040|no|4|
+|Italian (`IT_IT`)|LOWERCASE_GROUPS_ONLY|Radixor|0.996512|SNOWBALL ITALIAN DIRECT|0.130307087|no|4|
+|Norwegian Bokmal (`NB_NO`)|ALL_WORDS|Radixor|0.974783|SNOWBALL NORWEGIAN BOKMAL DIRECT|0.099819340|no|5|
+|Norwegian Bokmal (`NB_NO`)|LOWERCASE_GROUPS_ONLY|Radixor|0.975000|SNOWBALL NORWEGIAN BOKMAL DIRECT|0.100008544|no|5|
+|Norwegian Nynorsk (`NN_NO`)|ALL_WORDS|Radixor|0.935777|SNOWBALL NORWEGIAN NYNORSK DIRECT|0.076868986|no|3|
+|Norwegian Nynorsk (`NN_NO`)|LOWERCASE_GROUPS_ONLY|Radixor|0.935853|SNOWBALL NORWEGIAN NYNORSK DIRECT|0.076816096|no|3|
+|Persian (`FA_IR`)|ALL_WORDS|Radixor|0.974922|PERSIAN LUCENE PERSIAN STEM FILTER|0.472751327|no|2|
+|Persian (`FA_IR`)|LOWERCASE_GROUPS_ONLY|Radixor|0.974922|PERSIAN LUCENE PERSIAN STEM FILTER|0.472751327|no|2|
+|Polish (`PL_PL`)|ALL_WORDS|Radixor|0.990388|POLISH LUCENE MORFOLOGIK FILTER|0.042233990|no|5|
+|Polish (`PL_PL`)|LOWERCASE_GROUPS_ONLY|Radixor|0.990579|POLISH LUCENE MORFOLOGIK FILTER|0.042401633|no|5|
+|Portuguese (`PT_PT`)|ALL_WORDS|Radixor|0.998502|SNOWBALL PORTUGUESE DIRECT|0.059701750|no|6|
+|Portuguese (`PT_PT`)|LOWERCASE_GROUPS_ONLY|Radixor|0.998502|SNOWBALL PORTUGUESE DIRECT|0.059701750|no|6|
+|Russian (`RU_RU`)|ALL_WORDS|Radixor|0.989827|SNOWBALL RUSSIAN LUCENE FILTER|0.154951419|no|4|
+|Russian (`RU_RU`)|LOWERCASE_GROUPS_ONLY|Radixor|0.989852|SNOWBALL RUSSIAN DIRECT|0.154997931|no|4|
+|Spanish (`ES_ES`)|ALL_WORDS|Radixor|0.989295|SNOWBALL SPANISH LUCENE FILTER|0.336680479|no|7|
+|Spanish (`ES_ES`)|LOWERCASE_GROUPS_ONLY|Radixor|0.989429|SNOWBALL SPANISH DIRECT|0.336708826|no|7|
+|Swedish (`SV_SE`)|ALL_WORDS|Radixor|0.974636|SNOWBALL SWEDISH DIRECT|0.167101450|no|5|
+|Swedish (`SV_SE`)|LOWERCASE_GROUPS_ONLY|Radixor|0.974584|SNOWBALL SWEDISH DIRECT|0.166984893|no|5|
+|Ukrainian (`UK_UA`)|ALL_WORDS|Radixor|0.995343|UKRAINIAN LUCENE MORFOLOGIK FILTER|0.066574583|no|4|
+|Ukrainian (`UK_UA`)|LOWERCASE_GROUPS_ONLY|Radixor|0.995342|UKRAINIAN LUCENE MORFOLOGIK FILTER|0.066590889|no|4|
+|Yiddish (`YI`)|ALL_WORDS|Radixor|0.988241|SNOWBALL YIDDISH DIRECT|0.097253207|no|3|
+|Yiddish (`YI`)|LOWERCASE_GROUPS_ONLY|Radixor|0.988241|SNOWBALL YIDDISH DIRECT|0.097253207|no|3|
+
+### Secondary-metric trade-offs
+
+Balanced-accuracy leadership does not imply leadership on every error trade-off. The table below lists all **15** deterministic primary-output language-mode-metric cases where a non-Radixor adapter has the best displayed value. Equal values are resolved by the authoritative row ordering and should be read as ties when the unrounded values are equal. Throughput leadership remains in the separate performance tables.
+
+<details class="quality-details" markdown="1"><summary>Non-Radixor secondary-metric leaders</summary>
+
+| Language | Dictionary mode | Metric | Leader | Value |
+|---|---|---|---|---:|
+|English|ALL_WORDS|Over-stemming percentage|ENGLISH LUCENE POSSESSIVE FILTER|0.000604|
+|English|LOWERCASE_GROUPS_ONLY|Over-stemming percentage|ENGLISH LUCENE POSSESSIVE FILTER|0.000653|
+|French|ALL_WORDS|Over-stemming percentage|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|0.000177|
+|French|LOWERCASE_GROUPS_ONLY|Over-stemming percentage|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|0.000166|
+|German|LOWERCASE_GROUPS_ONLY|Over-stemming percentage|GERMAN LUCENE GERMAN MINIMAL STEM FILTER|0.000188|
+|Italian|ALL_WORDS|Over-stemming percentage|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|0.000020|
+|Italian|LOWERCASE_GROUPS_ONLY|Over-stemming percentage|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|0.000020|
+|Persian|ALL_WORDS|Over-stemming percentage|PERSIAN LUCENE PERSIAN STEM FILTER|0.002652|
+|Persian|LOWERCASE_GROUPS_ONLY|Over-stemming percentage|PERSIAN LUCENE PERSIAN STEM FILTER|0.002652|
+|Portuguese|ALL_WORDS|Over-stemming percentage|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|0.000003|
+|Portuguese|LOWERCASE_GROUPS_ONLY|Over-stemming percentage|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|0.000003|
+|Spanish|ALL_WORDS|Over-stemming percentage|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|0.000013|
+|Spanish|LOWERCASE_GROUPS_ONLY|Over-stemming percentage|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|0.000012|
+|Ukrainian|ALL_WORDS|Over-stemming percentage|HUNSPELL UKRAINIAN LUCENE FILTER|0.000783|
+|Ukrainian|LOWERCASE_GROUPS_ONLY|Over-stemming percentage|HUNSPELL UKRAINIAN LUCENE FILTER|0.000784|
+
+</details>
+
+### Win, tie, and placement summary
+
+Counts use `PRIMARY_OUTPUT` only and retain each adapter configuration as a separate stemmer except that language-specific Radixor identifiers are combined as Radixor. Coverage is displayed explicitly; unsupported languages are absent, not losses.
+
+<details class="quality-details" markdown="1"><summary>ALL_WORDS placements</summary>
+
+| Stemmer | Evaluated languages | Wins | Exact first-place ties | Top-three placements | Average rank | Median rank |
+|---|---:|---:|---:|---:|---:|---:|
+|Radixor|19|19|0|19|1.000|1.000|
+|CZECH LUCENE CZECH STEM FILTER|1|0|0|1|3.000|3.000|
+|ENGLISH LUCENE KSTEM FILTER|1|0|0|0|8.000|8.000|
+|ENGLISH LUCENE MINIMAL FILTER|1|0|0|0|9.000|9.000|
+|ENGLISH LUCENE PORTER COPIED|1|0|0|1|2.000|2.000|
+|ENGLISH LUCENE PORTER FILTER|1|0|0|1|3.000|3.000|
+|ENGLISH LUCENE POSSESSIVE FILTER|1|0|0|0|11.000|11.000|
+|ENGLISH OPENNLP PORTER|1|0|0|0|4.000|4.000|
+|ENGLISH PAICE HUSK LANCASTER|1|0|0|0|7.000|7.000|
+|ENGLISH SNOWBALL ORIGINAL PORTER|1|0|0|0|6.000|6.000|
+|ENGLISH SNOWBALL PORTER2|1|0|0|0|5.000|5.000|
+|FINNISH LUCENE FINNISH LIGHT STEM FILTER|1|0|0|0|4.000|4.000|
+|FRENCH LUCENE FRENCH LIGHT STEM FILTER|1|0|0|0|5.000|5.000|
+|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|1|0|0|0|6.000|6.000|
+|GERMAN CISTEM|1|0|0|1|2.000|2.000|
+|GERMAN LUCENE GERMAN LIGHT STEM FILTER|1|0|0|0|5.000|5.000|
+|GERMAN LUCENE GERMAN MINIMAL STEM FILTER|1|0|0|0|8.000|8.000|
+|GERMAN LUCENE GERMAN STEM FILTER|1|0|0|0|6.000|6.000|
+|HUNGARIAN LUCENE HUNGARIAN LIGHT STEM FILTER|1|0|0|0|4.000|4.000|
+|HUNSPELL CZECH LUCENE FILTER|1|0|0|1|2.000|2.000|
+|HUNSPELL DUTCH LUCENE FILTER|1|0|0|1|3.000|3.000|
+|HUNSPELL ENGLISH LUCENE FILTER|1|0|0|0|10.000|10.000|
+|HUNSPELL FRENCH LUCENE FILTER|1|0|0|0|4.000|4.000|
+|HUNSPELL GERMAN LUCENE FILTER|1|0|0|0|7.000|7.000|
+|HUNSPELL POLISH LUCENE FILTER|1|0|0|1|3.000|3.000|
+|HUNSPELL SPANISH LUCENE FILTER|1|0|0|0|4.000|4.000|
+|HUNSPELL UKRAINIAN LUCENE FILTER|1|0|0|0|4.000|4.000|
+|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|1|0|0|0|4.000|4.000|
+|NORWEGIAN BOKMAL LUCENE NORWEGIAN LIGHT STEM FILTER|1|0|0|0|4.000|4.000|
+|NORWEGIAN BOKMAL LUCENE NORWEGIAN MINIMAL STEM FILTER|1|0|0|0|5.000|5.000|
+|PERSIAN LUCENE PERSIAN STEM FILTER|1|0|0|1|2.000|2.000|
+|POLISH LUCENE MORFOLOGIK FILTER|1|0|0|1|2.000|2.000|
+|POLISH LUCENE STEMPEL DIRECT|1|0|0|0|4.000|4.000|
+|POLISH LUCENE STEMPEL FILTER|1|0|0|0|5.000|5.000|
+|PORTUGUESE LUCENE PORTUGUESE LIGHT STEM FILTER|1|0|0|0|5.000|5.000|
+|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|1|0|0|0|6.000|6.000|
+|PORTUGUESE LUCENE PORTUGUESE STEM FILTER|1|0|0|0|4.000|4.000|
+|RUSSIAN LUCENE RUSSIAN LIGHT STEM FILTER|1|0|0|0|4.000|4.000|
+|SNOWBALL DANISH DIRECT|1|0|0|1|3.000|3.000|
+|SNOWBALL DANISH LUCENE FILTER|1|0|0|1|2.000|2.000|
+|SNOWBALL DUTCH DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL DUTCH LUCENE FILTER|1|0|0|0|4.000|4.000|
+|SNOWBALL FINNISH DIRECT|1|0|0|1|3.000|3.000|
+|SNOWBALL FINNISH LUCENE FILTER|1|0|0|1|2.000|2.000|
+|SNOWBALL FRENCH DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL FRENCH LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL GERMAN DIRECT|1|0|0|1|3.000|3.000|
+|SNOWBALL GERMAN LUCENE FILTER|1|0|0|0|4.000|4.000|
+|SNOWBALL HUNGARIAN DIRECT|1|0|0|1|3.000|3.000|
+|SNOWBALL HUNGARIAN LUCENE FILTER|1|0|0|1|2.000|2.000|
+|SNOWBALL ITALIAN DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL ITALIAN LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL NORWEGIAN BOKMAL DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL NORWEGIAN BOKMAL LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL NORWEGIAN NYNORSK DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL NORWEGIAN NYNORSK LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL PORTUGUESE DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL PORTUGUESE LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL RUSSIAN DIRECT|1|0|0|1|3.000|3.000|
+|SNOWBALL RUSSIAN LUCENE FILTER|1|0|0|1|2.000|2.000|
+|SNOWBALL SPANISH DIRECT|1|0|0|1|3.000|3.000|
+|SNOWBALL SPANISH LUCENE FILTER|1|0|0|1|2.000|2.000|
+|SNOWBALL SWEDISH DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL SWEDISH LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL YIDDISH DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL YIDDISH LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SPANISH LUCENE SPANISH LIGHT STEM FILTER|1|0|0|0|5.000|5.000|
+|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|1|0|0|0|7.000|7.000|
+|SPANISH LUCENE SPANISH PLURAL STEM FILTER|1|0|0|0|6.000|6.000|
+|SWEDISH LUCENE SWEDISH LIGHT STEM FILTER|1|0|0|0|5.000|5.000|
+|SWEDISH LUCENE SWEDISH MINIMAL STEM FILTER|1|0|0|0|4.000|4.000|
+|UKRAINIAN LUCENE MORFOLOGIK FILTER|1|0|0|1|2.000|2.000|
+|UKRAINIAN MORFOLOGIK DIRECT|1|0|0|1|3.000|3.000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>LOWERCASE_GROUPS_ONLY placements</summary>
+
+| Stemmer | Evaluated languages | Wins | Exact first-place ties | Top-three placements | Average rank | Median rank |
+|---|---:|---:|---:|---:|---:|---:|
+|Radixor|19|19|0|19|1.000|1.000|
+|CZECH LUCENE CZECH STEM FILTER|1|0|0|1|3.000|3.000|
+|ENGLISH LUCENE KSTEM FILTER|1|0|0|0|8.000|8.000|
+|ENGLISH LUCENE MINIMAL FILTER|1|0|0|0|9.000|9.000|
+|ENGLISH LUCENE PORTER COPIED|1|0|0|1|2.000|2.000|
+|ENGLISH LUCENE PORTER FILTER|1|0|0|1|3.000|3.000|
+|ENGLISH LUCENE POSSESSIVE FILTER|1|0|0|0|11.000|11.000|
+|ENGLISH OPENNLP PORTER|1|0|0|0|4.000|4.000|
+|ENGLISH PAICE HUSK LANCASTER|1|0|0|0|7.000|7.000|
+|ENGLISH SNOWBALL ORIGINAL PORTER|1|0|0|0|6.000|6.000|
+|ENGLISH SNOWBALL PORTER2|1|0|0|0|5.000|5.000|
+|FINNISH LUCENE FINNISH LIGHT STEM FILTER|1|0|0|0|4.000|4.000|
+|FRENCH LUCENE FRENCH LIGHT STEM FILTER|1|0|0|0|5.000|5.000|
+|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|1|0|0|0|6.000|6.000|
+|GERMAN CISTEM|1|0|0|1|2.000|2.000|
+|GERMAN LUCENE GERMAN LIGHT STEM FILTER|1|0|0|0|5.000|5.000|
+|GERMAN LUCENE GERMAN MINIMAL STEM FILTER|1|0|0|0|8.000|8.000|
+|GERMAN LUCENE GERMAN STEM FILTER|1|0|0|0|6.000|6.000|
+|HUNGARIAN LUCENE HUNGARIAN LIGHT STEM FILTER|1|0|0|0|4.000|4.000|
+|HUNSPELL CZECH LUCENE FILTER|1|0|0|1|2.000|2.000|
+|HUNSPELL DUTCH LUCENE FILTER|1|0|0|1|3.000|3.000|
+|HUNSPELL ENGLISH LUCENE FILTER|1|0|0|0|10.000|10.000|
+|HUNSPELL FRENCH LUCENE FILTER|1|0|0|0|4.000|4.000|
+|HUNSPELL GERMAN LUCENE FILTER|1|0|0|0|7.000|7.000|
+|HUNSPELL POLISH LUCENE FILTER|1|0|0|1|3.000|3.000|
+|HUNSPELL SPANISH LUCENE FILTER|1|0|0|0|4.000|4.000|
+|HUNSPELL UKRAINIAN LUCENE FILTER|1|0|0|0|4.000|4.000|
+|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|1|0|0|0|4.000|4.000|
+|NORWEGIAN BOKMAL LUCENE NORWEGIAN LIGHT STEM FILTER|1|0|0|0|4.000|4.000|
+|NORWEGIAN BOKMAL LUCENE NORWEGIAN MINIMAL STEM FILTER|1|0|0|0|5.000|5.000|
+|PERSIAN LUCENE PERSIAN STEM FILTER|1|0|0|1|2.000|2.000|
+|POLISH LUCENE MORFOLOGIK FILTER|1|0|0|1|2.000|2.000|
+|POLISH LUCENE STEMPEL DIRECT|1|0|0|0|4.000|4.000|
+|POLISH LUCENE STEMPEL FILTER|1|0|0|0|5.000|5.000|
+|PORTUGUESE LUCENE PORTUGUESE LIGHT STEM FILTER|1|0|0|0|5.000|5.000|
+|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|1|0|0|0|6.000|6.000|
+|PORTUGUESE LUCENE PORTUGUESE STEM FILTER|1|0|0|0|4.000|4.000|
+|RUSSIAN LUCENE RUSSIAN LIGHT STEM FILTER|1|0|0|0|4.000|4.000|
+|SNOWBALL DANISH DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL DANISH LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL DUTCH DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL DUTCH LUCENE FILTER|1|0|0|0|4.000|4.000|
+|SNOWBALL FINNISH DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL FINNISH LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL FRENCH DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL FRENCH LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL GERMAN DIRECT|1|0|0|1|3.000|3.000|
+|SNOWBALL GERMAN LUCENE FILTER|1|0|0|0|4.000|4.000|
+|SNOWBALL HUNGARIAN DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL HUNGARIAN LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL ITALIAN DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL ITALIAN LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL NORWEGIAN BOKMAL DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL NORWEGIAN BOKMAL LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL NORWEGIAN NYNORSK DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL NORWEGIAN NYNORSK LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL PORTUGUESE DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL PORTUGUESE LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL RUSSIAN DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL RUSSIAN LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL SPANISH DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL SPANISH LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL SWEDISH DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL SWEDISH LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SNOWBALL YIDDISH DIRECT|1|0|0|1|2.000|2.000|
+|SNOWBALL YIDDISH LUCENE FILTER|1|0|0|1|3.000|3.000|
+|SPANISH LUCENE SPANISH LIGHT STEM FILTER|1|0|0|0|5.000|5.000|
+|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|1|0|0|0|7.000|7.000|
+|SPANISH LUCENE SPANISH PLURAL STEM FILTER|1|0|0|0|6.000|6.000|
+|SWEDISH LUCENE SWEDISH LIGHT STEM FILTER|1|0|0|0|5.000|5.000|
+|SWEDISH LUCENE SWEDISH MINIMAL STEM FILTER|1|0|0|0|4.000|4.000|
+|UKRAINIAN LUCENE MORFOLOGIK FILTER|1|0|0|1|2.000|2.000|
+|UKRAINIAN MORFOLOGIK DIRECT|1|0|0|1|3.000|3.000|
+
+</details>
+
+
+### Radixor full-coverage aggregates
+
+These aggregates cover all 19 documented languages. Macro balanced accuracy gives each language equal weight. Micro metrics first sum raw pair counts across languages. Unsupported third-party languages are never inserted as zero results, so this full-coverage table is not presented as a cross-stemmer common-language ranking.
+
+| Dictionary mode | Languages | Macro balanced accuracy | Micro balanced accuracy | Micro precision | Micro recall | Micro F1 |
+|---|---:|---:|---:|---:|---:|---:|
+|ALL_WORDS|19|0.978929|0.987664|0.975113|0.975328|0.975221|
+|LOWERCASE_GROUPS_ONLY|19|0.982354|0.989366|0.975322|0.978734|0.977025|
+
+### Reproducible data
+
+- [Machine-readable quality snapshot](data/stemming-quality.csv)
+- SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- [Linguistic quality methodology](reference/linguistic-quality.md)
+- [Tested stemmer inventory](reference/tested-stemmers.md)
+- [Reproducibility and raw data](reference/reproducibility.md)
+- Pearson and Spearman correlation files are generated under `build/reports/stemming-quality/`; they are separated by dictionary mode and output policy. Correlation does not establish metric equivalence.
+
+<!-- STEMMING-QUALITY-OVERVIEW:END -->
--- a/docs/benchmarks/languages/czech.md
+++ b/docs/benchmarks/languages/czech.md
@@ -0,0 +1,418 @@
+# Czech Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Czech. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `CS_CZ` | 5,113 | 56,612 | 10,049 | 46,563 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **56,612**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 675 | 1.192% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 22,681 | 40.064% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 14,980 | 26.461% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 10,109 | 17.857% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 8,167 | 14.426% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 99.465% | 99.439% | 99.582% | Full Radixor dictionary patch-command stemmer. |
+| Lucene HunspellStemFilter | 84.850% | 82.269% | 96.806% | Benchmark-only Czech Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Lucene CzechStemFilter | 16.784% | 15.538% | 22.559% | Lucene Czech suffix stemmer implemented as a TokenFilter. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `czechRadixor` | 3.332 | 0.240 | 71.6 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene HunspellStemFilter | `luceneHunspellStemFilter` | 346.819 | 3.622 | 7448.4 | 104.091 | Benchmark-only Czech Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Lucene CzechStemFilter | `czechLuceneCzechStemFilter` | 3.163 | 0.253 | 67.9 | 0.949 | Czech suffix stemmer implemented as a Lucene TokenFilter. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `CS_CZ` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/cs_cz/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.996565** among 3 deterministic stemmers. The runner-up is `HUNSPELL CZECH LUCENE FILTER` at 0.853752, a difference of 0.142813. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.997139** among 3 deterministic stemmers. The runner-up is `HUNSPELL CZECH LUCENE FILTER` at 0.852770, a difference of 0.144369. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **7 result rows**, **3 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.996565|3867 / 1334876815 (0.000290%)|2073 / 301835 (0.686799%)|0.988432|0.990189|0.990191|
+|2|HUNSPELL CZECH LUCENE FILTER|PRIMARY_OUTPUT|0.853752|11408 / 1334876815 (0.000855%)|88283 / 301835 (29.248762%)|0.888560|0.810759|0.819499|
+|3|CZECH LUCENE CZECH STEM FILTER|PRIMARY_OUTPUT|0.793614|14480 / 1334876815 (0.001085%)|124586 / 301835 (41.276194%)|0.829234|0.718241|0.736765|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.987264|0.993132|0.999997|0.996565|0.999996|0.000004|
+|2|HUNSPELL CZECH LUCENE FILTER|PRIMARY_OUTPUT|0.949289|0.707512|0.999991|0.853752|0.999925|0.000075|
+|3|CZECH LUCENE CZECH STEM FILTER|PRIMARY_OUTPUT|0.924477|0.587238|0.999989|0.793614|0.999896|0.000104|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.988432|0.990189|0.991953|0.980569|0.990194|0.990191|
+|2|HUNSPELL CZECH LUCENE FILTER|PRIMARY_OUTPUT|0.888560|0.810759|0.745486|0.681745|0.819533|0.819499|
+|3|CZECH LUCENE CZECH STEM FILTER|PRIMARY_OUTPUT|0.829234|0.718241|0.633453|0.560356|0.736809|0.736765|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.990187|0.998733|0.998686|0.998709|0.998709|
+|2|HUNSPELL CZECH LUCENE FILTER|PRIMARY_OUTPUT|0.810723|0.995777|0.952852|0.973842|0.973842|
+|3|CZECH LUCENE CZECH STEM FILTER|PRIMARY_OUTPUT|0.718192|0.993801|0.944977|0.968774|0.968774|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|299762|3867|2073|1334872948|3867 / 1334876815|2073 / 301835|
+|2|HUNSPELL CZECH LUCENE FILTER|PRIMARY_OUTPUT|213552|11408|88283|1334865407|11408 / 1334876815|88283 / 301835|
+|3|CZECH LUCENE CZECH STEM FILTER|PRIMARY_OUTPUT|177249|14480|124586|1334862335|14480 / 1334876815|124586 / 301835|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 1334876815 (0.000000%)|0 / 301835 (0.000000%)|1.000000|1.000000|1.000000|
+|2|HUNSPELL CZECH LUCENE FILTER|ANY_CANDIDATE|0.871577|10102 / 1334876815 (0.000757%)|77523 / 301835 (25.683900%)|0.904855|0.836596|0.843258|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+|2|HUNSPELL CZECH LUCENE FILTER|ANY_CANDIDATE|0.956905|0.743161|0.999992|0.871577|0.999934|0.000066|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+|2|HUNSPELL CZECH LUCENE FILTER|ANY_CANDIDATE|0.904855|0.836596|0.777914|0.719094|0.843288|0.843258|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL CZECH LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|301835|0|0|1334876815|0 / 1334876815|0 / 301835|
+|2|HUNSPELL CZECH LUCENE FILTER|ANY_CANDIDATE|224312|10102|77523|1334866713|10102 / 1334876815|77523 / 301835|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999998|5850 / 1334876815 (0.000438%)|0 / 301835 (0.000000%)|0.984732|0.990402|0.990446|
+|2|HUNSPELL CZECH LUCENE FILTER|ALL_CANDIDATES|0.871575|13917 / 1334876815 (0.001043%)|77523 / 301835 (25.683900%)|0.893851|0.830687|0.836477|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.980987|1.000000|0.999996|0.999998|0.999996|0.000004|
+|2|HUNSPELL CZECH LUCENE FILTER|ALL_CANDIDATES|0.941581|0.743161|0.999990|0.871575|0.999932|0.000068|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.984732|0.990402|0.996139|0.980987|0.990448|0.990446|
+|2|HUNSPELL CZECH LUCENE FILTER|ALL_CANDIDATES|0.893851|0.830687|0.775861|0.710406|0.836509|0.836477|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL CZECH LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|301835|5850|0|1334870965|5850 / 1334876815|0 / 301835|
+|2|HUNSPELL CZECH LUCENE FILTER|ALL_CANDIDATES|224312|13917|77523|1334862898|13917 / 1334876815|77523 / 301835|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|2073|3867|1983|596|1.153340%|4|52319|
+|HUNSPELL CZECH LUCENE FILTER|10760|1306|2509|3317|6.418840%|5|55596|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **7 result rows**, **3 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.997139|3863 / 1298544215 (0.000297%)|1709 / 298813 (0.571930%)|0.988580|0.990710|0.990714|
+|2|HUNSPELL CZECH LUCENE FILTER|PRIMARY_OUTPUT|0.852770|11239 / 1298544215 (0.000866%)|87986 / 298813 (29.445171%)|0.888009|0.809505|0.818403|
+|3|CZECH LUCENE CZECH STEM FILTER|PRIMARY_OUTPUT|0.791794|13950 / 1298544215 (0.001074%)|124426 / 298813 (41.640089%)|0.828709|0.715948|0.735055|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.987165|0.994281|0.999997|0.997139|0.999996|0.000004|
+|2|HUNSPELL CZECH LUCENE FILTER|PRIMARY_OUTPUT|0.949389|0.705548|0.999991|0.852770|0.999924|0.000076|
+|3|CZECH LUCENE CZECH STEM FILTER|PRIMARY_OUTPUT|0.925931|0.583599|0.999989|0.791794|0.999893|0.000107|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.988580|0.990710|0.992849|0.981591|0.990716|0.990714|
+|2|HUNSPELL CZECH LUCENE FILTER|PRIMARY_OUTPUT|0.888009|0.809505|0.743753|0.679973|0.818437|0.818403|
+|3|CZECH LUCENE CZECH STEM FILTER|PRIMARY_OUTPUT|0.828709|0.715948|0.630198|0.557569|0.735100|0.735055|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.990708|0.998726|0.999030|0.998878|0.998878|
+|2|HUNSPELL CZECH LUCENE FILTER|PRIMARY_OUTPUT|0.809467|0.995812|0.952394|0.973619|0.973619|
+|3|CZECH LUCENE CZECH STEM FILTER|PRIMARY_OUTPUT|0.715897|0.993897|0.944297|0.968463|0.968463|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|297104|3863|1709|1298540352|3863 / 1298544215|1709 / 298813|
+|2|HUNSPELL CZECH LUCENE FILTER|PRIMARY_OUTPUT|210827|11239|87986|1298532976|11239 / 1298544215|87986 / 298813|
+|3|CZECH LUCENE CZECH STEM FILTER|PRIMARY_OUTPUT|174387|13950|124426|1298530265|13950 / 1298544215|124426 / 298813|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 1298544215 (0.000000%)|0 / 298813 (0.000000%)|1.000000|1.000000|1.000000|
+|2|HUNSPELL CZECH LUCENE FILTER|ANY_CANDIDATE|0.870432|10028 / 1298544215 (0.000772%)|77431 / 298813 (25.912862%)|0.904004|0.835052|0.841852|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+|2|HUNSPELL CZECH LUCENE FILTER|ANY_CANDIDATE|0.956666|0.740871|0.999992|0.870432|0.999933|0.000067|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+|2|HUNSPELL CZECH LUCENE FILTER|ANY_CANDIDATE|0.904004|0.835052|0.775874|0.716815|0.841883|0.841852|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL CZECH LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|298813|0|0|1298544215|0 / 1298544215|0 / 298813|
+|2|HUNSPELL CZECH LUCENE FILTER|ANY_CANDIDATE|221382|10028|77431|1298534187|10028 / 1298544215|77431 / 298813|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999998|5782 / 1298544215 (0.000445%)|0 / 298813 (0.000000%)|0.984756|0.990418|0.990461|
+|2|HUNSPELL CZECH LUCENE FILTER|ALL_CANDIDATES|0.870430|13601 / 1298544215 (0.001047%)|77431 / 298813 (25.912862%)|0.893574|0.829463|0.835425|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.981017|1.000000|0.999996|0.999998|0.999996|0.000004|
+|2|HUNSPELL CZECH LUCENE FILTER|ALL_CANDIDATES|0.942119|0.740871|0.999990|0.870430|0.999930|0.000070|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.984756|0.990418|0.996145|0.981017|0.990463|0.990461|
+|2|HUNSPELL CZECH LUCENE FILTER|ALL_CANDIDATES|0.893574|0.829463|0.773936|0.708617|0.835457|0.835425|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL CZECH LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|298813|5782|0|1298538433|5782 / 1298544215|0 / 298813|
+|2|HUNSPELL CZECH LUCENE FILTER|ALL_CANDIDATES|221382|13601|77431|1298530614|13601 / 1298544215|77431 / 298813|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|1709|3863|1919|540|1.059488%|4|51543|
+|HUNSPELL CZECH LUCENE FILTER|10555|1211|2362|3237|6.351044%|5|54804|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `CS_CZ`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/danish.md
+++ b/docs/benchmarks/languages/danish.md
@@ -0,0 +1,396 @@
+# Danish Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Danish. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `DA_DK` | 4,179 | 32,256 | 8,356 | 23,900 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **32,256**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 137 | 0.425% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 1,127 | 3.494% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 22,586 | 70.021% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 8,405 | 26.057% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 1 | 0.003% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 99.371% | 99.527% | 98.923% | Full Radixor dictionary patch-command stemmer. |
+| Lucene SnowballFilter | 55.509% | 54.159% | 59.371% | Lucene TokenFilter integration path around the Snowball algorithm. |
+| Official Snowball direct | 55.509% | 54.159% | 59.371% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `radixor[DANISH]` | 1.143 | 0.017 | 47.8 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Official Snowball direct | `snowballDirect[DANISH]` | 2.168 | 0.058 | 90.7 | 1.896 | Official Snowball generated Java stemmer; direct API. |
+| Lucene SnowballFilter | `luceneSnowballFilter[DANISH]` | 2.975 | 0.143 | 124.5 | 2.602 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `DA_DK` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/da_dk/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.996066** among 3 deterministic stemmers. The runner-up is `SNOWBALL DANISH LUCENE FILTER` at 0.937969, a difference of 0.058097. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.996305** among 3 deterministic stemmers. The runner-up is `SNOWBALL DANISH DIRECT` at 0.938074, a difference of 0.058230. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **5 result rows**, **3 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.996066|1165 / 394111186 (0.000296%)|707 / 89895 (0.786473%)|0.988108|0.989614|0.989615|
+|2|SNOWBALL DANISH LUCENE FILTER|PRIMARY_OUTPUT|0.937969|6507 / 394111186 (0.001651%)|11151 / 89895 (12.404472%)|0.913718|0.899181|0.899475|
+|3|SNOWBALL DANISH DIRECT|PRIMARY_OUTPUT|0.937903|6341 / 394111186 (0.001609%)|11163 / 89895 (12.417821%)|0.915090|0.899959|0.900279|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.987106|0.992135|0.999997|0.996066|0.999995|0.000005|
+|2|SNOWBALL DANISH LUCENE FILTER|PRIMARY_OUTPUT|0.923672|0.875955|0.999983|0.937969|0.999955|0.000045|
+|3|SNOWBALL DANISH DIRECT|PRIMARY_OUTPUT|0.925464|0.875822|0.999984|0.937903|0.999956|0.000044|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.988108|0.989614|0.991125|0.979442|0.989618|0.989615|
+|2|SNOWBALL DANISH LUCENE FILTER|PRIMARY_OUTPUT|0.913718|0.899181|0.885100|0.816830|0.899498|0.899475|
+|3|SNOWBALL DANISH DIRECT|PRIMARY_OUTPUT|0.915090|0.899959|0.885320|0.818114|0.900301|0.900279|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.989612|0.998466|0.998719|0.998592|0.998592|
+|2|SNOWBALL DANISH LUCENE FILTER|PRIMARY_OUTPUT|0.899159|0.994053|0.978603|0.986268|0.986268|
+|3|SNOWBALL DANISH DIRECT|PRIMARY_OUTPUT|0.899937|0.994196|0.978579|0.986326|0.986326|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|89188|1165|707|394110021|1165 / 394111186|707 / 89895|
+|2|SNOWBALL DANISH LUCENE FILTER|PRIMARY_OUTPUT|78744|6507|11151|394104679|6507 / 394111186|11151 / 89895|
+|3|SNOWBALL DANISH DIRECT|PRIMARY_OUTPUT|78732|6341|11163|394104845|6341 / 394111186|11163 / 89895|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 394111186 (0.000000%)|0 / 89895 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|89895|0|0|394111186|0 / 394111186|0 / 89895|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999998|1849 / 394111186 (0.000469%)|0 / 89895 (0.000000%)|0.983812|0.989820|0.989869|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.979846|1.000000|0.999995|0.999998|0.999995|0.000005|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.983812|0.989820|0.995903|0.979846|0.989872|0.989869|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|89895|1849|0|394109337|1849 / 394111186|0 / 89895|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|707|1165|684|323|1.150326%|3|28405|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **5 result rows**, **3 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.996305|1165 / 392820788 (0.000297%)|663 / 89740 (0.738801%)|0.988190|0.989843|0.989845|
+|2|SNOWBALL DANISH DIRECT|PRIMARY_OUTPUT|0.938074|6341 / 392820788 (0.001614%)|11113 / 89740 (12.383552%)|0.915093|0.900096|0.900410|
+|3|SNOWBALL DANISH LUCENE FILTER|PRIMARY_OUTPUT|0.938074|6341 / 392820788 (0.001614%)|11113 / 89740 (12.383552%)|0.915093|0.900096|0.900410|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.987090|0.992612|0.999997|0.996305|0.999995|0.000005|
+|2|SNOWBALL DANISH DIRECT|PRIMARY_OUTPUT|0.925372|0.876164|0.999984|0.938074|0.999956|0.000044|
+|3|SNOWBALL DANISH LUCENE FILTER|PRIMARY_OUTPUT|0.925372|0.876164|0.999984|0.938074|0.999956|0.000044|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.988190|0.989843|0.991503|0.979891|0.989847|0.989845|
+|2|SNOWBALL DANISH DIRECT|PRIMARY_OUTPUT|0.915093|0.900096|0.885583|0.818341|0.900432|0.900410|
+|3|SNOWBALL DANISH LUCENE FILTER|PRIMARY_OUTPUT|0.915093|0.900096|0.885583|0.818341|0.900432|0.900410|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.989841|0.998463|0.998812|0.998637|0.998637|
+|2|SNOWBALL DANISH DIRECT|PRIMARY_OUTPUT|0.900074|0.994185|0.978644|0.986354|0.986354|
+|3|SNOWBALL DANISH LUCENE FILTER|PRIMARY_OUTPUT|0.900074|0.994185|0.978644|0.986354|0.986354|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|89077|1165|663|392819623|1165 / 392820788|663 / 89740|
+|2|SNOWBALL DANISH DIRECT|PRIMARY_OUTPUT|78627|6341|11113|392814447|6341 / 392820788|11113 / 89740|
+|3|SNOWBALL DANISH LUCENE FILTER|PRIMARY_OUTPUT|78627|6341|11113|392814447|6341 / 392820788|11113 / 89740|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 392820788 (0.000000%)|0 / 89740 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|89740|0|0|392820788|0 / 392820788|0 / 89740|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999998|1849 / 392820788 (0.000471%)|0 / 89740 (0.000000%)|0.983784|0.989803|0.989852|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.979812|1.000000|0.999995|0.999998|0.999995|0.000005|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.983784|0.989803|0.995896|0.979812|0.989855|0.989852|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|89740|1849|0|392818939|1849 / 392820788|0 / 89740|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|663|1165|684|315|1.123676%|3|28351|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `DA_DK`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/dutch.md
+++ b/docs/benchmarks/languages/dutch.md
@@ -0,0 +1,430 @@
+# Dutch Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Dutch. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `NL_NL` | 4,992 | 31,466 | 9,981 | 21,485 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **31,466**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 2,107 | 6.696% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 11,484 | 36.497% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 7,732 | 24.573% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 10,127 | 32.184% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 16 | 0.051% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 99.120% | 98.711% | 100.000% | Full Radixor dictionary patch-command stemmer. |
+| Lucene HunspellStemFilter | 46.590% | 22.718% | 97.976% | Benchmark-only Dutch Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Official Snowball direct | 15.954% | 8.992% | 30.939% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+| Lucene SnowballFilter | 12.620% | 5.441% | 28.073% | Lucene TokenFilter integration path around the Snowball algorithm. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `radixor[DUTCH]` | 1.331 | 0.114 | 61.9 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene HunspellStemFilter | `luceneHunspellStemFilter` | 22.760 | 1.387 | 1059.3 | 17.105 | Benchmark-only Dutch Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Official Snowball direct | `snowballDirect[DUTCH]` | 4.146 | 0.291 | 193.0 | 3.116 | Official Snowball generated Java stemmer; direct API. |
+| Lucene SnowballFilter | `luceneSnowballFilter[DUTCH]` | 7.375 | 0.595 | 343.3 | 5.543 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `NL_NL` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/nl_nl/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.988661** among 4 deterministic stemmers. The runner-up is `SNOWBALL DUTCH DIRECT` at 0.727087, a difference of 0.261574. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.989040** among 4 deterministic stemmers. The runner-up is `SNOWBALL DUTCH DIRECT` at 0.730495, a difference of 0.258544. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **8 result rows**, **4 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.988661|1214 / 350437960 (0.000346%)|1464 / 64566 (2.267447%)|0.980362|0.979221|0.979219|
+|2|SNOWBALL DUTCH DIRECT|PRIMARY_OUTPUT|0.727087|4382 / 350437960 (0.001250%)|35241 / 64566 (54.581359%)|0.735353|0.596807|0.628557|
+|3|HUNSPELL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.643123|1333 / 350437960 (0.000380%)|46084 / 64566 (71.375027%)|0.642512|0.438061|0.516674|
+|4|SNOWBALL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.618497|1588 / 350437960 (0.000453%)|49264 / 64566 (76.300220%)|0.579068|0.375712|0.463333|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.981124|0.977326|0.999997|0.988661|0.999992|0.000008|
+|2|SNOWBALL DUTCH DIRECT|PRIMARY_OUTPUT|0.869997|0.454186|0.999987|0.727087|0.999887|0.000113|
+|3|HUNSPELL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.932728|0.286250|0.999996|0.643123|0.999865|0.000135|
+|4|SNOWBALL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.905980|0.236998|0.999995|0.618497|0.999855|0.000145|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.980362|0.979221|0.978083|0.959289|0.979223|0.979219|
+|2|SNOWBALL DUTCH DIRECT|PRIMARY_OUTPUT|0.735353|0.596807|0.502190|0.425321|0.628602|0.628557|
+|3|HUNSPELL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.642512|0.438061|0.332316|0.280459|0.516714|0.516674|
+|4|SNOWBALL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.579068|0.375712|0.278062|0.231309|0.463374|0.463333|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.979217|0.997464|0.997003|0.997234|0.997234|
+|2|SNOWBALL DUTCH DIRECT|PRIMARY_OUTPUT|0.596756|0.992815|0.917346|0.953590|0.953590|
+|3|HUNSPELL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.438012|0.996932|0.889026|0.939892|0.939892|
+|4|SNOWBALL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.375664|0.995828|0.888410|0.939057|0.939057|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|63102|1214|1464|350436746|1214 / 350437960|1464 / 64566|
+|2|SNOWBALL DUTCH DIRECT|PRIMARY_OUTPUT|29325|4382|35241|350433578|4382 / 350437960|35241 / 64566|
+|3|HUNSPELL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|18482|1333|46084|350436627|1333 / 350437960|46084 / 64566|
+|4|SNOWBALL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|15302|1588|49264|350436372|1588 / 350437960|49264 / 64566|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 350437960 (0.000000%)|0 / 64566 (0.000000%)|1.000000|1.000000|1.000000|
+|2|HUNSPELL DUTCH LUCENE FILTER|ANY_CANDIDATE|0.665519|1164 / 350437960 (0.000332%)|43192 / 64566 (66.895889%)|0.690741|0.490770|0.560268|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+|2|HUNSPELL DUTCH LUCENE FILTER|ANY_CANDIDATE|0.948354|0.331041|0.999997|0.665519|0.999873|0.000127|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+|2|HUNSPELL DUTCH LUCENE FILTER|ANY_CANDIDATE|0.690741|0.490770|0.380588|0.325179|0.560307|0.560268|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL DUTCH LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|64566|0|0|350437960|0 / 350437960|0 / 64566|
+|2|HUNSPELL DUTCH LUCENE FILTER|ANY_CANDIDATE|21374|1164|43192|350436796|1164 / 350437960|43192 / 64566|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999996|2651 / 350437960 (0.000756%)|0 / 64566 (0.000000%)|0.968198|0.979884|0.980078|
+|2|HUNSPELL DUTCH LUCENE FILTER|ALL_CANDIDATES|0.665518|1738 / 350437960 (0.000496%)|43192 / 64566 (66.895889%)|0.680640|0.487557|0.553265|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.960561|1.000000|0.999992|0.999996|0.999992|0.000008|
+|2|HUNSPELL DUTCH LUCENE FILTER|ALL_CANDIDATES|0.924801|0.331041|0.999995|0.665518|0.999872|0.000128|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.968198|0.979884|0.991855|0.960561|0.980082|0.980078|
+|2|HUNSPELL DUTCH LUCENE FILTER|ALL_CANDIDATES|0.680640|0.487557|0.379812|0.322364|0.553306|0.553265|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL DUTCH LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|64566|2651|0|350435309|2651 / 350437960|0 / 64566|
+|2|HUNSPELL DUTCH LUCENE FILTER|ALL_CANDIDATES|21374|1738|43192|350436222|1738 / 350437960|43192 / 64566|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|HUNSPELL DUTCH LUCENE FILTER|2892|169|405|1254|4.736186%|3|27763|
+|Radixor|1464|1214|1437|572|2.160366%|3|27061|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **8 result rows**, **4 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.989040|1214 / 329603856 (0.000368%)|1384 / 63147 (2.191711%)|0.980194|0.979401|0.979398|
+|2|SNOWBALL DUTCH DIRECT|PRIMARY_OUTPUT|0.730495|4382 / 329603856 (0.001329%)|34036 / 63147 (53.899631%)|0.738412|0.602463|0.632953|
+|3|HUNSPELL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.645159|1310 / 329603856 (0.000397%)|44814 / 63147 (70.967742%)|0.646808|0.442880|0.520498|
+|4|SNOWBALL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.618546|1544 / 329603856 (0.000468%)|48175 / 63147 (76.290243%)|0.579362|0.375883|0.463566|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.980723|0.978083|0.999996|0.989040|0.999992|0.000008|
+|2|SNOWBALL DUTCH DIRECT|PRIMARY_OUTPUT|0.869167|0.461004|0.999987|0.730495|0.999883|0.000117|
+|3|HUNSPELL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.933310|0.290323|0.999996|0.645159|0.999860|0.000140|
+|4|SNOWBALL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.906515|0.237098|0.999995|0.618546|0.999849|0.000151|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.980194|0.979401|0.978610|0.959634|0.979402|0.979398|
+|2|SNOWBALL DUTCH DIRECT|PRIMARY_OUTPUT|0.738412|0.602463|0.508789|0.431089|0.633000|0.632953|
+|3|HUNSPELL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.646808|0.442880|0.336718|0.284422|0.520539|0.520498|
+|4|SNOWBALL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.579362|0.375883|0.278182|0.231439|0.463608|0.463566|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.979397|0.997373|0.997139|0.997256|0.997256|
+|2|SNOWBALL DUTCH DIRECT|PRIMARY_OUTPUT|0.602410|0.992557|0.918059|0.953856|0.953856|
+|3|HUNSPELL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.442829|0.996884|0.889061|0.939890|0.939890|
+|4|SNOWBALL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|0.375834|0.995817|0.887492|0.938539|0.938539|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|61763|1214|1384|329602642|1214 / 329603856|1384 / 63147|
+|2|SNOWBALL DUTCH DIRECT|PRIMARY_OUTPUT|29111|4382|34036|329599474|4382 / 329603856|34036 / 63147|
+|3|HUNSPELL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|18333|1310|44814|329602546|1310 / 329603856|44814 / 63147|
+|4|SNOWBALL DUTCH LUCENE FILTER|PRIMARY_OUTPUT|14972|1544|48175|329602312|1544 / 329603856|48175 / 63147|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 329603856 (0.000000%)|0 / 63147 (0.000000%)|1.000000|1.000000|1.000000|
+|2|HUNSPELL DUTCH LUCENE FILTER|ANY_CANDIDATE|0.667956|1141 / 329603856 (0.000346%)|41935 / 63147 (66.408539%)|0.695206|0.496187|0.564555|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+|2|HUNSPELL DUTCH LUCENE FILTER|ANY_CANDIDATE|0.948955|0.335915|0.999997|0.667956|0.999869|0.000131|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+|2|HUNSPELL DUTCH LUCENE FILTER|ANY_CANDIDATE|0.695206|0.496187|0.385755|0.329953|0.564595|0.564555|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL DUTCH LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|63147|0|0|329603856|0 / 329603856|0 / 63147|
+|2|HUNSPELL DUTCH LUCENE FILTER|ANY_CANDIDATE|21212|1141|41935|329602715|1141 / 329603856|41935 / 63147|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999996|2651 / 329603856 (0.000804%)|0 / 63147 (0.000000%)|0.967506|0.979441|0.979644|
+|2|HUNSPELL DUTCH LUCENE FILTER|ALL_CANDIDATES|0.667955|1712 / 329603856 (0.000519%)|41935 / 63147 (66.408539%)|0.684952|0.492895|0.557477|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.959710|1.000000|0.999992|0.999996|0.999992|0.000008|
+|2|HUNSPELL DUTCH LUCENE FILTER|ALL_CANDIDATES|0.925318|0.335915|0.999995|0.667955|0.999868|0.000132|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.967506|0.979441|0.991674|0.959710|0.979648|0.979644|
+|2|HUNSPELL DUTCH LUCENE FILTER|ALL_CANDIDATES|0.684952|0.492895|0.384956|0.327048|0.557519|0.557477|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL DUTCH LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|63147|2651|0|329601205|2651 / 329603856|0 / 63147|
+|2|HUNSPELL DUTCH LUCENE FILTER|ALL_CANDIDATES|21212|1712|41935|329602144|1712 / 329603856|41935 / 63147|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|HUNSPELL DUTCH LUCENE FILTER|2879|169|402|1186|4.618740%|3|26896|
+|Radixor|1384|1214|1437|549|2.138017%|3|26239|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `NL_NL`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/english.md
+++ b/docs/benchmarks/languages/english.md
@@ -0,0 +1,514 @@
+# English Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for English. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `US_UK` | 396,939 | 1,004,374 | 793,874 | 210,500 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **1,004,374**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 28 | 0.003% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 22,493 | 2.240% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 186,764 | 18.595% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 795,024 | 79.156% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 65 | 0.006% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 97.478% | 97.197% | 97.552% | Full Radixor dictionary patch-command stemmer. |
+| Lucene EnglishMinimalStemFilter | 90.981% | 65.189% | 97.820% | Minimal English plural reduction, not a full stemmer. |
+| Lucene KStemFilter | 80.076% | 76.608% | 80.996% | Krovetz-style English stemming TokenFilter; broader than minimal suffix reducers. |
+| Lucene HunspellStemFilter | 80.243% | 12.750% | 98.139% | Benchmark-only English Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Lucene EnglishPossessiveFilter | 79.032% | 0.003% | 99.987% | Possessive-ending remover only, not a full stemmer. |
+| Snowball English / Porter2 | 40.342% | 46.296% | 38.763% | Porter2 rule-based suffix stemmer, distinct from original Porter. |
+| Lucene PorterStemFilter | 39.538% | 46.201% | 37.772% | Lucene TokenFilter path for Porter suffix rules; not dictionary-root equivalent. |
+| Lucene PorterStemmer direct copy | 39.538% | 46.201% | 37.772% | Direct Porter suffix-rule implementation generated under build for benchmark-only use. |
+| OpenNLP PorterStemmer | 39.538% | 46.201% | 37.772% | Apache OpenNLP Porter suffix-rule implementation. |
+| Snowball original Porter | 39.529% | 46.179% | 37.766% | Classic Porter rule-based suffix stemmer. |
+| Paice/Husk Lancaster | 28.055% | 37.039% | 25.673% | Aggressive Paice/Husk rule stemmer that often produces shorter stems. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `radixorUsUkProfiPreferredStem` | 21.987 | 8.707 | 104.5 | 1.000 | Full dictionary patch-command stemmer using compiled patch commands. |
+| Lucene EnglishPossessiveFilter | `luceneEnglishPossessiveFilter` | 24.539 | 1.515 | 116.6 | 1.116 | Possessive-ending remover only; not a full stemmer. |
+| Lucene EnglishMinimalStemFilter | `luceneEnglishMinimalStemFilter` | 22.702 | 1.195 | 107.8 | 1.032 | Narrow plural reduction filter; not a full stemmer. |
+| Lucene PorterStemmer direct copy | `lucenePorterStemmerCopied` | 24.696 | 13.235 | 117.3 | 1.123 | Benchmark-only generated copy of Lucene package-private Porter implementation. |
+| OpenNLP PorterStemmer | `opennlpPorterStemmer` | 23.121 | 12.528 | 109.8 | 1.052 | Apache OpenNLP Porter implementation. |
+| Snowball original Porter | `snowballOriginalPorter` | 38.904 | 10.353 | 184.8 | 1.769 | Classic Porter suffix-rule stemmer; historical English baseline, not a dictionary-equivalent stemmer. |
+| Lucene PorterStemFilter | `lucenePorterStemFilter` | 37.021 | 1.196 | 175.9 | 1.684 | Lucene TokenFilter integration path for Porter; includes TokenStream overhead. |
+| Lucene KStemFilter | `luceneKStemFilter` | 51.640 | 2.591 | 245.3 | 2.349 | Krovetz-style English TokenFilter; broader than minimal suffix filters. |
+| Lucene HunspellStemFilter | `luceneHunspellStemFilter` | 79.785 | 1.347 | 379.0 | 3.629 | Benchmark-only English Hunspell comparison using the benchmark Hunspell corpus. |
+| Snowball English / Porter2 | `snowballEnglishPorter2` | 52.437 | 0.773 | 249.1 | 2.385 | Porter2 suffix-rule stemmer, distinct from original Porter. |
+| Paice/Husk Lancaster | `paiceHuskLancaster` | 141.556 | 12.324 | 672.5 | 6.438 | Aggressive rule-based English stemmer. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `US_UK` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/us_uk/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.965159** among 11 deterministic stemmers. The runner-up is `ENGLISH LUCENE PORTER COPIED` at 0.954627, a difference of 0.010533. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.965820** among 11 deterministic stemmers. The runner-up is `ENGLISH LUCENE PORTER COPIED` at 0.954900, a difference of 0.010920. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **15 result rows**, **11 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.965159|1149886 / 184490451771 (0.000623%)|21869 / 313870 (6.967534%)|0.240076|0.332621|0.434052|
+|2|ENGLISH LUCENE PORTER COPIED|PRIMARY_OUTPUT|0.954627|1557406 / 184490451771 (0.000844%)|28480 / 313870 (9.073820%)|0.185679|0.264659|0.375252|
+|3|ENGLISH LUCENE PORTER FILTER|PRIMARY_OUTPUT|0.954627|1557406 / 184490451771 (0.000844%)|28480 / 313870 (9.073820%)|0.185679|0.264659|0.375252|
+|4|ENGLISH OPENNLP PORTER|PRIMARY_OUTPUT|0.954627|1557406 / 184490451771 (0.000844%)|28480 / 313870 (9.073820%)|0.185679|0.264659|0.375252|
+|5|ENGLISH SNOWBALL PORTER2|PRIMARY_OUTPUT|0.954537|1566711 / 184490451771 (0.000849%)|28536 / 313870 (9.091662%)|0.184753|0.263477|0.374240|
+|6|ENGLISH SNOWBALL ORIGINAL PORTER|PRIMARY_OUTPUT|0.954490|1555293 / 184490451771 (0.000843%)|28566 / 313870 (9.101220%)|0.185835|0.264849|0.375363|
+|7|ENGLISH PAICE HUSK LANCASTER|PRIMARY_OUTPUT|0.952394|3062661 / 184490451771 (0.001660%)|29879 / 313870 (9.519546%)|0.103643|0.155164|0.277089|
+|8|ENGLISH LUCENE KSTEM FILTER|PRIMARY_OUTPUT|0.878441|1368501 / 184490451771 (0.000742%)|76305 / 313870 (24.311020%)|0.176284|0.247472|0.334598|
+|9|ENGLISH LUCENE MINIMAL FILTER|PRIMARY_OUTPUT|0.718599|1122264 / 184490451771 (0.000608%)|176645 / 313870 (56.279670%)|0.128204|0.174436|0.218251|
+|10|HUNSPELL ENGLISH LUCENE FILTER|PRIMARY_OUTPUT|0.573277|1981986 / 184490451771 (0.001074%)|267868 / 313870 (85.343614%)|0.027298|0.039287|0.057655|
+|11|ENGLISH LUCENE POSSESSIVE FILTER|PRIMARY_OUTPUT|0.500008|1115154 / 184490451771 (0.000604%)|313863 / 313870 (99.997770%)|0.000007|0.000010|0.000009|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.202513|0.930325|0.999994|0.965159|0.999994|0.000006|
+|2|ENGLISH LUCENE PORTER COPIED|PRIMARY_OUTPUT|0.154868|0.909262|0.999992|0.954627|0.999991|0.000009|
+|3|ENGLISH LUCENE PORTER FILTER|PRIMARY_OUTPUT|0.154868|0.909262|0.999992|0.954627|0.999991|0.000009|
+|4|ENGLISH OPENNLP PORTER|PRIMARY_OUTPUT|0.154868|0.909262|0.999992|0.954627|0.999991|0.000009|
+|5|ENGLISH SNOWBALL PORTER2|PRIMARY_OUTPUT|0.154064|0.909083|0.999992|0.954537|0.999991|0.000009|
+|6|ENGLISH SNOWBALL ORIGINAL PORTER|PRIMARY_OUTPUT|0.155006|0.908988|0.999992|0.954490|0.999991|0.000009|
+|7|ENGLISH PAICE HUSK LANCASTER|PRIMARY_OUTPUT|0.084858|0.904805|0.999983|0.952394|0.999983|0.000017|
+|8|ENGLISH LUCENE KSTEM FILTER|PRIMARY_OUTPUT|0.147917|0.756890|0.999993|0.878441|0.999992|0.000008|
+|9|ENGLISH LUCENE MINIMAL FILTER|PRIMARY_OUTPUT|0.108953|0.437203|0.999994|0.718599|0.999993|0.000007|
+|10|HUNSPELL ENGLISH LUCENE FILTER|PRIMARY_OUTPUT|0.022684|0.146564|0.999989|0.573277|0.999988|0.000012|
+|11|ENGLISH LUCENE POSSESSIVE FILTER|PRIMARY_OUTPUT|0.000006|0.000022|0.999994|0.500008|0.999992|0.000008|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.240076|0.332621|0.541270|0.199487|0.434054|0.434052|
+|2|ENGLISH LUCENE PORTER COPIED|PRIMARY_OUTPUT|0.185679|0.264659|0.460563|0.152511|0.375254|0.375252|
+|3|ENGLISH LUCENE PORTER FILTER|PRIMARY_OUTPUT|0.185679|0.264659|0.460563|0.152511|0.375254|0.375252|
+|4|ENGLISH OPENNLP PORTER|PRIMARY_OUTPUT|0.185679|0.264659|0.460563|0.152511|0.375254|0.375252|
+|5|ENGLISH SNOWBALL PORTER2|PRIMARY_OUTPUT|0.184753|0.263477|0.459102|0.151727|0.374242|0.374240|
+|6|ENGLISH SNOWBALL ORIGINAL PORTER|PRIMARY_OUTPUT|0.185835|0.264849|0.460751|0.152637|0.375365|0.375363|
+|7|ENGLISH PAICE HUSK LANCASTER|PRIMARY_OUTPUT|0.103643|0.155164|0.308543|0.084107|0.277092|0.277089|
+|8|ENGLISH LUCENE KSTEM FILTER|PRIMARY_OUTPUT|0.176284|0.247472|0.415099|0.141208|0.334600|0.334598|
+|9|ENGLISH LUCENE MINIMAL FILTER|PRIMARY_OUTPUT|0.128204|0.174436|0.272816|0.095552|0.218253|0.218251|
+|10|HUNSPELL ENGLISH LUCENE FILTER|PRIMARY_OUTPUT|0.027298|0.039287|0.070051|0.020037|0.057659|0.057655|
+|11|ENGLISH LUCENE POSSESSIVE FILTER|PRIMARY_OUTPUT|0.000007|0.000010|0.000015|0.000005|0.000012|0.000009|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.332619|0.994215|0.997770|0.995989|0.995989|
+|2|ENGLISH LUCENE PORTER COPIED|PRIMARY_OUTPUT|0.264656|0.969648|0.997199|0.983231|0.983231|
+|3|ENGLISH LUCENE PORTER FILTER|PRIMARY_OUTPUT|0.264656|0.969648|0.997199|0.983231|0.983231|
+|4|ENGLISH OPENNLP PORTER|PRIMARY_OUTPUT|0.264656|0.969648|0.997199|0.983231|0.983231|
+|5|ENGLISH SNOWBALL PORTER2|PRIMARY_OUTPUT|0.263474|0.969037|0.997182|0.982908|0.982908|
+|6|ENGLISH SNOWBALL ORIGINAL PORTER|PRIMARY_OUTPUT|0.264847|0.969891|0.997193|0.983353|0.983353|
+|7|ENGLISH PAICE HUSK LANCASTER|PRIMARY_OUTPUT|0.155162|0.937768|0.996600|0.966289|0.966289|
+|8|ENGLISH LUCENE KSTEM FILTER|PRIMARY_OUTPUT|0.247470|0.980687|0.992108|0.986364|0.986364|
+|9|ENGLISH LUCENE MINIMAL FILTER|PRIMARY_OUTPUT|0.174433|0.995202|0.981174|0.988138|0.988138|
+|10|HUNSPELL ENGLISH LUCENE FILTER|PRIMARY_OUTPUT|0.039284|0.993096|0.963677|0.978166|0.978166|
+|11|ENGLISH LUCENE POSSESSIVE FILTER|PRIMARY_OUTPUT|0.000007|0.995789|0.958019|0.976539|0.976539|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|292001|1149886|21869|184489301885|1149886 / 184490451771|21869 / 313870|
+|2|ENGLISH LUCENE PORTER COPIED|PRIMARY_OUTPUT|285390|1557406|28480|184488894365|1557406 / 184490451771|28480 / 313870|
+|3|ENGLISH LUCENE PORTER FILTER|PRIMARY_OUTPUT|285390|1557406|28480|184488894365|1557406 / 184490451771|28480 / 313870|
+|4|ENGLISH OPENNLP PORTER|PRIMARY_OUTPUT|285390|1557406|28480|184488894365|1557406 / 184490451771|28480 / 313870|
+|5|ENGLISH SNOWBALL PORTER2|PRIMARY_OUTPUT|285334|1566711|28536|184488885060|1566711 / 184490451771|28536 / 313870|
+|6|ENGLISH SNOWBALL ORIGINAL PORTER|PRIMARY_OUTPUT|285304|1555293|28566|184488896478|1555293 / 184490451771|28566 / 313870|
+|7|ENGLISH PAICE HUSK LANCASTER|PRIMARY_OUTPUT|283991|3062661|29879|184487389110|3062661 / 184490451771|29879 / 313870|
+|8|ENGLISH LUCENE KSTEM FILTER|PRIMARY_OUTPUT|237565|1368501|76305|184489083270|1368501 / 184490451771|76305 / 313870|
+|9|ENGLISH LUCENE MINIMAL FILTER|PRIMARY_OUTPUT|137225|1122264|176645|184489329507|1122264 / 184490451771|176645 / 313870|
+|10|HUNSPELL ENGLISH LUCENE FILTER|PRIMARY_OUTPUT|46002|1981986|267868|184488469785|1981986 / 184490451771|267868 / 313870|
+|11|ENGLISH LUCENE POSSESSIVE FILTER|PRIMARY_OUTPUT|7|1115154|313863|184489336617|1115154 / 184490451771|313863 / 313870|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|0.999976|12 / 184490451771 (0.000000%)|15 / 313870 (0.004779%)|0.999960|0.999957|0.999957|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ANY_CANDIDATE|0.581603|1978852 / 184490451771 (0.001073%)|262641 / 313870 (83.678274%)|0.030370|0.043712|0.064174|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|0.999962|0.999952|1.000000|0.999976|1.000000|0.000000|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ANY_CANDIDATE|0.025235|0.163217|0.999989|0.581603|0.999988|0.000012|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|0.999960|0.999957|0.999954|0.999914|0.999957|0.999957|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ANY_CANDIDATE|0.030370|0.043712|0.077961|0.022344|0.064178|0.064174|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|313855|12|15|184490451759|12 / 184490451771|15 / 313870|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ANY_CANDIDATE|51229|1978852|262641|184488472919|1978852 / 184490451771|262641 / 313870|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999945|11482166 / 184490451771 (0.006224%)|15 / 313870 (0.004779%)|0.033039|0.051834|0.163107|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ALL_CANDIDATES|0.581603|2008917 / 184490451771 (0.001089%)|262641 / 313870 (83.678274%)|0.029943|0.043158|0.063704|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.026607|0.999952|0.999938|0.999945|0.999938|0.000062|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ALL_CANDIDATES|0.024867|0.163217|0.999989|0.581603|0.999988|0.000012|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.033039|0.051834|0.120237|0.026607|0.163112|0.163107|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ALL_CANDIDATES|0.029943|0.043158|0.077254|0.022055|0.063708|0.063704|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|313855|11482166|15|184478969605|11482166 / 184490451771|15 / 313870|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ALL_CANDIDATES|51229|2008917|262641|184488442854|2008917 / 184490451771|262641 / 313870|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|21854|1149874|10332280|29208|4.808384%|1355|2838145|
+|HUNSPELL ENGLISH LUCENE FILTER|5227|3134|26931|6837|1.125545%|4|614296|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **15 result rows**, **11 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.965820|1148489 / 170474840204 (0.000674%)|21319 / 311891 (6.835401%)|0.239424|0.331902|0.433722|
+|2|ENGLISH LUCENE PORTER COPIED|PRIMARY_OUTPUT|0.954900|1552702 / 170474840204 (0.000911%)|28130 / 311891 (9.019177%)|0.185277|0.264166|0.374937|
+|3|ENGLISH LUCENE PORTER FILTER|PRIMARY_OUTPUT|0.954900|1552702 / 170474840204 (0.000911%)|28130 / 311891 (9.019177%)|0.185277|0.264166|0.374937|
+|4|ENGLISH OPENNLP PORTER|PRIMARY_OUTPUT|0.954900|1552702 / 170474840204 (0.000911%)|28130 / 311891 (9.019177%)|0.185277|0.264166|0.374937|
+|5|ENGLISH SNOWBALL PORTER2|PRIMARY_OUTPUT|0.954850|1561891 / 170474840204 (0.000916%)|28161 / 311891 (9.029116%)|0.184375|0.263016|0.373964|
+|6|ENGLISH SNOWBALL ORIGINAL PORTER|PRIMARY_OUTPUT|0.954762|1550615 / 170474840204 (0.000910%)|28216 / 311891 (9.046750%)|0.185431|0.264353|0.375045|
+|7|ENGLISH PAICE HUSK LANCASTER|PRIMARY_OUTPUT|0.952710|3045870 / 170474840204 (0.001787%)|29493 / 311891 (9.456188%)|0.103633|0.155157|0.277170|
+|8|ENGLISH LUCENE KSTEM FILTER|PRIMARY_OUTPUT|0.880820|1367069 / 170474840204 (0.000802%)|74340 / 311891 (23.835250%)|0.176477|0.247899|0.335789|
+|9|ENGLISH LUCENE MINIMAL FILTER|PRIMARY_OUTPUT|0.719516|1120871 / 170474840204 (0.000657%)|174959 / 311891 (56.096200%)|0.128139|0.174470|0.218621|
+|10|HUNSPELL ENGLISH LUCENE FILTER|PRIMARY_OUTPUT|0.573619|1978041 / 170474840204 (0.001160%)|265965 / 311891 (85.274984%)|0.027312|0.039323|0.057799|
+|11|ENGLISH LUCENE POSSESSIVE FILTER|PRIMARY_OUTPUT|0.500005|1113773 / 170474840204 (0.000653%)|311886 / 311891 (99.998397%)|0.000005|0.000007|0.000005|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.201918|0.931646|0.999993|0.965820|0.999993|0.000007|
+|2|ENGLISH LUCENE PORTER COPIED|PRIMARY_OUTPUT|0.154515|0.909808|0.999991|0.954900|0.999991|0.000009|
+|3|ENGLISH LUCENE PORTER FILTER|PRIMARY_OUTPUT|0.154515|0.909808|0.999991|0.954900|0.999991|0.000009|
+|4|ENGLISH OPENNLP PORTER|PRIMARY_OUTPUT|0.154515|0.909808|0.999991|0.954900|0.999991|0.000009|
+|5|ENGLISH SNOWBALL PORTER2|PRIMARY_OUTPUT|0.153731|0.909709|0.999991|0.954850|0.999991|0.000009|
+|6|ENGLISH SNOWBALL ORIGINAL PORTER|PRIMARY_OUTPUT|0.154651|0.909532|0.999991|0.954762|0.999991|0.000009|
+|7|ENGLISH PAICE HUSK LANCASTER|PRIMARY_OUTPUT|0.084848|0.905438|0.999982|0.952710|0.999982|0.000018|
+|8|ENGLISH LUCENE KSTEM FILTER|PRIMARY_OUTPUT|0.148042|0.761647|0.999992|0.880820|0.999992|0.000008|
+|9|ENGLISH LUCENE MINIMAL FILTER|PRIMARY_OUTPUT|0.108866|0.439038|0.999993|0.719516|0.999992|0.000008|
+|10|HUNSPELL ENGLISH LUCENE FILTER|PRIMARY_OUTPUT|0.022691|0.147250|0.999988|0.573619|0.999987|0.000013|
+|11|ENGLISH LUCENE POSSESSIVE FILTER|PRIMARY_OUTPUT|0.000004|0.000016|0.999993|0.500005|0.999992|0.000008|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.239424|0.331902|0.540775|0.198970|0.433723|0.433722|
+|2|ENGLISH LUCENE PORTER COPIED|PRIMARY_OUTPUT|0.185277|0.264166|0.460049|0.152184|0.374939|0.374937|
+|3|ENGLISH LUCENE PORTER FILTER|PRIMARY_OUTPUT|0.185277|0.264166|0.460049|0.152184|0.374939|0.374937|
+|4|ENGLISH OPENNLP PORTER|PRIMARY_OUTPUT|0.185277|0.264166|0.460049|0.152184|0.374939|0.374937|
+|5|ENGLISH SNOWBALL PORTER2|PRIMARY_OUTPUT|0.184375|0.263016|0.458637|0.151421|0.373966|0.373964|
+|6|ENGLISH SNOWBALL ORIGINAL PORTER|PRIMARY_OUTPUT|0.185431|0.264353|0.460234|0.152308|0.375047|0.375045|
+|7|ENGLISH PAICE HUSK LANCASTER|PRIMARY_OUTPUT|0.103633|0.155157|0.308576|0.084103|0.277173|0.277170|
+|8|ENGLISH LUCENE KSTEM FILTER|PRIMARY_OUTPUT|0.176477|0.247899|0.416437|0.141487|0.335791|0.335789|
+|9|ENGLISH LUCENE MINIMAL FILTER|PRIMARY_OUTPUT|0.128139|0.174470|0.273277|0.095572|0.218624|0.218621|
+|10|HUNSPELL ENGLISH LUCENE FILTER|PRIMARY_OUTPUT|0.027312|0.039323|0.070190|0.020056|0.057804|0.057799|
+|11|ENGLISH LUCENE POSSESSIVE FILTER|PRIMARY_OUTPUT|0.000005|0.000007|0.000011|0.000004|0.000008|0.000005|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.331900|0.993959|0.997731|0.995842|0.995842|
+|2|ENGLISH LUCENE PORTER COPIED|PRIMARY_OUTPUT|0.264164|0.968645|0.997109|0.982671|0.982671|
+|3|ENGLISH LUCENE PORTER FILTER|PRIMARY_OUTPUT|0.264164|0.968645|0.997109|0.982671|0.982671|
+|4|ENGLISH OPENNLP PORTER|PRIMARY_OUTPUT|0.264164|0.968645|0.997109|0.982671|0.982671|
+|5|ENGLISH SNOWBALL PORTER2|PRIMARY_OUTPUT|0.263014|0.968020|0.997096|0.982343|0.982343|
+|6|ENGLISH SNOWBALL ORIGINAL PORTER|PRIMARY_OUTPUT|0.264351|0.968894|0.997102|0.982795|0.982795|
+|7|ENGLISH PAICE HUSK LANCASTER|PRIMARY_OUTPUT|0.155154|0.936077|0.996487|0.965338|0.965338|
+|8|ENGLISH LUCENE KSTEM FILTER|PRIMARY_OUTPUT|0.247897|0.979822|0.991991|0.985869|0.985869|
+|9|ENGLISH LUCENE MINIMAL FILTER|PRIMARY_OUTPUT|0.174467|0.994994|0.980520|0.987704|0.987704|
+|10|HUNSPELL ENGLISH LUCENE FILTER|PRIMARY_OUTPUT|0.039320|0.993066|0.962317|0.977450|0.977450|
+|11|ENGLISH LUCENE POSSESSIVE FILTER|PRIMARY_OUTPUT|0.000004|0.995605|0.956423|0.975621|0.975621|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|290572|1148489|21319|170473691715|1148489 / 170474840204|21319 / 311891|
+|2|ENGLISH LUCENE PORTER COPIED|PRIMARY_OUTPUT|283761|1552702|28130|170473287502|1552702 / 170474840204|28130 / 311891|
+|3|ENGLISH LUCENE PORTER FILTER|PRIMARY_OUTPUT|283761|1552702|28130|170473287502|1552702 / 170474840204|28130 / 311891|
+|4|ENGLISH OPENNLP PORTER|PRIMARY_OUTPUT|283761|1552702|28130|170473287502|1552702 / 170474840204|28130 / 311891|
+|5|ENGLISH SNOWBALL PORTER2|PRIMARY_OUTPUT|283730|1561891|28161|170473278313|1561891 / 170474840204|28161 / 311891|
+|6|ENGLISH SNOWBALL ORIGINAL PORTER|PRIMARY_OUTPUT|283675|1550615|28216|170473289589|1550615 / 170474840204|28216 / 311891|
+|7|ENGLISH PAICE HUSK LANCASTER|PRIMARY_OUTPUT|282398|3045870|29493|170471794334|3045870 / 170474840204|29493 / 311891|
+|8|ENGLISH LUCENE KSTEM FILTER|PRIMARY_OUTPUT|237551|1367069|74340|170473473135|1367069 / 170474840204|74340 / 311891|
+|9|ENGLISH LUCENE MINIMAL FILTER|PRIMARY_OUTPUT|136932|1120871|174959|170473719333|1120871 / 170474840204|174959 / 311891|
+|10|HUNSPELL ENGLISH LUCENE FILTER|PRIMARY_OUTPUT|45926|1978041|265965|170472862163|1978041 / 170474840204|265965 / 311891|
+|11|ENGLISH LUCENE POSSESSIVE FILTER|PRIMARY_OUTPUT|5|1113773|311886|170473726431|1113773 / 170474840204|311886 / 311891|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 170474840204 (0.000000%)|0 / 311891 (0.000000%)|1.000000|1.000000|1.000000|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ANY_CANDIDATE|0.581994|1974950 / 170474840204 (0.001158%)|260741 / 311891 (83.600040%)|0.030387|0.043756|0.064341|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ANY_CANDIDATE|0.025246|0.164000|0.999988|0.581994|0.999987|0.000013|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ANY_CANDIDATE|0.030387|0.043756|0.078123|0.022367|0.064345|0.064341|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|311891|0|0|170474840204|0 / 170474840204|0 / 311891|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ANY_CANDIDATE|51150|1974950|260741|170472865254|1974950 / 170474840204|260741 / 311891|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999966|11470018 / 170474840204 (0.006728%)|0 / 311891 (0.000000%)|0.032872|0.051579|0.162697|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ALL_CANDIDATES|0.581994|2004598 / 170474840204 (0.001176%)|260741 / 311891 (83.600040%)|0.029965|0.043208|0.063875|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.026472|1.000000|0.999933|0.999966|0.999933|0.000067|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ALL_CANDIDATES|0.024881|0.164000|0.999988|0.581994|0.999987|0.000013|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.032872|0.051579|0.119687|0.026472|0.162702|0.162697|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ALL_CANDIDATES|0.029965|0.043208|0.077422|0.022081|0.063879|0.063875|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|311891|11470018|0|170463370186|11470018 / 170474840204|0 / 311891|
+|2|HUNSPELL ENGLISH LUCENE FILTER|ALL_CANDIDATES|51150|2004598|260741|170472835606|2004598 / 170474840204|260741 / 311891|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|21319|1148489|10321529|28826|4.936720%|1355|2812871|
+|HUNSPELL ENGLISH LUCENE FILTER|5224|3091|26557|6786|1.162165%|4|590716|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `US_UK`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/finnish.md
+++ b/docs/benchmarks/languages/finnish.md
@@ -0,0 +1,408 @@
+# Finnish Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Finnish. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `FI_FI` | 57,027 | 1,865,215 | 110,525 | 1,754,690 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **1,865,215**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 745 | 0.040% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 1,176,003 | 63.049% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 565,585 | 30.323% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 116,946 | 6.270% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 5,936 | 0.318% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 98.661% | 98.803% | 96.408% | Full Radixor dictionary patch-command stemmer. |
+| Lucene SnowballFilter | 10.991% | 10.268% | 22.471% | Lucene TokenFilter integration path around the Snowball algorithm. |
+| Official Snowball direct | 10.991% | 10.268% | 22.471% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+| Lucene FinnishLightStemFilter | 4.351% | 4.294% | 5.264% | Light suffix stemmer; intentionally narrower than a dictionary-derived stemmer. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `finnishRadixor` | 308.076 | 15.529 | 175.6 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene FinnishLightStemFilter | `finnishLuceneFinnishLightStemFilter` | 175.250 | 46.995 | 99.9 | 0.569 | Light Finnish suffix stemmer. |
+| Official Snowball direct | `snowballDirect[FINNISH]` | 264.652 | 63.054 | 150.8 | 0.859 | Official Snowball generated Java stemmer; direct API. |
+| Lucene SnowballFilter | `luceneSnowballFilter[FINNISH]` | 374.883 | 238.157 | 213.6 | 1.217 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `FI_FI` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/fi_fi/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.984594** among 4 deterministic stemmers. The runner-up is `SNOWBALL FINNISH LUCENE FILTER` at 0.740353, a difference of 0.244242. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.988068** among 4 deterministic stemmers. The runner-up is `SNOWBALL FINNISH DIRECT` at 0.738400, a difference of 0.249668. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **6 result rows**, **4 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.984594|731279 / 1641126814491 (0.000045%)|971268 / 31523695 (3.081073%)|0.975128|0.972893|0.972899|
+|2|SNOWBALL FINNISH LUCENE FILTER|PRIMARY_OUTPUT|0.740353|1922153 / 1641126814491 (0.000117%)|16370057 / 31523695 (51.929372%)|0.758996|0.623613|0.653138|
+|3|SNOWBALL FINNISH DIRECT|PRIMARY_OUTPUT|0.739729|1544812 / 1641126814491 (0.000094%)|16409363 / 31523695 (52.054060%)|0.769880|0.627374|0.659540|
+|4|FINNISH LUCENE FINNISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.695969|2223150 / 1641126814491 (0.000135%)|19168306 / 31523695 (60.806025%)|0.687649|0.536000|0.576338|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.976624|0.969189|1.000000|0.984594|0.999999|0.000001|
+|2|SNOWBALL FINNISH LUCENE FILTER|PRIMARY_OUTPUT|0.887434|0.480706|0.999999|0.740353|0.999989|0.000011|
+|3|SNOWBALL FINNISH DIRECT|PRIMARY_OUTPUT|0.907269|0.479459|0.999999|0.739729|0.999989|0.000011|
+|4|FINNISH LUCENE FINNISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.847505|0.391940|0.999999|0.695969|0.999987|0.000013|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.975128|0.972893|0.970667|0.947216|0.972900|0.972899|
+|2|SNOWBALL FINNISH LUCENE FILTER|PRIMARY_OUTPUT|0.758996|0.623613|0.529216|0.453080|0.653142|0.653138|
+|3|SNOWBALL FINNISH DIRECT|PRIMARY_OUTPUT|0.769880|0.627374|0.529384|0.457061|0.659544|0.659540|
+|4|FINNISH LUCENE FINNISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.687649|0.536000|0.439152|0.366120|0.576343|0.576338|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.972892|0.996085|0.993746|0.994914|0.994914|
+|2|SNOWBALL FINNISH LUCENE FILTER|PRIMARY_OUTPUT|0.623608|0.990718|0.904385|0.945585|0.945585|
+|3|SNOWBALL FINNISH DIRECT|PRIMARY_OUTPUT|0.627369|0.991872|0.904139|0.945975|0.945975|
+|4|FINNISH LUCENE FINNISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.535994|0.988126|0.886473|0.934544|0.934544|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|30552427|731279|971268|1641126083212|731279 / 1641126814491|971268 / 31523695|
+|2|SNOWBALL FINNISH LUCENE FILTER|PRIMARY_OUTPUT|15153638|1922153|16370057|1641124892338|1922153 / 1641126814491|16370057 / 31523695|
+|3|SNOWBALL FINNISH DIRECT|PRIMARY_OUTPUT|15114332|1544812|16409363|1641125269679|1544812 / 1641126814491|16409363 / 31523695|
+|4|FINNISH LUCENE FINNISH LIGHT STEM FILTER|PRIMARY_OUTPUT|12355389|2223150|19168306|1641124591341|2223150 / 1641126814491|19168306 / 31523695|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 1641126814491 (0.000000%)|0 / 31523695 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|31523695|0|0|1641126814491|0 / 1641126814491|0 / 31523695|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999999|1683575 / 1641126814491 (0.000103%)|0 / 31523695 (0.000000%)|0.959025|0.973991|0.974320|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.949301|1.000000|0.999999|0.999999|0.999999|0.000001|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.959025|0.973991|0.989432|0.949301|0.974321|0.974320|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|31523695|1683575|0|1641125130916|1683575 / 1641126814491|0 / 31523695|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|971268|731279|952296|57328|3.164291%|6|1876272|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **6 result rows**, **4 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.988068|730145 / 1543589444152 (0.000047%)|735305 / 30813833 (2.386282%)|0.976268|0.976219|0.976218|
+|2|SNOWBALL FINNISH DIRECT|PRIMARY_OUTPUT|0.738400|1513705 / 1543589444152 (0.000098%)|16121763 / 30813833 (52.319888%)|0.768117|0.624934|0.657464|
+|3|SNOWBALL FINNISH LUCENE FILTER|PRIMARY_OUTPUT|0.738400|1513705 / 1543589444152 (0.000098%)|16121763 / 30813833 (52.319888%)|0.768117|0.624934|0.657464|
+|4|FINNISH LUCENE FINNISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.694529|1806392 / 1543589444152 (0.000117%)|18825444 / 30813833 (61.094133%)|0.697056|0.537492|0.581469|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.976301|0.976137|1.000000|0.988068|0.999999|0.000001|
+|2|SNOWBALL FINNISH DIRECT|PRIMARY_OUTPUT|0.906595|0.476801|0.999999|0.738400|0.999989|0.000011|
+|3|SNOWBALL FINNISH LUCENE FILTER|PRIMARY_OUTPUT|0.906595|0.476801|0.999999|0.738400|0.999989|0.000011|
+|4|FINNISH LUCENE FINNISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.869053|0.389059|0.999999|0.694529|0.999987|0.000013|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.976268|0.976219|0.976170|0.953543|0.976219|0.976218|
+|2|SNOWBALL FINNISH DIRECT|PRIMARY_OUTPUT|0.768117|0.624934|0.526744|0.454475|0.657469|0.657464|
+|3|SNOWBALL FINNISH LUCENE FILTER|PRIMARY_OUTPUT|0.768117|0.624934|0.526744|0.454475|0.657469|0.657464|
+|4|FINNISH LUCENE FINNISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.697056|0.537492|0.437372|0.367514|0.581474|0.581469|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.976218|0.996000|0.996069|0.996035|0.996035|
+|2|SNOWBALL FINNISH DIRECT|PRIMARY_OUTPUT|0.624929|0.991732|0.902933|0.945252|0.945252|
+|3|SNOWBALL FINNISH LUCENE FILTER|PRIMARY_OUTPUT|0.624929|0.991732|0.902933|0.945252|0.945252|
+|4|FINNISH LUCENE FINNISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.537486|0.989268|0.885294|0.934397|0.934397|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|30078528|730145|735305|1543588714007|730145 / 1543589444152|735305 / 30813833|
+|2|SNOWBALL FINNISH DIRECT|PRIMARY_OUTPUT|14692070|1513705|16121763|1543587930447|1513705 / 1543589444152|16121763 / 30813833|
+|3|SNOWBALL FINNISH LUCENE FILTER|PRIMARY_OUTPUT|14692070|1513705|16121763|1543587930447|1513705 / 1543589444152|16121763 / 30813833|
+|4|FINNISH LUCENE FINNISH LIGHT STEM FILTER|PRIMARY_OUTPUT|11988389|1806392|18825444|1543587637760|1806392 / 1543589444152|18825444 / 30813833|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 1543589444152 (0.000000%)|0 / 30813833 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|30813833|0|0|1543589444152|0 / 1543589444152|0 / 30813833|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999999|1653320 / 1543589444152 (0.000107%)|0 / 30813833 (0.000000%)|0.958843|0.973873|0.974205|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.949077|1.000000|0.999999|0.999999|0.999999|0.000001|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.958843|0.973873|0.989383|0.949077|0.974206|0.974205|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|30813833|1653320|0|1543587790832|1653320 / 1543589444152|0 / 30813833|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|735305|730145|923175|44331|2.523029%|6|1805864|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `FI_FI`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/french.md
+++ b/docs/benchmarks/languages/french.md
@@ -0,0 +1,454 @@
+# French Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for French. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `FR_FR` | 59,240 | 474,110 | 108,141 | 365,969 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **474,110**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 5,370 | 1.133% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 185,263 | 39.076% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 153,886 | 32.458% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 116,519 | 24.576% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 13,072 | 2.757% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 94.831% | 94.859% | 94.734% | Full Radixor dictionary patch-command stemmer. |
+| Lucene HunspellStemFilter | 68.923% | 63.617% | 86.876% | Benchmark-only French Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Lucene FrenchMinimalStemFilter | 11.472% | 6.236% | 29.192% | Minimal suffix reducer; narrow baseline, not a full stemmer. |
+| Lucene SnowballFilter | 8.551% | 5.183% | 19.952% | Lucene TokenFilter integration path around the Snowball algorithm. |
+| Official Snowball direct | 8.462% | 5.067% | 19.952% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+| Lucene FrenchLightStemFilter | 6.377% | 3.965% | 14.540% | Light suffix stemmer; intentionally narrower than a dictionary-derived stemmer. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `frenchRadixor` | 47.033 | 4.146 | 128.5 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene HunspellStemFilter | `luceneHunspellStemFilter` | 1664.935 | 65.928 | 4549.4 | 35.399 | Benchmark-only French Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Lucene FrenchMinimalStemFilter | `frenchLuceneFrenchMinimalStemFilter` | 19.234 | 2.098 | 52.6 | 0.409 | Minimal French suffix reducer; narrow baseline. |
+| Lucene FrenchLightStemFilter | `frenchLuceneFrenchLightStemFilter` | 30.560 | 3.680 | 83.5 | 0.650 | Light French suffix stemmer. |
+| Official Snowball direct | `snowballDirect[FRENCH]` | 111.057 | 8.172 | 303.5 | 2.361 | Official Snowball generated Java stemmer; direct API. |
+| Lucene SnowballFilter | `luceneSnowballFilter[FRENCH]` | 123.648 | 3.500 | 337.9 | 2.629 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `FR_FR` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/fr_fr/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.956992** among 6 deterministic stemmers. The runner-up is `SNOWBALL FRENCH DIRECT` at 0.845262, a difference of 0.111731. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.957224** among 6 deterministic stemmers. The runner-up is `SNOWBALL FRENCH DIRECT` at 0.845414, a difference of 0.111810. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **10 result rows**, **6 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.956992|318767 / 90396104830 (0.000353%)|469160 / 5454615 (8.601157%)|0.934603|0.926765|0.926851|
+|2|SNOWBALL FRENCH DIRECT|PRIMARY_OUTPUT|0.845262|1654723 / 90396104830 (0.001831%)|1687975 / 5454615 (30.945814%)|0.693926|0.692653|0.692638|
+|3|SNOWBALL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.844999|1661388 / 90396104830 (0.001838%)|1690838 / 5454615 (30.998301%)|0.693010|0.691885|0.691869|
+|4|HUNSPELL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.813742|776728 / 90396104830 (0.000859%)|2031881 / 5454615 (37.250677%)|0.769069|0.709075|0.715131|
+|5|FRENCH LUCENE FRENCH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.518587|276403 / 90396104830 (0.000306%)|5251833 / 5454615 (96.282377%)|0.137547|0.068348|0.125415|
+|6|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.516830|160438 / 90396104830 (0.000177%)|5271003 / 5454615 (96.633823%)|0.134400|0.063329|0.134021|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.939903|0.913988|0.999996|0.956992|0.999991|0.000009|
+|2|SNOWBALL FRENCH DIRECT|PRIMARY_OUTPUT|0.694777|0.690542|0.999982|0.845262|0.999963|0.000037|
+|3|SNOWBALL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.693763|0.690017|0.999982|0.844999|0.999963|0.000037|
+|4|HUNSPELL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.815041|0.627493|0.999991|0.813742|0.999969|0.000031|
+|5|FRENCH LUCENE FRENCH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.423181|0.037176|0.999997|0.518587|0.999939|0.000061|
+|6|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.533678|0.033662|0.999998|0.516830|0.999940|0.000060|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.934603|0.926765|0.919056|0.863524|0.926855|0.926851|
+|2|SNOWBALL FRENCH DIRECT|PRIMARY_OUTPUT|0.693926|0.692653|0.691385|0.529816|0.692656|0.692638|
+|3|SNOWBALL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.693010|0.691885|0.690763|0.528917|0.691887|0.691869|
+|4|HUNSPELL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.769069|0.709075|0.657765|0.549277|0.715145|0.715131|
+|5|FRENCH LUCENE FRENCH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.137547|0.068348|0.045472|0.035383|0.125428|0.125415|
+|6|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.134400|0.063329|0.041424|0.032700|0.134032|0.134021|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.926760|0.988772|0.985214|0.986990|0.986990|
+|2|SNOWBALL FRENCH DIRECT|PRIMARY_OUTPUT|0.692635|0.959459|0.944948|0.952148|0.952148|
+|3|SNOWBALL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.691866|0.958698|0.944715|0.951655|0.951655|
+|4|HUNSPELL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.709060|0.978337|0.913706|0.944918|0.944918|
+|5|FRENCH LUCENE FRENCH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.068339|0.974110|0.812376|0.885922|0.885922|
+|6|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.063322|0.984019|0.810979|0.889158|0.889158|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|4985455|318767|469160|90395786063|318767 / 90396104830|469160 / 5454615|
+|2|SNOWBALL FRENCH DIRECT|PRIMARY_OUTPUT|3766640|1654723|1687975|90394450107|1654723 / 90396104830|1687975 / 5454615|
+|3|SNOWBALL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|3763777|1661388|1690838|90394443442|1661388 / 90396104830|1690838 / 5454615|
+|4|HUNSPELL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|3422734|776728|2031881|90395328102|776728 / 90396104830|2031881 / 5454615|
+|5|FRENCH LUCENE FRENCH LIGHT STEM FILTER|PRIMARY_OUTPUT|202782|276403|5251833|90395828427|276403 / 90396104830|5251833 / 5454615|
+|6|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|PRIMARY_OUTPUT|183612|160438|5271003|90395944392|160438 / 90396104830|5271003 / 5454615|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|0.999979|12 / 90396104830 (0.000000%)|232 / 5454615 (0.004253%)|0.999990|0.999978|0.999978|
+|2|HUNSPELL FRENCH LUCENE FILTER|ANY_CANDIDATE|0.830964|745831 / 90396104830 (0.000825%)|1844003 / 5454615 (33.806291%)|0.789019|0.736029|0.740670|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|0.999998|0.999957|1.000000|0.999979|1.000000|0.000000|
+|2|HUNSPELL FRENCH LUCENE FILTER|ANY_CANDIDATE|0.828798|0.661937|0.999992|0.830964|0.999971|0.000029|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|0.999990|0.999978|0.999966|0.999955|0.999978|0.999978|
+|2|HUNSPELL FRENCH LUCENE FILTER|ANY_CANDIDATE|0.789019|0.736029|0.689709|0.582315|0.740684|0.740670|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL FRENCH LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|5454383|12|232|90396104818|12 / 90396104830|232 / 5454615|
+|2|HUNSPELL FRENCH LUCENE FILTER|ANY_CANDIDATE|3610612|745831|1844003|90395358999|745831 / 90396104830|1844003 / 5454615|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999973|1056255 / 90396104830 (0.001168%)|232 / 5454615 (0.004253%)|0.865853|0.911704|0.915270|
+|2|HUNSPELL FRENCH LUCENE FILTER|ALL_CANDIDATES|0.830963|1043199 / 90396104830 (0.001154%)|1844003 / 5454615 (33.806291%)|0.750028|0.714377|0.716613|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.837765|0.999957|0.999988|0.999973|0.999988|0.000012|
+|2|HUNSPELL FRENCH LUCENE FILTER|ALL_CANDIDATES|0.775840|0.661937|0.999988|0.830963|0.999968|0.000032|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.865853|0.911704|0.962682|0.837735|0.915275|0.915270|
+|2|HUNSPELL FRENCH LUCENE FILTER|ALL_CANDIDATES|0.750028|0.714377|0.681961|0.555666|0.716629|0.716613|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL FRENCH LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|5454383|1056255|232|90395048575|1056255 / 90396104830|232 / 5454615|
+|2|HUNSPELL FRENCH LUCENE FILTER|ALL_CANDIDATES|3610612|1043199|1844003|90395061631|1043199 / 90396104830|1844003 / 5454615|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|468928|318755|737488|43040|10.122057%|56|477024|
+|HUNSPELL FRENCH LUCENE FILTER|187878|30897|266471|13511|3.177489%|4|439015|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **10 result rows**, **6 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.957224|315266 / 88712126506 (0.000355%)|465436 / 5440559 (8.554930%)|0.935099|0.927248|0.927334|
+|2|SNOWBALL FRENCH DIRECT|PRIMARY_OUTPUT|0.845414|1646111 / 88712126506 (0.001856%)|1681970 / 5440559 (30.915389%)|0.694508|0.693130|0.693115|
+|3|SNOWBALL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.845163|1641925 / 88712126506 (0.001851%)|1684703 / 5440559 (30.965623%)|0.694714|0.693068|0.693055|
+|4|HUNSPELL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.813617|763305 / 88712126506 (0.000860%)|2028011 / 5440559 (37.275784%)|0.770537|0.709734|0.715938|
+|5|FRENCH LUCENE FRENCH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.518442|262689 / 88712126506 (0.000296%)|5239869 / 5440559 (96.311225%)|0.137571|0.067985|0.126383|
+|6|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.516697|147476 / 88712126506 (0.000166%)|5258873 / 5440559 (96.660527%)|0.134439|0.062979|0.135757|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.940408|0.914451|0.999996|0.957224|0.999991|0.000009|
+|2|SNOWBALL FRENCH DIRECT|PRIMARY_OUTPUT|0.695430|0.690846|0.999981|0.845414|0.999962|0.000038|
+|3|SNOWBALL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.695815|0.690344|0.999981|0.845163|0.999963|0.000037|
+|4|HUNSPELL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.817210|0.627242|0.999991|0.813617|0.999969|0.000031|
+|5|FRENCH LUCENE FRENCH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.433101|0.036888|0.999997|0.518442|0.999938|0.000062|
+|6|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.551965|0.033395|0.999998|0.516697|0.999939|0.000061|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.935099|0.927248|0.919527|0.864363|0.927338|0.927334|
+|2|SNOWBALL FRENCH DIRECT|PRIMARY_OUTPUT|0.694508|0.693130|0.691758|0.530374|0.693134|0.693115|
+|3|SNOWBALL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.694714|0.693068|0.691431|0.530302|0.693074|0.693055|
+|4|HUNSPELL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.770537|0.709734|0.657826|0.550068|0.715953|0.715938|
+|5|FRENCH LUCENE FRENCH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.137571|0.067985|0.045148|0.035189|0.126397|0.126383|
+|6|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.134439|0.062979|0.041121|0.032513|0.135767|0.135757|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.927243|0.988916|0.985550|0.987230|0.987230|
+|2|SNOWBALL FRENCH DIRECT|PRIMARY_OUTPUT|0.693112|0.959521|0.944537|0.951970|0.951970|
+|3|SNOWBALL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.693050|0.959566|0.944385|0.951915|0.951915|
+|4|HUNSPELL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|0.709719|0.979328|0.913162|0.945088|0.945088|
+|5|FRENCH LUCENE FRENCH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.067976|0.975086|0.811144|0.885591|0.885591|
+|6|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.062973|0.985086|0.809774|0.888868|0.888868|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|4975123|315266|465436|88711811240|315266 / 88712126506|465436 / 5440559|
+|2|SNOWBALL FRENCH DIRECT|PRIMARY_OUTPUT|3758589|1646111|1681970|88710480395|1646111 / 88712126506|1681970 / 5440559|
+|3|SNOWBALL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|3755856|1641925|1684703|88710484581|1641925 / 88712126506|1684703 / 5440559|
+|4|HUNSPELL FRENCH LUCENE FILTER|PRIMARY_OUTPUT|3412548|763305|2028011|88711363201|763305 / 88712126506|2028011 / 5440559|
+|5|FRENCH LUCENE FRENCH LIGHT STEM FILTER|PRIMARY_OUTPUT|200690|262689|5239869|88711863817|262689 / 88712126506|5239869 / 5440559|
+|6|FRENCH LUCENE FRENCH MINIMAL STEM FILTER|PRIMARY_OUTPUT|181686|147476|5258873|88711979030|147476 / 88712126506|5258873 / 5440559|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 88712126506 (0.000000%)|0 / 5440559 (0.000000%)|1.000000|1.000000|1.000000|
+|2|HUNSPELL FRENCH LUCENE FILTER|ANY_CANDIDATE|0.830852|733584 / 88712126506 (0.000827%)|1840476 / 5440559 (33.828803%)|0.790351|0.736648|0.741404|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+|2|HUNSPELL FRENCH LUCENE FILTER|ANY_CANDIDATE|0.830724|0.661712|0.999992|0.830852|0.999971|0.000029|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+|2|HUNSPELL FRENCH LUCENE FILTER|ANY_CANDIDATE|0.790351|0.736648|0.689779|0.583090|0.741418|0.741404|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL FRENCH LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|5440559|0|0|88712126506|0 / 88712126506|0 / 5440559|
+|2|HUNSPELL FRENCH LUCENE FILTER|ANY_CANDIDATE|3600083|733584|1840476|88711392922|733584 / 88712126506|1840476 / 5440559|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999995|938985 / 88712126506 (0.001058%)|0 / 5440559 (0.000000%)|0.878679|0.920560|0.923474|
+|2|HUNSPELL FRENCH LUCENE FILTER|ALL_CANDIDATES|0.830850|1027635 / 88712126506 (0.001158%)|1840476 / 5440559 (33.828803%)|0.751538|0.715134|0.717460|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.852813|1.000000|0.999989|0.999995|0.999989|0.000011|
+|2|HUNSPELL FRENCH LUCENE FILTER|ALL_CANDIDATES|0.777939|0.661712|0.999988|0.830850|0.999968|0.000032|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.878679|0.920560|0.966634|0.852813|0.923479|0.923474|
+|2|HUNSPELL FRENCH LUCENE FILTER|ALL_CANDIDATES|0.751538|0.715134|0.682093|0.556582|0.717476|0.717460|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL FRENCH LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|5440559|938985|0|88711187521|938985 / 88712126506|0 / 5440559|
+|2|HUNSPELL FRENCH LUCENE FILTER|ALL_CANDIDATES|3600083|1027635|1840476|88711098871|1027635 / 88712126506|1840476 / 5440559|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|465436|315266|623719|41130|9.764239%|56|468574|
+|HUNSPELL FRENCH LUCENE FILTER|187535|29721|264330|13437|3.189936%|4|434961|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `FR_FR`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/german.md
+++ b/docs/benchmarks/languages/german.md
@@ -0,0 +1,478 @@
+# German Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for German. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `DE_DE` | 39,315 | 213,440 | 73,799 | 139,641 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **213,440**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 3,627 | 1.699% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 48,605 | 22.772% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 80,443 | 37.689% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 75,717 | 35.475% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 5,048 | 2.365% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 92.725% | 92.847% | 92.396% | Full Radixor dictionary patch-command stemmer. |
+| Lucene HunspellStemFilter | 47.064% | 29.661% | 93.678% | Benchmark-only German Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| CISTEM (German) | 24.675% | 23.724% | 27.222% | Benchmark-only CISTEM implementation. |
+| Lucene GermanLightStemFilter | 37.434% | 35.465% | 42.707% | Light suffix stemmer; intentionally narrower than a dictionary-derived stemmer. |
+| Lucene GermanMinimalStemFilter | 27.640% | 24.951% | 34.844% | Minimal suffix reducer; narrow baseline, not a full stemmer. |
+| Lucene SnowballFilter | 30.956% | 28.853% | 36.589% | Lucene TokenFilter integration path around the Snowball algorithm. |
+| Official Snowball direct | 30.481% | 29.027% | 34.376% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+| Lucene GermanStemFilter | 21.559% | 19.312% | 27.576% | German Lucene stemming TokenFilter; broader than minimal/light variants. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `germanRadixor` | 41.166 | 2.396 | 294.8 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| CISTEM | `germanCistem` | 248.392 | 12.294 | 1778.8 | 6.034 | Benchmark-only CISTEM implementation. |
+| Lucene HunspellStemFilter | `luceneHunspellStemFilter` | 281.322 | 3.411 | 2014.6 | 6.834 | Benchmark-only German Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Lucene GermanMinimalStemFilter | `germanLuceneGermanMinimalStemFilter` | 23.562 | 0.969 | 168.7 | 0.572 | Minimal German suffix reduction; narrow baseline. |
+| Lucene GermanLightStemFilter | `germanLuceneGermanLightStemFilter` | 24.410 | 1.034 | 174.8 | 0.593 | Light German suffix stemmer; narrower than a dictionary stemmer. |
+| Lucene GermanStemFilter | `germanLuceneGermanStemFilter` | 71.039 | 4.443 | 508.7 | 1.726 | Older German stemming TokenFilter with normalization requirements. |
+| Lucene SnowballFilter | `luceneSnowballFilter[GERMAN]` | 105.771 | 9.617 | 757.4 | 2.569 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+| Official Snowball direct | `snowballDirect[GERMAN]` | 100.688 | 9.018 | 721.0 | 2.446 | Official Snowball generated Java stemmer; direct API. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `DE_DE` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/de_de/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.907901** among 8 deterministic stemmers. The runner-up is `GERMAN CISTEM` at 0.880770, a difference of 0.027131. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.966157** among 8 deterministic stemmers. The runner-up is `GERMAN CISTEM` at 0.915288, a difference of 0.050869. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **12 result rows**, **8 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.907901|98192 / 44095245979 (0.000223%)|254903 / 1383872 (18.419550%)|0.897073|0.864768|0.866326|
+|2|GERMAN CISTEM|PRIMARY_OUTPUT|0.880770|477122 / 44095245979 (0.001082%)|329983 / 1383872 (23.844908%)|0.701852|0.723109|0.724023|
+|3|SNOWBALL GERMAN DIRECT|PRIMARY_OUTPUT|0.778614|190680 / 44095245979 (0.000432%)|612734 / 1383872 (44.276783%)|0.737064|0.657494|0.668394|
+|4|SNOWBALL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.771357|295701 / 44095245979 (0.000671%)|632816 / 1383872 (45.727929%)|0.674089|0.617993|0.624014|
+|5|GERMAN LUCENE GERMAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.756258|205740 / 44095245979 (0.000467%)|674609 / 1383872 (48.747933%)|0.703092|0.617052|0.630292|
+|6|GERMAN LUCENE GERMAN STEM FILTER|PRIMARY_OUTPUT|0.723772|331871 / 44095245979 (0.000753%)|764518 / 1383872 (55.244849%)|0.596821|0.530474|0.539809|
+|7|HUNSPELL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.641579|203883 / 44095245979 (0.000462%)|992010 / 1383872 (71.683653%)|0.520145|0.395897|0.431563|
+|8|GERMAN LUCENE GERMAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.598139|110840 / 44095245979 (0.000251%)|1112246 / 1383872 (80.372029%)|0.466113|0.307558|0.373350|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.919984|0.815804|0.999998|0.907901|0.999992|0.000008|
+|2|GERMAN CISTEM|PRIMARY_OUTPUT|0.688361|0.761551|0.999989|0.880770|0.999982|0.000018|
+|3|SNOWBALL GERMAN DIRECT|PRIMARY_OUTPUT|0.801750|0.557232|0.999996|0.778614|0.999982|0.000018|
+|4|SNOWBALL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.717508|0.542721|0.999993|0.771357|0.999979|0.000021|
+|5|GERMAN LUCENE GERMAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.775148|0.512521|0.999995|0.756258|0.999980|0.000020|
+|6|GERMAN LUCENE GERMAN STEM FILTER|PRIMARY_OUTPUT|0.651112|0.447552|0.999992|0.723772|0.999975|0.000025|
+|7|HUNSPELL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.657768|0.283163|0.999995|0.641579|0.999973|0.000027|
+|8|GERMAN LUCENE GERMAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.710196|0.196280|0.999997|0.598139|0.999972|0.000028|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.897073|0.864768|0.834709|0.761755|0.866330|0.866326|
+|2|GERMAN CISTEM|PRIMARY_OUTPUT|0.701852|0.723109|0.745694|0.566304|0.724032|0.724023|
+|3|SNOWBALL GERMAN DIRECT|PRIMARY_OUTPUT|0.737064|0.657494|0.593429|0.489751|0.668402|0.668394|
+|4|SNOWBALL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.674089|0.617993|0.570517|0.447171|0.624024|0.624014|
+|5|GERMAN LUCENE GERMAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.703092|0.617052|0.549774|0.446186|0.630301|0.630292|
+|6|GERMAN LUCENE GERMAN STEM FILTER|PRIMARY_OUTPUT|0.596821|0.530474|0.477402|0.360983|0.539820|0.539809|
+|7|HUNSPELL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.520145|0.395897|0.319562|0.246803|0.431574|0.431563|
+|8|GERMAN LUCENE GERMAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.466113|0.307558|0.229493|0.181725|0.373359|0.373350|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.864764|0.989946|0.975085|0.982460|0.982460|
+|2|GERMAN CISTEM|PRIMARY_OUTPUT|0.723100|0.974048|0.975147|0.974597|0.974597|
+|3|SNOWBALL GERMAN DIRECT|PRIMARY_OUTPUT|0.657485|0.983725|0.949324|0.966218|0.966218|
+|4|SNOWBALL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.617983|0.975845|0.942925|0.959102|0.959102|
+|5|GERMAN LUCENE GERMAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.617043|0.980753|0.936533|0.958133|0.958133|
+|6|GERMAN LUCENE GERMAN STEM FILTER|PRIMARY_OUTPUT|0.530462|0.975550|0.942890|0.958942|0.958942|
+|7|HUNSPELL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.395885|0.980463|0.886873|0.931322|0.931322|
+|8|GERMAN LUCENE GERMAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.307549|0.983615|0.896264|0.937910|0.937910|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|1128969|98192|254903|44095147787|98192 / 44095245979|254903 / 1383872|
+|2|GERMAN CISTEM|PRIMARY_OUTPUT|1053889|477122|329983|44094768857|477122 / 44095245979|329983 / 1383872|
+|3|SNOWBALL GERMAN DIRECT|PRIMARY_OUTPUT|771138|190680|612734|44095055299|190680 / 44095245979|612734 / 1383872|
+|4|SNOWBALL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|751056|295701|632816|44094950278|295701 / 44095245979|632816 / 1383872|
+|5|GERMAN LUCENE GERMAN LIGHT STEM FILTER|PRIMARY_OUTPUT|709263|205740|674609|44095040239|205740 / 44095245979|674609 / 1383872|
+|6|GERMAN LUCENE GERMAN STEM FILTER|PRIMARY_OUTPUT|619354|331871|764518|44094914108|331871 / 44095245979|764518 / 1383872|
+|7|HUNSPELL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|391862|203883|992010|44095042096|203883 / 44095245979|992010 / 1383872|
+|8|GERMAN LUCENE GERMAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|271626|110840|1112246|44095135139|110840 / 44095245979|1112246 / 1383872|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|0.959835|1375 / 44095245979 (0.000003%)|111167 / 1383872 (8.033041%)|0.981996|0.957658|0.958475|
+|2|HUNSPELL GERMAN LUCENE FILTER|ANY_CANDIDATE|0.647474|158403 / 44095245979 (0.000359%)|975697 / 1383872 (70.504859%)|0.559116|0.418544|0.460956|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|0.998921|0.919670|1.000000|0.959835|0.999997|0.000003|
+|2|HUNSPELL GERMAN LUCENE FILTER|ANY_CANDIDATE|0.720422|0.294951|0.999996|0.647474|0.999974|0.000026|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|0.981996|0.957658|0.934498|0.918757|0.958476|0.958475|
+|2|HUNSPELL GERMAN LUCENE FILTER|ANY_CANDIDATE|0.559116|0.418544|0.334456|0.264658|0.460966|0.460956|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL GERMAN LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1272705|1375|111167|44095244604|1375 / 44095245979|111167 / 1383872|
+|2|HUNSPELL GERMAN LUCENE FILTER|ANY_CANDIDATE|408175|158403|975697|44095087576|158403 / 44095245979|975697 / 1383872|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.959832|244817 / 44095245979 (0.000555%)|111167 / 1383872 (8.033041%)|0.853711|0.877306|0.878234|
+|2|HUNSPELL GERMAN LUCENE FILTER|ALL_CANDIDATES|0.647473|242551 / 44095245979 (0.000550%)|975697 / 1383872 (70.504859%)|0.511911|0.401234|0.430118|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.838673|0.919670|0.999994|0.959832|0.999992|0.000008|
+|2|HUNSPELL GERMAN LUCENE FILTER|ALL_CANDIDATES|0.627261|0.294951|0.999994|0.647473|0.999972|0.000028|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.853711|0.877306|0.902242|0.781429|0.878238|0.878234|
+|2|HUNSPELL GERMAN LUCENE FILTER|ALL_CANDIDATES|0.511911|0.401234|0.329907|0.250965|0.430130|0.430118|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL GERMAN LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|1272705|244817|111167|44095001162|244817 / 44095245979|111167 / 1383872|
+|2|HUNSPELL GERMAN LUCENE FILTER|ALL_CANDIDATES|408175|242551|975697|44095003428|242551 / 44095245979|975697 / 1383872|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|143736|96817|146625|48574|16.356314%|8|361016|
+|HUNSPELL GERMAN LUCENE FILTER|16313|45480|38668|7891|2.657135%|3|305052|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **12 result rows**, **8 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.966157|47898 / 11263756342 (0.000425%)|59114 / 873411 (6.768177%)|0.941996|0.938343|0.938358|
+|2|GERMAN CISTEM|PRIMARY_OUTPUT|0.915288|156784 / 11263756342 (0.001392%)|147964 / 873411 (16.940936%)|0.823934|0.826418|0.826415|
+|3|SNOWBALL GERMAN DIRECT|PRIMARY_OUTPUT|0.795926|87697 / 11263756342 (0.000779%)|356475 / 873411 (40.814118%)|0.785153|0.699487|0.711329|
+|4|SNOWBALL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.775641|77653 / 11263756342 (0.000689%)|391910 / 873411 (44.871200%)|0.774111|0.672222|0.688986|
+|5|GERMAN LUCENE GERMAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.769953|55477 / 11263756342 (0.000493%)|401846 / 873411 (46.008809%)|0.790797|0.673446|0.695023|
+|6|GERMAN LUCENE GERMAN STEM FILTER|PRIMARY_OUTPUT|0.716810|78723 / 11263756342 (0.000699%)|494677 / 873411 (56.637368%)|0.700519|0.569153|0.599149|
+|7|HUNSPELL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.659196|84679 / 11263756342 (0.000752%)|595318 / 873411 (68.160122%)|0.598178|0.449922|0.494019|
+|8|GERMAN LUCENE GERMAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.575691|21214 / 11263756342 (0.000188%)|741190 / 873411 (84.861537%)|0.444545|0.257528|0.361168|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.944446|0.932318|0.999996|0.966157|0.999991|0.000009|
+|2|GERMAN CISTEM|PRIMARY_OUTPUT|0.822287|0.830591|0.999986|0.915288|0.999973|0.000027|
+|3|SNOWBALL GERMAN DIRECT|PRIMARY_OUTPUT|0.854958|0.591859|0.999992|0.795926|0.999961|0.000039|
+|4|SNOWBALL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.861124|0.551288|0.999993|0.775641|0.999958|0.000042|
+|5|GERMAN LUCENE GERMAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.894739|0.539912|0.999995|0.769953|0.999959|0.000041|
+|6|GERMAN LUCENE GERMAN STEM FILTER|PRIMARY_OUTPUT|0.827912|0.433626|0.999993|0.716810|0.999949|0.000051|
+|7|HUNSPELL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.766578|0.318399|0.999992|0.659196|0.999940|0.000060|
+|8|GERMAN LUCENE GERMAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.861739|0.151385|0.999998|0.575691|0.999932|0.000068|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.941996|0.938343|0.934719|0.883848|0.938363|0.938358|
+|2|GERMAN CISTEM|PRIMARY_OUTPUT|0.823934|0.826418|0.828917|0.704184|0.826428|0.826415|
+|3|SNOWBALL GERMAN DIRECT|PRIMARY_OUTPUT|0.785153|0.699487|0.630675|0.537854|0.711347|0.711329|
+|4|SNOWBALL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.774111|0.672222|0.594035|0.506276|0.689005|0.688986|
+|5|GERMAN LUCENE GERMAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.790797|0.673446|0.586424|0.507666|0.695040|0.695023|
+|6|GERMAN LUCENE GERMAN STEM FILTER|PRIMARY_OUTPUT|0.700519|0.569153|0.479277|0.397774|0.599170|0.599149|
+|7|HUNSPELL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.598178|0.449922|0.360559|0.290258|0.494042|0.494019|
+|8|GERMAN LUCENE GERMAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.444545|0.257528|0.181270|0.147795|0.361184|0.361168|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.938338|0.994062|0.990664|0.992360|0.992360|
+|2|GERMAN CISTEM|PRIMARY_OUTPUT|0.826404|0.985936|0.973570|0.979714|0.979714|
+|3|SNOWBALL GERMAN DIRECT|PRIMARY_OUTPUT|0.699468|0.988418|0.932452|0.959619|0.959619|
+|4|SNOWBALL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.672202|0.989021|0.919542|0.953017|0.953017|
+|5|GERMAN LUCENE GERMAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.673427|0.991320|0.915070|0.951670|0.951670|
+|6|GERMAN LUCENE GERMAN STEM FILTER|PRIMARY_OUTPUT|0.569130|0.988584|0.918718|0.952371|0.952371|
+|7|HUNSPELL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|0.449897|0.988041|0.865581|0.922766|0.922766|
+|8|GERMAN LUCENE GERMAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.257511|0.992643|0.854403|0.918349|0.918349|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|814297|47898|59114|11263708444|47898 / 11263756342|59114 / 873411|
+|2|GERMAN CISTEM|PRIMARY_OUTPUT|725447|156784|147964|11263599558|156784 / 11263756342|147964 / 873411|
+|3|SNOWBALL GERMAN DIRECT|PRIMARY_OUTPUT|516936|87697|356475|11263668645|87697 / 11263756342|356475 / 873411|
+|4|SNOWBALL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|481501|77653|391910|11263678689|77653 / 11263756342|391910 / 873411|
+|5|GERMAN LUCENE GERMAN LIGHT STEM FILTER|PRIMARY_OUTPUT|471565|55477|401846|11263700865|55477 / 11263756342|401846 / 873411|
+|6|GERMAN LUCENE GERMAN STEM FILTER|PRIMARY_OUTPUT|378734|78723|494677|11263677619|78723 / 11263756342|494677 / 873411|
+|7|HUNSPELL GERMAN LUCENE FILTER|PRIMARY_OUTPUT|278093|84679|595318|11263671663|84679 / 11263756342|595318 / 873411|
+|8|GERMAN LUCENE GERMAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|132221|21214|741190|11263735128|21214 / 11263756342|741190 / 873411|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 11263756342 (0.000000%)|0 / 873411 (0.000000%)|1.000000|1.000000|1.000000|
+|2|HUNSPELL GERMAN LUCENE FILTER|ANY_CANDIDATE|0.665363|60996 / 11263756342 (0.000542%)|584547 / 873411 (66.926911%)|0.635466|0.472281|0.522540|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+|2|HUNSPELL GERMAN LUCENE FILTER|ANY_CANDIDATE|0.825656|0.330731|0.999995|0.665363|0.999943|0.000057|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+|2|HUNSPELL GERMAN LUCENE FILTER|ANY_CANDIDATE|0.635466|0.472281|0.375782|0.309142|0.522561|0.522540|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL GERMAN LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|873411|0|0|11263756342|0 / 11263756342|0 / 873411|
+|2|HUNSPELL GERMAN LUCENE FILTER|ANY_CANDIDATE|288864|60996|584547|11263695346|60996 / 11263756342|584547 / 873411|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999996|97544 / 11263756342 (0.000866%)|0 / 873411 (0.000000%)|0.917983|0.947112|0.948436|
+|2|HUNSPELL GERMAN LUCENE FILTER|ALL_CANDIDATES|0.665361|96545 / 11263756342 (0.000857%)|584547 / 873411 (66.926911%)|0.598050|0.458944|0.497855|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.899538|1.000000|0.999991|0.999996|0.999991|0.000009|
+|2|HUNSPELL GERMAN LUCENE FILTER|ALL_CANDIDATES|0.749500|0.330731|0.999991|0.665361|0.999940|0.000060|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.917983|0.947112|0.978152|0.899538|0.948440|0.948436|
+|2|HUNSPELL GERMAN LUCENE FILTER|ALL_CANDIDATES|0.598050|0.458944|0.372338|0.297811|0.497878|0.497855|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL GERMAN LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|873411|97544|0|11263658798|97544 / 11263756342|0 / 873411|
+|2|HUNSPELL GERMAN LUCENE FILTER|ALL_CANDIDATES|288864|96545|584547|11263659797|96545 / 11263756342|584547 / 873411|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|59114|47898|49646|14978|9.978814%|8|167157|
+|HUNSPELL GERMAN LUCENE FILTER|10771|23683|11866|4989|3.323828%|3|155207|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `DE_DE`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/hungarian.md
+++ b/docs/benchmarks/languages/hungarian.md
@@ -0,0 +1,408 @@
+# Hungarian Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Hungarian. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `HU_HU` | 19,406 | 935,713 | 38,775 | 896,938 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **935,713**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 15 | 0.002% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 149,173 | 15.942% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 746,296 | 79.757% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 40,125 | 4.288% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 104 | 0.011% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 99.222% | 99.537% | 91.948% | Full Radixor dictionary patch-command stemmer. |
+| Lucene SnowballFilter | 66.445% | 66.938% | 55.043% | Lucene TokenFilter integration path around the Snowball algorithm. |
+| Official Snowball direct | 66.445% | 66.938% | 55.043% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+| Lucene HungarianLightStemFilter | 14.748% | 14.777% | 14.086% | Light suffix stemmer; intentionally narrower than a dictionary-derived stemmer. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `hungarianRadixor` | 62.232 | 6.412 | 69.4 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene HungarianLightStemFilter | `hungarianLuceneHungarianLightStemFilter` | 92.813 | 6.929 | 103.5 | 1.491 | Light Hungarian suffix stemmer. |
+| Official Snowball direct | `snowballDirect[HUNGARIAN]` | 157.765 | 13.202 | 175.9 | 2.535 | Official Snowball generated Java stemmer; direct API. |
+| Lucene SnowballFilter | `luceneSnowballFilter[HUNGARIAN]` | 188.863 | 15.880 | 210.6 | 3.035 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `HU_HU` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/hu_hu/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.995491** among 4 deterministic stemmers. The runner-up is `SNOWBALL HUNGARIAN LUCENE FILTER` at 0.822606, a difference of 0.172885. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.996163** among 4 deterministic stemmers. The runner-up is `SNOWBALL HUNGARIAN DIRECT` at 0.821708, a difference of 0.174455. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **6 result rows**, **4 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.995491|272900 / 419820542893 (0.000065%)|199837 / 22162103 (0.901706%)|0.988376|0.989352|0.989353|
+|2|SNOWBALL HUNGARIAN LUCENE FILTER|PRIMARY_OUTPUT|0.822606|1792049 / 419820542893 (0.000427%)|7862745 / 22162103 (35.478334%)|0.826288|0.747610|0.757196|
+|3|SNOWBALL HUNGARIAN DIRECT|PRIMARY_OUTPUT|0.822348|1506056 / 419820542893 (0.000359%)|7874191 / 22162103 (35.529981%)|0.837137|0.752866|0.763681|
+|4|HUNGARIAN LUCENE HUNGARIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.816668|4132555 / 419820542893 (0.000984%)|8125833 / 22162103 (36.665442%)|0.740018|0.696055|0.699478|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.987727|0.990983|0.999999|0.995491|0.999999|0.000001|
+|2|SNOWBALL HUNGARIAN LUCENE FILTER|PRIMARY_OUTPUT|0.888633|0.645217|0.999996|0.822606|0.999977|0.000023|
+|3|SNOWBALL HUNGARIAN DIRECT|PRIMARY_OUTPUT|0.904644|0.644700|0.999996|0.822348|0.999978|0.000022|
+|4|HUNGARIAN LUCENE HUNGARIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.772547|0.633346|0.999990|0.816668|0.999971|0.000029|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.988376|0.989352|0.990330|0.978929|0.989353|0.989353|
+|2|SNOWBALL HUNGARIAN LUCENE FILTER|PRIMARY_OUTPUT|0.826288|0.747610|0.682613|0.596947|0.757206|0.757196|
+|3|SNOWBALL HUNGARIAN DIRECT|PRIMARY_OUTPUT|0.837137|0.752866|0.684009|0.603677|0.763691|0.763681|
+|4|HUNGARIAN LUCENE HUNGARIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.740018|0.696055|0.657023|0.533807|0.699492|0.699478|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.989352|0.998036|0.997809|0.997922|0.997922|
+|2|SNOWBALL HUNGARIAN LUCENE FILTER|PRIMARY_OUTPUT|0.747599|0.990687|0.924490|0.956445|0.956445|
+|3|SNOWBALL HUNGARIAN DIRECT|PRIMARY_OUTPUT|0.752855|0.991948|0.924304|0.956932|0.956932|
+|4|HUNGARIAN LUCENE HUNGARIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.696040|0.982615|0.926772|0.953877|0.953877|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|21962266|272900|199837|419820269993|272900 / 419820542893|199837 / 22162103|
+|2|SNOWBALL HUNGARIAN LUCENE FILTER|PRIMARY_OUTPUT|14299358|1792049|7862745|419818750844|1792049 / 419820542893|7862745 / 22162103|
+|3|SNOWBALL HUNGARIAN DIRECT|PRIMARY_OUTPUT|14287912|1506056|7874191|419819036837|1506056 / 419820542893|7874191 / 22162103|
+|4|HUNGARIAN LUCENE HUNGARIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|14036270|4132555|8125833|419816410338|4132555 / 419820542893|8125833 / 22162103|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 419820542893 (0.000000%)|0 / 22162103 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|22162103|0|0|419820542893|0 / 419820542893|0 / 22162103|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999999|460158 / 419820542893 (0.000110%)|0 / 22162103 (0.000000%)|0.983661|0.989725|0.989777|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.979659|1.000000|0.999999|0.999999|0.999999|0.000001|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.983661|0.989725|0.995865|0.979659|0.989777|0.989777|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|22162103|460158|0|419820082735|460158 / 419820542893|0 / 22162103|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|199837|272900|187258|12320|1.344473%|5|929326|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **6 result rows**, **4 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.996163|272775 / 385870694917 (0.000071%)|164277 / 21411411 (0.767240%)|0.988321|0.989820|0.989822|
+|2|SNOWBALL HUNGARIAN DIRECT|PRIMARY_OUTPUT|0.821708|1496670 / 385870694917 (0.000388%)|7634885 / 21411411 (35.658019%)|0.834899|0.751079|0.761809|
+|3|SNOWBALL HUNGARIAN LUCENE FILTER|PRIMARY_OUTPUT|0.821708|1496670 / 385870694917 (0.000388%)|7634885 / 21411411 (35.658019%)|0.834899|0.751079|0.761809|
+|4|HUNGARIAN LUCENE HUNGARIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.815077|3639046 / 385870694917 (0.000943%)|7918708 / 21411411 (36.983588%)|0.750108|0.700135|0.704477|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.987325|0.992328|0.999999|0.996163|0.999999|0.000001|
+|2|SNOWBALL HUNGARIAN DIRECT|PRIMARY_OUTPUT|0.902007|0.643420|0.999996|0.821708|0.999976|0.000024|
+|3|SNOWBALL HUNGARIAN LUCENE FILTER|PRIMARY_OUTPUT|0.902007|0.643420|0.999996|0.821708|0.999976|0.000024|
+|4|HUNGARIAN LUCENE HUNGARIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.787585|0.630164|0.999991|0.815077|0.999970|0.000030|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.988321|0.989820|0.991323|0.979845|0.989823|0.989822|
+|2|SNOWBALL HUNGARIAN DIRECT|PRIMARY_OUTPUT|0.834899|0.751079|0.682555|0.601383|0.761820|0.761809|
+|3|SNOWBALL HUNGARIAN LUCENE FILTER|PRIMARY_OUTPUT|0.834899|0.751079|0.682555|0.601383|0.761820|0.761809|
+|4|HUNGARIAN LUCENE HUNGARIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.750108|0.700135|0.656404|0.538621|0.704491|0.704477|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.989819|0.997945|0.998273|0.998109|0.998109|
+|2|SNOWBALL HUNGARIAN DIRECT|PRIMARY_OUTPUT|0.751068|0.991610|0.923288|0.956230|0.956230|
+|3|SNOWBALL HUNGARIAN LUCENE FILTER|PRIMARY_OUTPUT|0.751068|0.991610|0.923288|0.956230|0.956230|
+|4|HUNGARIAN LUCENE HUNGARIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.700120|0.983687|0.925487|0.953700|0.953700|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|21247134|272775|164277|385870422142|272775 / 385870694917|164277 / 21411411|
+|2|SNOWBALL HUNGARIAN DIRECT|PRIMARY_OUTPUT|13776526|1496670|7634885|385869198247|1496670 / 385870694917|7634885 / 21411411|
+|3|SNOWBALL HUNGARIAN LUCENE FILTER|PRIMARY_OUTPUT|13776526|1496670|7634885|385869198247|1496670 / 385870694917|7634885 / 21411411|
+|4|HUNGARIAN LUCENE HUNGARIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|13492703|3639046|7918708|385867055871|3639046 / 385870694917|7918708 / 21411411|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 385870694917 (0.000000%)|0 / 21411411 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|21411411|0|0|385870694917|0 / 385870694917|0 / 21411411|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999999|458462 / 385870694917 (0.000119%)|0 / 21411411 (0.000000%)|0.983159|0.989407|0.989462|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.979037|1.000000|0.999999|0.999999|0.999999|0.000001|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.983159|0.989407|0.995736|0.979037|0.989463|0.989462|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|21411411|458462|0|385870236455|458462 / 385870694917|0 / 21411411|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|164277|272775|185687|11153|1.269532%|5|890245|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `HU_HU`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/index.md
+++ b/docs/benchmarks/languages/index.md
@@ -0,0 +1,45 @@
+# Language Benchmark Pages
+
+This section splits Radixor stemmer benchmark results by language. Each language page preserves the existing exact-root accuracy and runtime-performance results and adds pairwise stemming-quality tables for both dictionary-processing modes.
+
+## Reference Pages
+
+| Page | Purpose |
+| --- | --- |
+| [Methodology](../reference/methodology.md) | Workload design, normalization, speed metrics, and exact-root quality metrics. Pairwise quality definitions are also reproduced on every language page. |
+| [Corpora](../reference/corpora.md) | Dictionary sizes and changed-token timing workloads. |
+| [Environment and reports](../reference/environment.md) | Hardware, JVM, JMH settings, report files, and badge policy. |
+| [English dictionary coverage](../reference/english-coverage.md) | Quality/speed operating curve for contracted Radixor tries built from 100% down to 10% of English dictionary rows. |
+| [Candidate evaluation](../reference/candidates.md) | Included and skipped stemmer candidates. |
+
+## Languages
+
+| Language | Resource | Benchmark page |
+| --- | --- | --- |
+| Czech | `CS_CZ` | [Czech](czech.md) |
+| Danish | `DA_DK` | [Danish](danish.md) |
+| Dutch | `NL_NL` | [Dutch](dutch.md) |
+| English | `US_UK` | [English](english.md) |
+| Finnish | `FI_FI` | [Finnish](finnish.md) |
+| French | `FR_FR` | [French](french.md) |
+| German | `DE_DE` | [German](german.md) |
+| Hungarian | `HU_HU` | [Hungarian](hungarian.md) |
+| Italian | `IT_IT` | [Italian](italian.md) |
+| Norwegian Bokmal | `NB_NO` | [Norwegian Bokmal](norwegian-bokmal.md) |
+| Norwegian Nynorsk | `NN_NO` | [Norwegian Nynorsk](norwegian-nynorsk.md) |
+| Persian | `FA_IR` | [Persian](persian.md) |
+| Polish | `PL_PL` | [Polish](polish.md) |
+| Portuguese | `PT_PT` | [Portuguese](portuguese.md) |
+| Russian | `RU_RU` | [Russian](russian.md) |
+| Spanish | `ES_ES` | [Spanish](spanish.md) |
+| Swedish | `SV_SE` | [Swedish](swedish.md) |
+| Ukrainian | `UK_UA` | [Ukrainian](ukrainian.md) |
+| Yiddish | `YI` | [Yiddish](yiddish.md) |
+
+## Methodology Notes
+
+- Speed benchmarks process only changed dictionary tokens where the surface form differs from the expected root.
+- Accuracy benchmarks process the complete dictionary and report `All exact`, `Changed exact`, and `Root preserved`.
+- Radixor speed must be interpreted together with exact-root quality. A slower Radixor row must not be read as a simple performance weakness when Radixor is also the row with accuracy close to 100% and competing stemmers are much lower. Many fast light, minimal, possessive, or aggressive rule-based stemmers are fast because they do much less linguistic work. The measured Radixor cost buys dictionary-trained precision, and that precision is what improves search quality when queries and indexed text are reduced to the same intended roots. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows this contracted-trie operating curve explicitly.
+- Results are comparable only within the same language and benchmark family.
+- The historical Porter badge is retired; no JMH badge JSON is generated.
--- a/docs/benchmarks/languages/italian.md
+++ b/docs/benchmarks/languages/italian.md
@@ -0,0 +1,407 @@
+# Italian Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Italian. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `IT_IT` | 10,009 | 337,546 | 20,004 | 317,542 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **337,546**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 302,171 | 89.520% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 12,348 | 3.658% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 20,013 | 5.929% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 3,014 | 0.893% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 99.056% | 98.997% | 100.000% | Full Radixor dictionary patch-command stemmer. |
+| Lucene ItalianLightStemFilter | 0.466% | 0.479% | 0.270% | Light suffix stemmer; intentionally narrower than a dictionary-derived stemmer. |
+| Lucene SnowballFilter | 0.041% | 0.043% | 0.010% | Lucene TokenFilter integration path around the Snowball algorithm. |
+| Official Snowball direct | 0.041% | 0.043% | 0.010% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `italianRadixor` | 24.491 | 3.128 | 77.1 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene ItalianLightStemFilter | `italianLuceneItalianLightStemFilter` | 15.977 | 1.041 | 50.3 | 0.652 | Light Italian suffix stemmer. |
+| Official Snowball direct | `snowballDirect[ITALIAN]` | 109.526 | 12.572 | 344.9 | 4.472 | Official Snowball generated Java stemmer; direct API. |
+| Lucene SnowballFilter | `luceneSnowballFilter[ITALIAN]` | 116.260 | 7.459 | 366.1 | 4.747 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `IT_IT` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/it_it/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.996507** among 4 deterministic stemmers. The runner-up is `SNOWBALL ITALIAN DIRECT` at 0.866189, a difference of 0.130318. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.996512** among 4 deterministic stemmers. The runner-up is `SNOWBALL ITALIAN DIRECT` at 0.866205, a difference of 0.130307. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **6 result rows**, **4 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.996507|124172 / 53638521211 (0.000231%)|42908 / 6143814 (0.698394%)|0.982618|0.986492|0.986512|
+|2|SNOWBALL ITALIAN DIRECT|PRIMARY_OUTPUT|0.866189|504775 / 53638521211 (0.000941%)|1644164 / 6143814 (26.761292%)|0.859975|0.807240|0.811470|
+|3|SNOWBALL ITALIAN LUCENE FILTER|PRIMARY_OUTPUT|0.866189|504775 / 53638521211 (0.000941%)|1644164 / 6143814 (26.761292%)|0.859975|0.807240|0.811470|
+|4|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.508926|10589 / 53638521211 (0.000020%)|6034130 / 6143814 (98.214725%)|0.082782|0.035020|0.127588|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.980053|0.993016|0.999998|0.996507|0.999997|0.000003|
+|2|SNOWBALL ITALIAN DIRECT|PRIMARY_OUTPUT|0.899134|0.732387|0.999991|0.866189|0.999960|0.000040|
+|3|SNOWBALL ITALIAN LUCENE FILTER|PRIMARY_OUTPUT|0.899134|0.732387|0.999991|0.866189|0.999960|0.000040|
+|4|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.911959|0.017853|1.000000|0.508926|0.999887|0.000113|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.982618|0.986492|0.990396|0.973344|0.986513|0.986512|
+|2|SNOWBALL ITALIAN DIRECT|PRIMARY_OUTPUT|0.859975|0.807240|0.760598|0.676783|0.811489|0.811470|
+|3|SNOWBALL ITALIAN LUCENE FILTER|PRIMARY_OUTPUT|0.859975|0.807240|0.760598|0.676783|0.811489|0.811470|
+|4|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.082782|0.035020|0.022207|0.017822|0.127597|0.127588|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.986490|0.995780|0.997113|0.996446|0.996446|
+|2|SNOWBALL ITALIAN DIRECT|PRIMARY_OUTPUT|0.807220|0.987994|0.933408|0.959925|0.959925|
+|3|SNOWBALL ITALIAN LUCENE FILTER|PRIMARY_OUTPUT|0.807220|0.987994|0.933408|0.959925|0.959925|
+|4|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.035016|0.997481|0.737537|0.848037|0.848037|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|6100906|124172|42908|53638397039|124172 / 53638521211|42908 / 6143814|
+|2|SNOWBALL ITALIAN DIRECT|PRIMARY_OUTPUT|4499650|504775|1644164|53638016436|504775 / 53638521211|1644164 / 6143814|
+|3|SNOWBALL ITALIAN LUCENE FILTER|PRIMARY_OUTPUT|4499650|504775|1644164|53638016436|504775 / 53638521211|1644164 / 6143814|
+|4|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|109684|10589|6034130|53638510622|10589 / 53638521211|6034130 / 6143814|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|0.999993|0 / 53638521211 (0.000000%)|80 / 6143814 (0.001302%)|0.999997|0.999993|0.999993|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0.999987|1.000000|0.999993|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|0.999997|0.999993|0.999990|0.999987|0.999993|0.999993|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|6143734|0|80|53638521211|0 / 53638521211|80 / 6143814|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999992|170950 / 53638521211 (0.000319%)|80 / 6143814 (0.001302%)|0.978222|0.986272|0.986363|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.972928|0.999987|0.999997|0.999992|0.999997|0.000003|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.978222|0.986272|0.994455|0.972916|0.986365|0.986363|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|6143734|170950|80|53638350261|170950 / 53638521211|80 / 6143814|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|42828|124172|46778|6254|1.909321%|4|334175|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **6 result rows**, **4 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.996512|124171 / 53611667072 (0.000232%)|42828 / 6142174 (0.697278%)|0.982617|0.986495|0.986515|
+|2|SNOWBALL ITALIAN DIRECT|PRIMARY_OUTPUT|0.866205|504774 / 53611667072 (0.000942%)|1643522 / 6142174 (26.757985%)|0.859970|0.807252|0.811479|
+|3|SNOWBALL ITALIAN LUCENE FILTER|PRIMARY_OUTPUT|0.866205|504774 / 53611667072 (0.000942%)|1643522 / 6142174 (26.757985%)|0.859970|0.807252|0.811479|
+|4|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.508927|10588 / 53611667072 (0.000020%)|6032516 / 6142174 (98.214671%)|0.082784|0.035021|0.127589|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.980048|0.993027|0.999998|0.996512|0.999997|0.000003|
+|2|SNOWBALL ITALIAN DIRECT|PRIMARY_OUTPUT|0.899114|0.732420|0.999991|0.866205|0.999960|0.000040|
+|3|SNOWBALL ITALIAN LUCENE FILTER|PRIMARY_OUTPUT|0.899114|0.732420|0.999991|0.866205|0.999960|0.000040|
+|4|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.911947|0.017853|1.000000|0.508927|0.999887|0.000113|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.982617|0.986495|0.990404|0.973350|0.986516|0.986515|
+|2|SNOWBALL ITALIAN DIRECT|PRIMARY_OUTPUT|0.859970|0.807252|0.760624|0.676800|0.811498|0.811479|
+|3|SNOWBALL ITALIAN LUCENE FILTER|PRIMARY_OUTPUT|0.859970|0.807252|0.760624|0.676800|0.811498|0.811479|
+|4|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.082784|0.035021|0.022208|0.017823|0.127598|0.127589|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.986493|0.995780|0.997115|0.996447|0.996447|
+|2|SNOWBALL ITALIAN DIRECT|PRIMARY_OUTPUT|0.807232|0.987991|0.933413|0.959927|0.959927|
+|3|SNOWBALL ITALIAN LUCENE FILTER|PRIMARY_OUTPUT|0.807232|0.987991|0.933413|0.959927|0.959927|
+|4|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.035017|0.997481|0.737534|0.848035|0.848035|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|6099346|124171|42828|53611542901|124171 / 53611667072|42828 / 6142174|
+|2|SNOWBALL ITALIAN DIRECT|PRIMARY_OUTPUT|4498652|504774|1643522|53611162298|504774 / 53611667072|1643522 / 6142174|
+|3|SNOWBALL ITALIAN LUCENE FILTER|PRIMARY_OUTPUT|4498652|504774|1643522|53611162298|504774 / 53611667072|1643522 / 6142174|
+|4|ITALIAN LUCENE ITALIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|109658|10588|6032516|53611656484|10588 / 53611667072|6032516 / 6142174|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 53611667072 (0.000000%)|0 / 6142174 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|6142174|0|0|53611667072|0 / 53611667072|0 / 6142174|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999998|170949 / 53611667072 (0.000319%)|0 / 6142174 (0.000000%)|0.978219|0.986275|0.986366|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.972922|1.000000|0.999997|0.999998|0.999997|0.000003|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.978219|0.986275|0.994464|0.972922|0.986368|0.986366|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|6142174|170949|0|53611496123|170949 / 53611667072|0 / 6142174|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|42828|124171|46778|6252|1.909188%|4|334089|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `IT_IT`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/norwegian-bokmal.md
+++ b/docs/benchmarks/languages/norwegian-bokmal.md
@@ -0,0 +1,420 @@
+# Norwegian Bokmal Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Norwegian Bokmal. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `NB_NO` | 17,929 | 90,757 | 33,376 | 57,381 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **90,757**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 1,500 | 1.653% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 4,296 | 4.734% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 47,619 | 52.469% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 34,420 | 37.925% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 2,922 | 3.220% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 96.852% | 97.637% | 95.503% | Full Radixor dictionary patch-command stemmer. |
+| Lucene NorwegianMinimalStemFilter | 57.107% | 53.913% | 62.599% | Minimal suffix reducer; narrow baseline, not a full stemmer. |
+| Official Snowball direct | 54.824% | 51.791% | 60.040% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+| Lucene SnowballFilter | 54.803% | 51.780% | 60.001% | Lucene TokenFilter integration path around the Snowball algorithm. |
+| Lucene NorwegianLightStemFilter | 52.136% | 50.616% | 54.749% | Light suffix stemmer; intentionally narrower than a dictionary-derived stemmer. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `norwegianBokmalRadixor` | 3.631 | 1.377 | 63.3 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene NorwegianMinimalStemFilter | `norwegianBokmalLuceneNorwegianMinimalStemFilter` | 2.910 | 0.177 | 50.7 | 0.801 | Minimal Norwegian suffix reducer. |
+| Lucene NorwegianLightStemFilter | `norwegianBokmalLuceneNorwegianLightStemFilter` | 3.335 | 0.116 | 58.1 | 0.919 | Light Norwegian suffix stemmer. |
+| Official Snowball direct | `snowballDirect[NORWEGIAN_BOKMAL]` | 4.277 | 0.082 | 74.5 | 1.178 | Official Snowball generated Java stemmer; direct API. |
+| Lucene SnowballFilter | `luceneSnowballFilter[NORWEGIAN_BOKMAL]` | 6.077 | 0.208 | 105.9 | 1.674 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `NB_NO` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/nb_no/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.974783** among 5 deterministic stemmers. The runner-up is `SNOWBALL NORWEGIAN BOKMAL DIRECT` at 0.874964, a difference of 0.099819. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.975000** among 5 deterministic stemmers. The runner-up is `SNOWBALL NORWEGIAN BOKMAL DIRECT` at 0.874991, a difference of 0.100009. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **7 result rows**, **5 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.974783|11482 / 2835618215 (0.000405%)|7170 / 142180 (5.042903%)|0.927078|0.935387|0.935488|
+|2|SNOWBALL NORWEGIAN BOKMAL DIRECT|PRIMARY_OUTPUT|0.874964|23997 / 2835618215 (0.000846%)|35554 / 142180 (25.006330%)|0.802095|0.781707|0.782399|
+|3|SNOWBALL NORWEGIAN BOKMAL LUCENE FILTER|PRIMARY_OUTPUT|0.874834|24046 / 2835618215 (0.000848%)|35591 / 142180 (25.032353%)|0.801759|0.781401|0.782091|
+|4|NORWEGIAN BOKMAL LUCENE NORWEGIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.850006|25171 / 2835618215 (0.000888%)|42651 / 142180 (29.997890%)|0.776381|0.745871|0.747464|
+|5|NORWEGIAN BOKMAL LUCENE NORWEGIAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.832414|14772 / 2835618215 (0.000521%)|47654 / 142180 (33.516669%)|0.815763|0.751764|0.758263|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.921620|0.949571|0.999996|0.974783|0.999993|0.000007|
+|2|SNOWBALL NORWEGIAN BOKMAL DIRECT|PRIMARY_OUTPUT|0.816288|0.749937|0.999992|0.874964|0.999979|0.000021|
+|3|SNOWBALL NORWEGIAN BOKMAL LUCENE FILTER|PRIMARY_OUTPUT|0.815930|0.749676|0.999992|0.874834|0.999979|0.000021|
+|4|NORWEGIAN BOKMAL LUCENE NORWEGIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.798148|0.700021|0.999991|0.850006|0.999976|0.000024|
+|5|NORWEGIAN BOKMAL LUCENE NORWEGIAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.864847|0.664833|0.999995|0.832414|0.999978|0.000022|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.927078|0.935387|0.943846|0.878617|0.935491|0.935488|
+|2|SNOWBALL NORWEGIAN BOKMAL DIRECT|PRIMARY_OUTPUT|0.802095|0.781707|0.762330|0.641641|0.782409|0.782399|
+|3|SNOWBALL NORWEGIAN BOKMAL LUCENE FILTER|PRIMARY_OUTPUT|0.801759|0.781401|0.762052|0.641229|0.782102|0.782091|
+|4|NORWEGIAN BOKMAL LUCENE NORWEGIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.776381|0.745871|0.717668|0.594732|0.747476|0.747464|
+|5|NORWEGIAN BOKMAL LUCENE NORWEGIAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.815763|0.751764|0.697076|0.602261|0.758274|0.758263|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.935384|0.993354|0.994615|0.993984|0.993984|
+|2|SNOWBALL NORWEGIAN BOKMAL DIRECT|PRIMARY_OUTPUT|0.781696|0.988328|0.971120|0.979648|0.979648|
+|3|SNOWBALL NORWEGIAN BOKMAL LUCENE FILTER|PRIMARY_OUTPUT|0.781391|0.988295|0.971086|0.979615|0.979615|
+|4|NORWEGIAN BOKMAL LUCENE NORWEGIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.745859|0.987774|0.965622|0.976573|0.976573|
+|5|NORWEGIAN BOKMAL LUCENE NORWEGIAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.751753|0.992089|0.962516|0.977079|0.977079|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|135010|11482|7170|2835606733|11482 / 2835618215|7170 / 142180|
+|2|SNOWBALL NORWEGIAN BOKMAL DIRECT|PRIMARY_OUTPUT|106626|23997|35554|2835594218|23997 / 2835618215|35554 / 142180|
+|3|SNOWBALL NORWEGIAN BOKMAL LUCENE FILTER|PRIMARY_OUTPUT|106589|24046|35591|2835594169|24046 / 2835618215|35591 / 142180|
+|4|NORWEGIAN BOKMAL LUCENE NORWEGIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|99529|25171|42651|2835593044|25171 / 2835618215|42651 / 142180|
+|5|NORWEGIAN BOKMAL LUCENE NORWEGIAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|94526|14772|47654|2835603443|14772 / 2835618215|47654 / 142180|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 2835618215 (0.000000%)|0 / 142180 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|142180|0|0|2835618215|0 / 2835618215|0 / 142180|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999996|20161 / 2835618215 (0.000711%)|0 / 142180 (0.000000%)|0.898118|0.933794|0.935844|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.875811|1.000000|0.999993|0.999996|0.999993|0.000007|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.898118|0.933794|0.972422|0.875811|0.935848|0.935844|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|142180|20161|0|2835598054|20161 / 2835618215|0 / 142180|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|7170|11482|8679|4237|5.626079%|9|79825|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **7 result rows**, **5 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.975000|11482 / 2831176784 (0.000406%)|7104 / 142091 (4.999613%)|0.927151|0.935591|0.935695|
+|2|SNOWBALL NORWEGIAN BOKMAL DIRECT|PRIMARY_OUTPUT|0.874991|23997 / 2831176784 (0.000848%)|35524 / 142091 (25.000880%)|0.802043|0.781698|0.782388|
+|3|SNOWBALL NORWEGIAN BOKMAL LUCENE FILTER|PRIMARY_OUTPUT|0.874798|23993 / 2831176784 (0.000847%)|35579 / 142091 (25.039587%)|0.801914|0.781464|0.782161|
+|4|NORWEGIAN BOKMAL LUCENE NORWEGIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.849947|25118 / 2831176784 (0.000887%)|42641 / 142091 (30.009642%)|0.776513|0.745896|0.747500|
+|5|NORWEGIAN BOKMAL LUCENE NORWEGIAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.832344|14719 / 2831176784 (0.000520%)|47644 / 142091 (33.530625%)|0.815950|0.751796|0.758325|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.921608|0.950004|0.999996|0.975000|0.999993|0.000007|
+|2|SNOWBALL NORWEGIAN BOKMAL DIRECT|PRIMARY_OUTPUT|0.816205|0.749991|0.999992|0.874991|0.999979|0.000021|
+|3|SNOWBALL NORWEGIAN BOKMAL LUCENE FILTER|PRIMARY_OUTPUT|0.816153|0.749604|0.999992|0.874798|0.999979|0.000021|
+|4|NORWEGIAN BOKMAL LUCENE NORWEGIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.798359|0.699904|0.999991|0.849947|0.999976|0.000024|
+|5|NORWEGIAN BOKMAL LUCENE NORWEGIAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.865169|0.664694|0.999995|0.832344|0.999978|0.000022|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.927151|0.935591|0.944186|0.878976|0.935698|0.935695|
+|2|SNOWBALL NORWEGIAN BOKMAL DIRECT|PRIMARY_OUTPUT|0.802043|0.781698|0.762360|0.641630|0.782398|0.782388|
+|3|SNOWBALL NORWEGIAN BOKMAL LUCENE FILTER|PRIMARY_OUTPUT|0.801914|0.781464|0.762031|0.641314|0.782171|0.782161|
+|4|NORWEGIAN BOKMAL LUCENE NORWEGIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.776513|0.745896|0.717603|0.594765|0.747512|0.747500|
+|5|NORWEGIAN BOKMAL LUCENE NORWEGIAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.815950|0.751796|0.696995|0.602302|0.758335|0.758325|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.935587|0.993348|0.994694|0.994020|0.994020|
+|2|SNOWBALL NORWEGIAN BOKMAL DIRECT|PRIMARY_OUTPUT|0.781688|0.988318|0.971127|0.979647|0.979647|
+|3|SNOWBALL NORWEGIAN BOKMAL LUCENE FILTER|PRIMARY_OUTPUT|0.781454|0.988310|0.971074|0.979616|0.979616|
+|4|NORWEGIAN BOKMAL LUCENE NORWEGIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.745885|0.987789|0.965603|0.976570|0.976570|
+|5|NORWEGIAN BOKMAL LUCENE NORWEGIAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.751785|0.992107|0.962494|0.977076|0.977076|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|134987|11482|7104|2831165302|11482 / 2831176784|7104 / 142091|
+|2|SNOWBALL NORWEGIAN BOKMAL DIRECT|PRIMARY_OUTPUT|106567|23997|35524|2831152787|23997 / 2831176784|35524 / 142091|
+|3|SNOWBALL NORWEGIAN BOKMAL LUCENE FILTER|PRIMARY_OUTPUT|106512|23993|35579|2831152791|23993 / 2831176784|35579 / 142091|
+|4|NORWEGIAN BOKMAL LUCENE NORWEGIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|99450|25118|42641|2831151666|25118 / 2831176784|42641 / 142091|
+|5|NORWEGIAN BOKMAL LUCENE NORWEGIAN MINIMAL STEM FILTER|PRIMARY_OUTPUT|94447|14719|47644|2831162065|14719 / 2831176784|47644 / 142091|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 2831176784 (0.000000%)|0 / 142091 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|142091|0|0|2831176784|0 / 2831176784|0 / 142091|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999996|20161 / 2831176784 (0.000712%)|0 / 142091 (0.000000%)|0.898061|0.933756|0.935808|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.875743|1.000000|0.999993|0.999996|0.999993|0.000007|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.898061|0.933756|0.972405|0.875743|0.935811|0.935808|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|142091|20161|0|2831156623|20161 / 2831176784|0 / 142091|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|7104|11482|8679|4204|5.586637%|9|79733|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `NB_NO`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/norwegian-nynorsk.md
+++ b/docs/benchmarks/languages/norwegian-nynorsk.md
@@ -0,0 +1,396 @@
+# Norwegian Nynorsk Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Norwegian Nynorsk. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `NN_NO` | 4,688 | 19,651 | 6,089 | 13,562 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **19,651**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 224 | 1.140% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 1,505 | 7.659% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 11,017 | 56.063% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 6,427 | 32.706% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 478 | 2.432% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 93.089% | 91.395% | 96.863% | Full Radixor dictionary patch-command stemmer. |
+| Official Snowball direct | 60.974% | 60.212% | 62.670% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+| Lucene SnowballFilter | 60.918% | 60.146% | 62.638% | Lucene TokenFilter integration path around the Snowball algorithm. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `radixor[NORWEGIAN_NYNORSK]` | 0.571 | 0.012 | 42.1 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Official Snowball direct | `snowballDirect[NORWEGIAN_NYNORSK]` | 0.919 | 0.038 | 67.7 | 1.609 | Official Snowball generated Java stemmer; direct API. |
+| Lucene SnowballFilter | `luceneSnowballFilter[NORWEGIAN_NYNORSK]` | 1.309 | 0.015 | 96.5 | 2.292 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `NN_NO` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/nn_no/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.935777** among 3 deterministic stemmers. The runner-up is `SNOWBALL NORWEGIAN NYNORSK DIRECT` at 0.858908, a difference of 0.076869. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.935853** among 3 deterministic stemmers. The runner-up is `SNOWBALL NORWEGIAN NYNORSK DIRECT` at 0.859037, a difference of 0.076816. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **5 result rows**, **3 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.935777|6230 / 166491473 (0.003742%)|3936 / 30652 (12.840924%)|0.822355|0.840152|0.840669|
+|2|SNOWBALL NORWEGIAN NYNORSK DIRECT|PRIMARY_OUTPUT|0.858908|8274 / 166491473 (0.004970%)|8648 / 30652 (28.213493%)|0.724941|0.722271|0.722234|
+|3|SNOWBALL NORWEGIAN NYNORSK LUCENE FILTER|PRIMARY_OUTPUT|0.858484|8295 / 166491473 (0.004982%)|8674 / 30652 (28.298317%)|0.724180|0.721477|0.721440|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.810903|0.871591|0.999963|0.935777|0.999939|0.000061|
+|2|SNOWBALL NORWEGIAN NYNORSK DIRECT|PRIMARY_OUTPUT|0.726732|0.717865|0.999950|0.858908|0.999898|0.000102|
+|3|SNOWBALL NORWEGIAN NYNORSK LUCENE FILTER|PRIMARY_OUTPUT|0.725993|0.717017|0.999950|0.858484|0.999898|0.000102|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.822355|0.840152|0.858737|0.724364|0.840699|0.840669|
+|2|SNOWBALL NORWEGIAN NYNORSK DIRECT|PRIMARY_OUTPUT|0.724941|0.722271|0.719621|0.565278|0.722285|0.722234|
+|3|SNOWBALL NORWEGIAN NYNORSK LUCENE FILTER|PRIMARY_OUTPUT|0.724180|0.721477|0.718794|0.564305|0.721491|0.721440|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.840122|0.983845|0.986802|0.985321|0.985321|
+|2|SNOWBALL NORWEGIAN NYNORSK DIRECT|PRIMARY_OUTPUT|0.722221|0.980542|0.964998|0.972708|0.972708|
+|3|SNOWBALL NORWEGIAN NYNORSK LUCENE FILTER|PRIMARY_OUTPUT|0.721426|0.980461|0.964862|0.972599|0.972599|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|26716|6230|3936|166485243|6230 / 166491473|3936 / 30652|
+|2|SNOWBALL NORWEGIAN NYNORSK DIRECT|PRIMARY_OUTPUT|22004|8274|8648|166483199|8274 / 166491473|8648 / 30652|
+|3|SNOWBALL NORWEGIAN NYNORSK LUCENE FILTER|PRIMARY_OUTPUT|21978|8295|8674|166483178|8295 / 166491473|8674 / 30652|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 166491473 (0.000000%)|0 / 30652 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|30652|0|0|166491473|0 / 166491473|0 / 30652|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999960|13214 / 166491473 (0.007937%)|0 / 30652 (0.000000%)|0.743562|0.822674|0.835888|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.698764|1.000000|0.999921|0.999960|0.999921|0.000079|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.743562|0.822674|0.920624|0.698764|0.835921|0.835888|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|30652|13214|0|166478259|13214 / 166491473|0 / 30652|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|3936|6230|6984|2404|13.172603%|5|21513|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **5 result rows**, **3 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.935853|6230 / 165926276 (0.003755%)|3924 / 30595 (12.825625%)|0.822169|0.840084|0.840609|
+|2|SNOWBALL NORWEGIAN NYNORSK DIRECT|PRIMARY_OUTPUT|0.859037|8274 / 165926276 (0.004987%)|8624 / 30595 (28.187612%)|0.724757|0.722255|0.722216|
+|3|SNOWBALL NORWEGIAN NYNORSK LUCENE FILTER|PRIMARY_OUTPUT|0.858661|8274 / 165926276 (0.004987%)|8647 / 30595 (28.262788%)|0.724438|0.721772|0.721734|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.810644|0.871744|0.999962|0.935853|0.999939|0.000061|
+|2|SNOWBALL NORWEGIAN NYNORSK DIRECT|PRIMARY_OUTPUT|0.726434|0.718124|0.999950|0.859037|0.999898|0.000102|
+|3|SNOWBALL NORWEGIAN NYNORSK LUCENE FILTER|PRIMARY_OUTPUT|0.726226|0.717372|0.999950|0.858661|0.999898|0.000102|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.822169|0.840084|0.858798|0.724263|0.840639|0.840609|
+|2|SNOWBALL NORWEGIAN NYNORSK DIRECT|PRIMARY_OUTPUT|0.724757|0.722255|0.719771|0.565258|0.722267|0.722216|
+|3|SNOWBALL NORWEGIAN NYNORSK LUCENE FILTER|PRIMARY_OUTPUT|0.724438|0.721772|0.719126|0.564666|0.721785|0.721734|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.840054|0.983815|0.986842|0.985326|0.985326|
+|2|SNOWBALL NORWEGIAN NYNORSK DIRECT|PRIMARY_OUTPUT|0.722204|0.980506|0.965065|0.972724|0.972724|
+|3|SNOWBALL NORWEGIAN NYNORSK LUCENE FILTER|PRIMARY_OUTPUT|0.721721|0.980506|0.964945|0.972663|0.972663|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|26671|6230|3924|165920046|6230 / 165926276|3924 / 30595|
+|2|SNOWBALL NORWEGIAN NYNORSK DIRECT|PRIMARY_OUTPUT|21971|8274|8624|165918002|8274 / 165926276|8624 / 30595|
+|3|SNOWBALL NORWEGIAN NYNORSK LUCENE FILTER|PRIMARY_OUTPUT|21948|8274|8647|165918002|8274 / 165926276|8647 / 30595|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 165926276 (0.000000%)|0 / 30595 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|30595|0|0|165926276|0 / 165926276|0 / 30595|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999960|13214 / 165926276 (0.007964%)|0 / 30595 (0.000000%)|0.743207|0.822402|0.835654|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.698372|1.000000|0.999920|0.999960|0.999920|0.000080|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.743207|0.822402|0.920488|0.698372|0.835687|0.835654|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|30595|13214|0|165913062|13214 / 165926276|0 / 30595|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|3924|6230|6984|2399|13.167572%|5|21477|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `NN_NO`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/persian.md
+++ b/docs/benchmarks/languages/persian.md
@@ -0,0 +1,382 @@
+# Persian Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Persian. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `FA_IR` | 69 | 3,770 | 138 | 3,632 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **3,770**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `DeletePrefixCommand` | Deletes one or more leading characters from the word form in forward traversal. | 65 | 1.724% |
+| `ForwardCompoundCommand` | Applies a multi-step forward patch made from skip, delete, insert, and replace operations. | 3,567 | 94.615% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 138 | 3.660% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 95.836% | 95.677% | 100.000% | Full Radixor dictionary patch-command stemmer. |
+| Lucene PersianStemFilter | 1.485% | 0.000% | 40.580% | Lucene Persian suffix stemmer with required normalization in the measured path. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `persianRadixor` | 0.245 | 0.025 | 49.0 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene PersianStemFilter | `persianLucenePersianStemFilter` | 0.466 | 0.015 | 93.3 | 1.902 | Persian suffix stemmer with Lucene normalization in the measured path. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `FA_IR` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/fa_ir/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.974922** among 2 deterministic stemmers. The runner-up is `PERSIAN LUCENE PERSIAN STEM FILTER` at 0.502171, a difference of 0.472751. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.974922** among 2 deterministic stemmers. The runner-up is `PERSIAN LUCENE PERSIAN STEM FILTER` at 0.502171, a difference of 0.472751. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **4 result rows**, **2 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.974922|8621 / 6748402 (0.127749%)|4812 / 98448 (4.887860%)|0.922566|0.933071|0.932249|
+|2|PERSIAN LUCENE PERSIAN STEM FILTER|PRIMARY_OUTPUT|0.502171|179 / 6748402 (0.002652%)|98018 / 98448 (99.563221%)|0.021312|0.008682|0.054801|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.915693|0.951121|0.998723|0.974922|0.998038|0.001962|
+|2|PERSIAN LUCENE PERSIAN STEM FILTER|PRIMARY_OUTPUT|0.706076|0.004368|0.999973|0.502171|0.985658|0.014342|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.922566|0.933071|0.943818|0.874539|0.933239|0.932249|
+|2|PERSIAN LUCENE PERSIAN STEM FILTER|PRIMARY_OUTPUT|0.021312|0.008682|0.005451|0.004360|0.055534|0.054801|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.932076|0.980343|0.984586|0.982460|0.982460|
+|2|PERSIAN LUCENE PERSIAN STEM FILTER|PRIMARY_OUTPUT|0.008507|0.985686|0.520347|0.681125|0.681125|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|93636|8621|4812|6739781|8621 / 6748402|4812 / 98448|
+|2|PERSIAN LUCENE PERSIAN STEM FILTER|PRIMARY_OUTPUT|430|179|98018|6748223|179 / 6748402|98018 / 98448|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 6748402 (0.000000%)|0 / 98448 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|98448|0|0|6748402|0 / 6748402|0 / 98448|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999005|13433 / 6748402 (0.199055%)|0 / 98448 (0.000000%)|0.901585|0.936133|0.937114|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.879935|1.000000|0.998009|0.999005|0.998038|0.001962|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.901585|0.936133|0.973435|0.879935|0.938048|0.937114|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|98448|13433|0|6734969|13433 / 6748402|0 / 98448|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|4812|8621|4812|314|8.484193%|2|4015|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **4 result rows**, **2 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.974922|8621 / 6748402 (0.127749%)|4812 / 98448 (4.887860%)|0.922566|0.933071|0.932249|
+|2|PERSIAN LUCENE PERSIAN STEM FILTER|PRIMARY_OUTPUT|0.502171|179 / 6748402 (0.002652%)|98018 / 98448 (99.563221%)|0.021312|0.008682|0.054801|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.915693|0.951121|0.998723|0.974922|0.998038|0.001962|
+|2|PERSIAN LUCENE PERSIAN STEM FILTER|PRIMARY_OUTPUT|0.706076|0.004368|0.999973|0.502171|0.985658|0.014342|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.922566|0.933071|0.943818|0.874539|0.933239|0.932249|
+|2|PERSIAN LUCENE PERSIAN STEM FILTER|PRIMARY_OUTPUT|0.021312|0.008682|0.005451|0.004360|0.055534|0.054801|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.932076|0.980343|0.984586|0.982460|0.982460|
+|2|PERSIAN LUCENE PERSIAN STEM FILTER|PRIMARY_OUTPUT|0.008507|0.985686|0.520347|0.681125|0.681125|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|93636|8621|4812|6739781|8621 / 6748402|4812 / 98448|
+|2|PERSIAN LUCENE PERSIAN STEM FILTER|PRIMARY_OUTPUT|430|179|98018|6748223|179 / 6748402|98018 / 98448|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 6748402 (0.000000%)|0 / 98448 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|98448|0|0|6748402|0 / 6748402|0 / 98448|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999005|13433 / 6748402 (0.199055%)|0 / 98448 (0.000000%)|0.901585|0.936133|0.937114|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.879935|1.000000|0.998009|0.999005|0.998038|0.001962|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.901585|0.936133|0.973435|0.879935|0.938048|0.937114|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|98448|13433|0|6734969|13433 / 6748402|0 / 98448|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|4812|8621|4812|314|8.484193%|2|4015|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `FA_IR`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/polish.md
+++ b/docs/benchmarks/languages/polish.md
@@ -0,0 +1,464 @@
+# Polish Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Polish. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `PL_PL` | 9,990 | 132,308 | 19,957 | 112,351 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **132,308**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 1,719 | 1.299% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 53,303 | 40.287% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 37,051 | 28.004% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 20,415 | 15.430% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 19,820 | 14.980% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 98.837% | 98.744% | 99.359% | Full Radixor dictionary patch-command stemmer. |
+| Lucene HunspellStemFilter | 89.545% | 88.272% | 96.713% | Benchmark-only Polish Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Lucene MorfologikFilter | 87.729% | 86.606% | 94.047% | Dictionary-based path; Morfologik can emit multiple terms. |
+| Lucene StempelFilter | 70.009% | 69.262% | 74.220% | Lucene TokenFilter integration path for table-driven Polish Stempel. |
+| Lucene StempelStemmer direct | 70.009% | 69.262% | 74.220% | Direct table-driven Polish Stempel stemmer API. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `polishRadixor` | 9.049 | 0.485 | 80.5 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene HunspellStemFilter | `luceneHunspellStemFilter` | 483.316 | 11.455 | 4301.8 | 53.408 | Benchmark-only Polish Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Lucene StempelStemmer direct | `polishLuceneStempelStemmerDirect` | 41.932 | 1.916 | 373.2 | 4.634 | Direct table-driven Polish Stempel stemmer API. |
+| Lucene StempelFilter | `polishLuceneStempelFilter` | 45.277 | 13.693 | 403.0 | 5.003 | Lucene TokenFilter integration path for table-driven Polish Stempel. |
+| Lucene MorfologikFilter | `polishLuceneMorfologikFilter` | 135.763 | 31.634 | 1208.4 | 15.002 | Dictionary-based Morfologik TokenFilter; may emit multiple terms. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `PL_PL` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/pl_pl/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.990388** among 5 deterministic stemmers. The runner-up is `POLISH LUCENE MORFOLOGIK FILTER` at 0.948154, a difference of 0.042234. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.990579** among 5 deterministic stemmers. The runner-up is `POLISH LUCENE MORFOLOGIK FILTER` at 0.948177, a difference of 0.042402. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **11 result rows**, **5 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.990388|13669 / 7482478003 (0.000183%)|21547 / 1120967 (1.922180%)|0.986324|0.984237|0.984241|
+|2|POLISH LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.948154|99228 / 7482478003 (0.001326%)|116220 / 1120967 (10.367834%)|0.907324|0.903167|0.903179|
+|3|HUNSPELL POLISH LUCENE FILTER|PRIMARY_OUTPUT|0.933222|52652 / 7482478003 (0.000704%)|149705 / 1120967 (13.354987%)|0.930930|0.905656|0.906571|
+|4|POLISH LUCENE STEMPEL DIRECT|PRIMARY_OUTPUT|0.855748|66669 / 7482478003 (0.000891%)|323394 / 1120967 (28.849556%)|0.871106|0.803515|0.810296|
+|5|POLISH LUCENE STEMPEL FILTER|PRIMARY_OUTPUT|0.855748|66669 / 7482478003 (0.000891%)|323394 / 1120967 (28.849556%)|0.871106|0.803515|0.810296|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.987720|0.980778|0.999998|0.990388|0.999995|0.000005|
+|2|POLISH LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.910118|0.896322|0.999987|0.948154|0.999971|0.000029|
+|3|HUNSPELL POLISH LUCENE FILTER|PRIMARY_OUTPUT|0.948578|0.866450|0.999993|0.933222|0.999973|0.000027|
+|4|POLISH LUCENE STEMPEL DIRECT|PRIMARY_OUTPUT|0.922858|0.711504|0.999991|0.855748|0.999948|0.000052|
+|5|POLISH LUCENE STEMPEL FILTER|PRIMARY_OUTPUT|0.922858|0.711504|0.999991|0.855748|0.999948|0.000052|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.986324|0.984237|0.982159|0.968963|0.984243|0.984241|
+|2|POLISH LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.907324|0.903167|0.899047|0.823432|0.903193|0.903179|
+|3|HUNSPELL POLISH LUCENE FILTER|PRIMARY_OUTPUT|0.930930|0.905656|0.881718|0.827579|0.906584|0.906571|
+|4|POLISH LUCENE STEMPEL DIRECT|PRIMARY_OUTPUT|0.871106|0.803515|0.745659|0.671564|0.810320|0.810296|
+|5|POLISH LUCENE STEMPEL FILTER|PRIMARY_OUTPUT|0.871106|0.803515|0.745659|0.671564|0.810320|0.810296|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.984234|0.996967|0.996469|0.996718|0.996718|
+|2|POLISH LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.903153|0.990022|0.977054|0.983495|0.983495|
+|3|HUNSPELL POLISH LUCENE FILTER|PRIMARY_OUTPUT|0.905642|0.994546|0.970520|0.982386|0.982386|
+|4|POLISH LUCENE STEMPEL DIRECT|PRIMARY_OUTPUT|0.803490|0.991767|0.931069|0.960460|0.960460|
+|5|POLISH LUCENE STEMPEL FILTER|PRIMARY_OUTPUT|0.803490|0.991767|0.931069|0.960460|0.960460|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|1099420|13669|21547|7482464334|13669 / 7482478003|21547 / 1120967|
+|2|POLISH LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|1004747|99228|116220|7482378775|99228 / 7482478003|116220 / 1120967|
+|3|HUNSPELL POLISH LUCENE FILTER|PRIMARY_OUTPUT|971262|52652|149705|7482425351|52652 / 7482478003|149705 / 1120967|
+|4|POLISH LUCENE STEMPEL DIRECT|PRIMARY_OUTPUT|797573|66669|323394|7482411334|66669 / 7482478003|323394 / 1120967|
+|5|POLISH LUCENE STEMPEL FILTER|PRIMARY_OUTPUT|797573|66669|323394|7482411334|66669 / 7482478003|323394 / 1120967|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 7482478003 (0.000000%)|0 / 1120967 (0.000000%)|1.000000|1.000000|1.000000|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|0.987570|85532 / 7482478003 (0.001143%)|27855 / 1120967 (2.484908%)|0.936598|0.950693|0.950985|
+|3|HUNSPELL POLISH LUCENE FILTER|ANY_CANDIDATE|0.963982|42213 / 7482478003 (0.000564%)|80743 / 1120967 (7.202977%)|0.954209|0.944197|0.944333|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|0.927432|0.975151|0.999989|0.987570|0.999985|0.000015|
+|3|HUNSPELL POLISH LUCENE FILTER|ANY_CANDIDATE|0.961002|0.927970|0.999994|0.963982|0.999984|0.000016|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|0.936598|0.950693|0.965218|0.906020|0.950992|0.950985|
+|3|HUNSPELL POLISH LUCENE FILTER|ANY_CANDIDATE|0.954209|0.944197|0.934394|0.894293|0.944342|0.944333|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|3|HUNSPELL POLISH LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1120967|0|0|7482478003|0 / 7482478003|0 / 1120967|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|1093112|85532|27855|7482392471|85532 / 7482478003|27855 / 1120967|
+|3|HUNSPELL POLISH LUCENE FILTER|ANY_CANDIDATE|1040224|42213|80743|7482435790|42213 / 7482478003|80743 / 1120967|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999997|38073 / 7482478003 (0.000509%)|0 / 1120967 (0.000000%)|0.973547|0.983301|0.983436|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|0.987566|143096 / 7482478003 (0.001912%)|27855 / 1120967 (2.484908%)|0.901045|0.927476|0.928576|
+|3|HUNSPELL POLISH LUCENE FILTER|ALL_CANDIDATES|0.963980|82745 / 7482478003 (0.001106%)|80743 / 1120967 (7.202977%)|0.926646|0.927142|0.927132|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.967151|1.000000|0.999995|0.999997|0.999995|0.000005|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|0.884246|0.975151|0.999981|0.987566|0.999977|0.000023|
+|3|HUNSPELL POLISH LUCENE FILTER|ALL_CANDIDATES|0.926316|0.927970|0.999989|0.963980|0.999978|0.000022|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.973547|0.983301|0.993253|0.967151|0.983438|0.983436|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|0.901045|0.927476|0.955505|0.864761|0.928587|0.928576|
+|3|HUNSPELL POLISH LUCENE FILTER|ALL_CANDIDATES|0.926646|0.927142|0.927639|0.864180|0.927143|0.927132|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|3|HUNSPELL POLISH LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|1120967|38073|0|7482439930|38073 / 7482478003|0 / 1120967|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|1093112|143096|27855|7482334907|143096 / 7482478003|27855 / 1120967|
+|3|HUNSPELL POLISH LUCENE FILTER|ALL_CANDIDATES|1040224|82745|80743|7482395258|82745 / 7482478003|80743 / 1120967|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|HUNSPELL POLISH LUCENE FILTER|68962|10439|30093|11447|9.356634%|6|135231|
+|POLISH LUCENE MORFOLOGIK FILTER|88365|13696|43868|12873|10.522229%|5|136636|
+|Radixor|21547|13669|24404|2866|2.342632%|4|125778|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **11 result rows**, **5 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.990579|13669 / 7310252699 (0.000187%)|21000 / 1114651 (1.883998%)|0.986350|0.984397|0.984400|
+|2|POLISH LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.948177|99224 / 7310252699 (0.001357%)|115513 / 1114651 (10.363154%)|0.906972|0.902966|0.902976|
+|3|HUNSPELL POLISH LUCENE FILTER|PRIMARY_OUTPUT|0.933309|51950 / 7310252699 (0.000711%)|148667 / 1114651 (13.337538%)|0.931269|0.905928|0.906847|
+|4|POLISH LUCENE STEMPEL DIRECT|PRIMARY_OUTPUT|0.856382|66274 / 7310252699 (0.000907%)|320158 / 1114651 (28.722712%)|0.871591|0.804380|0.811082|
+|5|POLISH LUCENE STEMPEL FILTER|PRIMARY_OUTPUT|0.856382|66274 / 7310252699 (0.000907%)|320158 / 1114651 (28.722712%)|0.871591|0.804380|0.811082|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.987656|0.981160|0.999998|0.990579|0.999995|0.000005|
+|2|POLISH LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.909662|0.896368|0.999986|0.948177|0.999971|0.000029|
+|3|HUNSPELL POLISH LUCENE FILTER|PRIMARY_OUTPUT|0.948965|0.866625|0.999993|0.933309|0.999973|0.000027|
+|4|POLISH LUCENE STEMPEL DIRECT|PRIMARY_OUTPUT|0.923006|0.712773|0.999991|0.856382|0.999947|0.000053|
+|5|POLISH LUCENE STEMPEL FILTER|PRIMARY_OUTPUT|0.923006|0.712773|0.999991|0.856382|0.999947|0.000053|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.986350|0.984397|0.982452|0.969274|0.984403|0.984400|
+|2|POLISH LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.906972|0.902966|0.898996|0.823098|0.902991|0.902976|
+|3|HUNSPELL POLISH LUCENE FILTER|PRIMARY_OUTPUT|0.931269|0.905928|0.881929|0.828033|0.906861|0.906847|
+|4|POLISH LUCENE STEMPEL DIRECT|PRIMARY_OUTPUT|0.871591|0.804380|0.746792|0.672772|0.811106|0.811082|
+|5|POLISH LUCENE STEMPEL FILTER|PRIMARY_OUTPUT|0.871591|0.804380|0.746792|0.672772|0.811106|0.811082|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.984395|0.996926|0.996647|0.996786|0.996786|
+|2|POLISH LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.902952|0.989889|0.977012|0.983408|0.983408|
+|3|HUNSPELL POLISH LUCENE FILTER|PRIMARY_OUTPUT|0.905914|0.994584|0.970514|0.982402|0.982402|
+|4|POLISH LUCENE STEMPEL DIRECT|PRIMARY_OUTPUT|0.804354|0.991711|0.931318|0.960566|0.960566|
+|5|POLISH LUCENE STEMPEL FILTER|PRIMARY_OUTPUT|0.804354|0.991711|0.931318|0.960566|0.960566|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|1093651|13669|21000|7310239030|13669 / 7310252699|21000 / 1114651|
+|2|POLISH LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|999138|99224|115513|7310153475|99224 / 7310252699|115513 / 1114651|
+|3|HUNSPELL POLISH LUCENE FILTER|PRIMARY_OUTPUT|965984|51950|148667|7310200749|51950 / 7310252699|148667 / 1114651|
+|4|POLISH LUCENE STEMPEL DIRECT|PRIMARY_OUTPUT|794493|66274|320158|7310186425|66274 / 7310252699|320158 / 1114651|
+|5|POLISH LUCENE STEMPEL FILTER|PRIMARY_OUTPUT|794493|66274|320158|7310186425|66274 / 7310252699|320158 / 1114651|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 7310252699 (0.000000%)|0 / 1114651 (0.000000%)|1.000000|1.000000|1.000000|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|0.987661|85532 / 7310252699 (0.001170%)|27494 / 1114651 (2.466602%)|0.936331|0.950586|0.950885|
+|3|HUNSPELL POLISH LUCENE FILTER|ANY_CANDIDATE|0.963946|41671 / 7310252699 (0.000570%)|80368 / 1114651 (7.210149%)|0.954406|0.944290|0.944429|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|0.927063|0.975334|0.999988|0.987661|0.999985|0.000015|
+|3|HUNSPELL POLISH LUCENE FILTER|ANY_CANDIDATE|0.961271|0.927899|0.999994|0.963946|0.999983|0.000017|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|0.936331|0.950586|0.965282|0.905826|0.950892|0.950885|
+|3|HUNSPELL POLISH LUCENE FILTER|ANY_CANDIDATE|0.954406|0.944290|0.934386|0.894459|0.944437|0.944429|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|3|HUNSPELL POLISH LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1114651|0|0|7310252699|0 / 7310252699|0 / 1114651|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|1087157|85532|27494|7310167167|85532 / 7310252699|27494 / 1114651|
+|3|HUNSPELL POLISH LUCENE FILTER|ANY_CANDIDATE|1034283|41671|80368|7310211028|41671 / 7310252699|80368 / 1114651|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999997|38073 / 7310252699 (0.000521%)|0 / 1114651 (0.000000%)|0.973401|0.983208|0.983344|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|0.987657|143085 / 7310252699 (0.001957%)|27494 / 1114651 (2.466602%)|0.900618|0.927255|0.928372|
+|3|HUNSPELL POLISH LUCENE FILTER|ALL_CANDIDATES|0.963944|81865 / 7310252699 (0.001120%)|80368 / 1114651 (7.210149%)|0.926903|0.927276|0.927265|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.966971|1.000000|0.999995|0.999997|0.999995|0.000005|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|0.883694|0.975334|0.999980|0.987657|0.999977|0.000023|
+|3|HUNSPELL POLISH LUCENE FILTER|ALL_CANDIDATES|0.926654|0.927899|0.999989|0.963944|0.999978|0.000022|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.973401|0.983208|0.993215|0.966971|0.983347|0.983344|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|0.900618|0.927255|0.955516|0.864376|0.928384|0.928372|
+|3|HUNSPELL POLISH LUCENE FILTER|ALL_CANDIDATES|0.926903|0.927276|0.927649|0.864412|0.927276|0.927265|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|3|HUNSPELL POLISH LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|1114651|38073|0|7310214626|38073 / 7310252699|0 / 1114651|
+|2|POLISH LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|1087157|143085|27494|7310109614|143085 / 7310252699|27494 / 1114651|
+|3|HUNSPELL POLISH LUCENE FILTER|ALL_CANDIDATES|1034283|81865|80368|7310170834|81865 / 7310252699|80368 / 1114651|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|HUNSPELL POLISH LUCENE FILTER|68299|10279|29915|11265|9.315692%|6|133595|
+|POLISH LUCENE MORFOLOGIK FILTER|88019|13692|43861|12763|10.554476%|5|135105|
+|Radixor|21000|13669|24404|2780|2.298946%|4|124274|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `PL_PL`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/portuguese.md
+++ b/docs/benchmarks/languages/portuguese.md
@@ -0,0 +1,432 @@
+# Portuguese Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Portuguese. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `PT_PT` | 4,001 | 215,490 | 8,002 | 207,488 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **215,490**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 3,806 | 1.766% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 120,691 | 56.008% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 71,284 | 33.080% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 8,003 | 3.714% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 11,706 | 5.432% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 99.815% | 99.808% | 100.000% | Full Radixor dictionary patch-command stemmer. |
+| Lucene PortugueseLightStemFilter | 8.966% | 5.558% | 97.326% | Light suffix stemmer; intentionally narrower than a dictionary-derived stemmer. |
+| Lucene PortugueseMinimalStemFilter | 5.539% | 1.896% | 100.000% | Minimal suffix reducer; narrow baseline, not a full stemmer. |
+| Lucene SnowballFilter | 0.625% | 0.558% | 2.374% | Lucene TokenFilter integration path around the Snowball algorithm. |
+| Official Snowball direct | 0.625% | 0.558% | 2.374% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+| Lucene PortugueseStemFilter | 0.312% | 0.308% | 0.425% | Portuguese RSLP-style Lucene TokenFilter stemmer. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `portugueseRadixor` | 12.109 | 0.698 | 58.4 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene PortugueseLightStemFilter | `portugueseLucenePortugueseLightStemFilter` | 11.172 | 1.870 | 53.8 | 0.923 | Light Portuguese suffix stemmer. |
+| Lucene PortugueseMinimalStemFilter | `portugueseLucenePortugueseMinimalStemFilter` | 16.038 | 1.752 | 77.3 | 1.325 | Minimal Portuguese suffix reducer. |
+| Official Snowball direct | `snowballDirect[PORTUGUESE]` | 53.725 | 5.356 | 258.9 | 4.437 | Official Snowball generated Java stemmer; direct API. |
+| Lucene SnowballFilter | `luceneSnowballFilter[PORTUGUESE]` | 57.457 | 1.182 | 276.9 | 4.745 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+| Lucene PortugueseStemFilter | `portugueseLucenePortugueseStemFilter` | 165.447 | 40.334 | 797.4 | 13.663 | Portuguese RSLP-style Lucene TokenFilter. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `PT_PT` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/pt_pt/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.998502** among 6 deterministic stemmers. The runner-up is `SNOWBALL PORTUGUESE DIRECT` at 0.938800, a difference of 0.059702. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.998502** among 6 deterministic stemmers. The runner-up is `SNOWBALL PORTUGUESE DIRECT` at 0.938800, a difference of 0.059702. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **8 result rows**, **6 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.998502|20678 / 22358203756 (0.000092%)|16444 / 5489060 (0.299578%)|0.996389|0.996620|0.996619|
+|2|SNOWBALL PORTUGUESE DIRECT|PRIMARY_OUTPUT|0.938800|167230 / 22358203756 (0.000748%)|671821 / 5489060 (12.239272%)|0.947271|0.919888|0.920940|
+|3|SNOWBALL PORTUGUESE LUCENE FILTER|PRIMARY_OUTPUT|0.938800|167230 / 22358203756 (0.000748%)|671821 / 5489060 (12.239272%)|0.947271|0.919888|0.920940|
+|4|PORTUGUESE LUCENE PORTUGUESE STEM FILTER|PRIMARY_OUTPUT|0.846459|99075 / 22358203756 (0.000443%)|1685572 / 5489060 (30.707844%)|0.901330|0.809975|0.821750|
+|5|PORTUGUESE LUCENE PORTUGUESE LIGHT STEM FILTER|PRIMARY_OUTPUT|0.513648|2511 / 22358203756 (0.000011%)|5339230 / 5489060 (97.270389%)|0.122843|0.053118|0.163828|
+|6|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.503954|598 / 22358203756 (0.000003%)|5445654 / 5489060 (99.209227%)|0.038310|0.015690|0.088308|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.996236|0.997004|0.999999|0.998502|0.999998|0.000002|
+|2|SNOWBALL PORTUGUESE DIRECT|PRIMARY_OUTPUT|0.966450|0.877607|0.999993|0.938800|0.999962|0.000038|
+|3|SNOWBALL PORTUGUESE LUCENE FILTER|PRIMARY_OUTPUT|0.966450|0.877607|0.999993|0.938800|0.999962|0.000038|
+|4|PORTUGUESE LUCENE PORTUGUESE STEM FILTER|PRIMARY_OUTPUT|0.974613|0.692922|0.999996|0.846459|0.999920|0.000080|
+|5|PORTUGUESE LUCENE PORTUGUESE LIGHT STEM FILTER|PRIMARY_OUTPUT|0.983517|0.027296|1.000000|0.513648|0.999761|0.000239|
+|6|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.986410|0.007908|1.000000|0.503954|0.999756|0.000244|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.996389|0.996620|0.996850|0.993262|0.996620|0.996619|
+|2|SNOWBALL PORTUGUESE DIRECT|PRIMARY_OUTPUT|0.947271|0.919888|0.894045|0.851661|0.920958|0.920940|
+|3|SNOWBALL PORTUGUESE LUCENE FILTER|PRIMARY_OUTPUT|0.947271|0.919888|0.894045|0.851661|0.920958|0.920940|
+|4|PORTUGUESE LUCENE PORTUGUESE STEM FILTER|PRIMARY_OUTPUT|0.901330|0.809975|0.735434|0.680636|0.821785|0.821750|
+|5|PORTUGUESE LUCENE PORTUGUESE LIGHT STEM FILTER|PRIMARY_OUTPUT|0.122843|0.053118|0.033885|0.027284|0.163848|0.163828|
+|6|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.038310|0.015690|0.009865|0.007907|0.088319|0.088308|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.996619|0.999299|0.999347|0.999323|0.999323|
+|2|SNOWBALL PORTUGUESE DIRECT|PRIMARY_OUTPUT|0.919870|0.996663|0.967924|0.982083|0.982083|
+|3|SNOWBALL PORTUGUESE LUCENE FILTER|PRIMARY_OUTPUT|0.919870|0.996663|0.967924|0.982083|0.982083|
+|4|PORTUGUESE LUCENE PORTUGUESE STEM FILTER|PRIMARY_OUTPUT|0.809936|0.996729|0.918475|0.956003|0.956003|
+|5|PORTUGUESE LUCENE PORTUGUESE LIGHT STEM FILTER|PRIMARY_OUTPUT|0.053105|0.999226|0.720580|0.837330|0.837330|
+|6|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.015686|0.999664|0.692383|0.818122|0.818122|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|5472616|20678|16444|22358183078|20678 / 22358203756|16444 / 5489060|
+|2|SNOWBALL PORTUGUESE DIRECT|PRIMARY_OUTPUT|4817239|167230|671821|22358036526|167230 / 22358203756|671821 / 5489060|
+|3|SNOWBALL PORTUGUESE LUCENE FILTER|PRIMARY_OUTPUT|4817239|167230|671821|22358036526|167230 / 22358203756|671821 / 5489060|
+|4|PORTUGUESE LUCENE PORTUGUESE STEM FILTER|PRIMARY_OUTPUT|3803488|99075|1685572|22358104681|99075 / 22358203756|1685572 / 5489060|
+|5|PORTUGUESE LUCENE PORTUGUESE LIGHT STEM FILTER|PRIMARY_OUTPUT|149830|2511|5339230|22358201245|2511 / 22358203756|5339230 / 5489060|
+|6|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|PRIMARY_OUTPUT|43406|598|5445654|22358203158|598 / 22358203756|5445654 / 5489060|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 22358203756 (0.000000%)|0 / 5489060 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|5489060|0|0|22358203756|0 / 22358203756|0 / 5489060|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999999|38310 / 22358203756 (0.000171%)|0 / 5489060 (0.000000%)|0.994448|0.996522|0.996528|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.993069|1.000000|0.999998|0.999999|0.999998|0.000002|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.994448|0.996522|0.998606|0.993069|0.996528|0.996528|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|5489060|38310|0|22358165446|38310 / 22358203756|0 / 5489060|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|16444|20678|17632|790|0.373542%|3|212297|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **8 result rows**, **6 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.998502|20678 / 22358203756 (0.000092%)|16444 / 5489060 (0.299578%)|0.996389|0.996620|0.996619|
+|2|SNOWBALL PORTUGUESE DIRECT|PRIMARY_OUTPUT|0.938800|167230 / 22358203756 (0.000748%)|671821 / 5489060 (12.239272%)|0.947271|0.919888|0.920940|
+|3|SNOWBALL PORTUGUESE LUCENE FILTER|PRIMARY_OUTPUT|0.938800|167230 / 22358203756 (0.000748%)|671821 / 5489060 (12.239272%)|0.947271|0.919888|0.920940|
+|4|PORTUGUESE LUCENE PORTUGUESE STEM FILTER|PRIMARY_OUTPUT|0.846459|99075 / 22358203756 (0.000443%)|1685572 / 5489060 (30.707844%)|0.901330|0.809975|0.821750|
+|5|PORTUGUESE LUCENE PORTUGUESE LIGHT STEM FILTER|PRIMARY_OUTPUT|0.513648|2511 / 22358203756 (0.000011%)|5339230 / 5489060 (97.270389%)|0.122843|0.053118|0.163828|
+|6|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.503954|598 / 22358203756 (0.000003%)|5445654 / 5489060 (99.209227%)|0.038310|0.015690|0.088308|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.996236|0.997004|0.999999|0.998502|0.999998|0.000002|
+|2|SNOWBALL PORTUGUESE DIRECT|PRIMARY_OUTPUT|0.966450|0.877607|0.999993|0.938800|0.999962|0.000038|
+|3|SNOWBALL PORTUGUESE LUCENE FILTER|PRIMARY_OUTPUT|0.966450|0.877607|0.999993|0.938800|0.999962|0.000038|
+|4|PORTUGUESE LUCENE PORTUGUESE STEM FILTER|PRIMARY_OUTPUT|0.974613|0.692922|0.999996|0.846459|0.999920|0.000080|
+|5|PORTUGUESE LUCENE PORTUGUESE LIGHT STEM FILTER|PRIMARY_OUTPUT|0.983517|0.027296|1.000000|0.513648|0.999761|0.000239|
+|6|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.986410|0.007908|1.000000|0.503954|0.999756|0.000244|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.996389|0.996620|0.996850|0.993262|0.996620|0.996619|
+|2|SNOWBALL PORTUGUESE DIRECT|PRIMARY_OUTPUT|0.947271|0.919888|0.894045|0.851661|0.920958|0.920940|
+|3|SNOWBALL PORTUGUESE LUCENE FILTER|PRIMARY_OUTPUT|0.947271|0.919888|0.894045|0.851661|0.920958|0.920940|
+|4|PORTUGUESE LUCENE PORTUGUESE STEM FILTER|PRIMARY_OUTPUT|0.901330|0.809975|0.735434|0.680636|0.821785|0.821750|
+|5|PORTUGUESE LUCENE PORTUGUESE LIGHT STEM FILTER|PRIMARY_OUTPUT|0.122843|0.053118|0.033885|0.027284|0.163848|0.163828|
+|6|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.038310|0.015690|0.009865|0.007907|0.088319|0.088308|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.996619|0.999299|0.999347|0.999323|0.999323|
+|2|SNOWBALL PORTUGUESE DIRECT|PRIMARY_OUTPUT|0.919870|0.996663|0.967924|0.982083|0.982083|
+|3|SNOWBALL PORTUGUESE LUCENE FILTER|PRIMARY_OUTPUT|0.919870|0.996663|0.967924|0.982083|0.982083|
+|4|PORTUGUESE LUCENE PORTUGUESE STEM FILTER|PRIMARY_OUTPUT|0.809936|0.996729|0.918475|0.956003|0.956003|
+|5|PORTUGUESE LUCENE PORTUGUESE LIGHT STEM FILTER|PRIMARY_OUTPUT|0.053105|0.999226|0.720580|0.837330|0.837330|
+|6|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.015686|0.999664|0.692383|0.818122|0.818122|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|5472616|20678|16444|22358183078|20678 / 22358203756|16444 / 5489060|
+|2|SNOWBALL PORTUGUESE DIRECT|PRIMARY_OUTPUT|4817239|167230|671821|22358036526|167230 / 22358203756|671821 / 5489060|
+|3|SNOWBALL PORTUGUESE LUCENE FILTER|PRIMARY_OUTPUT|4817239|167230|671821|22358036526|167230 / 22358203756|671821 / 5489060|
+|4|PORTUGUESE LUCENE PORTUGUESE STEM FILTER|PRIMARY_OUTPUT|3803488|99075|1685572|22358104681|99075 / 22358203756|1685572 / 5489060|
+|5|PORTUGUESE LUCENE PORTUGUESE LIGHT STEM FILTER|PRIMARY_OUTPUT|149830|2511|5339230|22358201245|2511 / 22358203756|5339230 / 5489060|
+|6|PORTUGUESE LUCENE PORTUGUESE MINIMAL STEM FILTER|PRIMARY_OUTPUT|43406|598|5445654|22358203158|598 / 22358203756|5445654 / 5489060|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 22358203756 (0.000000%)|0 / 5489060 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|5489060|0|0|22358203756|0 / 22358203756|0 / 5489060|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999999|38310 / 22358203756 (0.000171%)|0 / 5489060 (0.000000%)|0.994448|0.996522|0.996528|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.993069|1.000000|0.999998|0.999999|0.999998|0.000002|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.994448|0.996522|0.998606|0.993069|0.996528|0.996528|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|5489060|38310|0|22358165446|38310 / 22358203756|0 / 5489060|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|16444|20678|17632|790|0.373542%|3|212297|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `PT_PT`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/russian.md
+++ b/docs/benchmarks/languages/russian.md
@@ -0,0 +1,408 @@
+# Russian Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Russian. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `RU_RU` | 37,410 | 806,279 | 74,808 | 731,471 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **806,279**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 9,260 | 1.148% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 584,785 | 72.529% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 82,864 | 10.277% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 75,646 | 9.382% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 53,724 | 6.663% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 98.807% | 98.696% | 99.896% | Full Radixor dictionary patch-command stemmer. |
+| Lucene RussianLightStemFilter | 9.658% | 8.452% | 21.447% | Light suffix stemmer; intentionally narrower than a dictionary-derived stemmer. |
+| Lucene SnowballFilter | 9.162% | 8.162% | 18.936% | Lucene TokenFilter integration path around the Snowball algorithm. |
+| Official Snowball direct | 9.162% | 8.162% | 18.936% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `russianRadixor` | 89.671 | 3.886 | 122.6 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene RussianLightStemFilter | `russianLuceneRussianLightStemFilter` | 60.522 | 5.310 | 82.7 | 0.675 | Light Russian suffix stemmer. |
+| Official Snowball direct | `snowballDirect[RUSSIAN]` | 106.031 | 9.287 | 145.0 | 1.182 | Official Snowball generated Java stemmer; direct API. |
+| Lucene SnowballFilter | `luceneSnowballFilter[RUSSIAN]` | 137.512 | 10.801 | 188.0 | 1.534 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `RU_RU` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/ru_ru/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.989827** among 4 deterministic stemmers. The runner-up is `SNOWBALL RUSSIAN LUCENE FILTER` at 0.834876, a difference of 0.154951. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.989852** among 4 deterministic stemmers. The runner-up is `SNOWBALL RUSSIAN DIRECT` at 0.834854, a difference of 0.154998. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **6 result rows**, **4 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.989827|155850 / 295576291016 (0.000053%)|266302 / 13089505 (2.034470%)|0.986313|0.983806|0.983814|
+|2|SNOWBALL RUSSIAN LUCENE FILTER|PRIMARY_OUTPUT|0.834876|3785790 / 295576291016 (0.001281%)|4322616 / 13089505 (33.023525%)|0.692485|0.683786|0.683923|
+|3|SNOWBALL RUSSIAN DIRECT|PRIMARY_OUTPUT|0.834867|3782908 / 295576291016 (0.001280%)|4322849 / 13089505 (33.025305%)|0.692603|0.683851|0.683989|
+|4|RUSSIAN LUCENE RUSSIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.617692|321183 / 295576291016 (0.000109%)|10008438 / 13089505 (76.461547%)|0.577011|0.373649|0.461687|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.987992|0.979655|0.999999|0.989827|0.999999|0.000001|
+|2|SNOWBALL RUSSIAN LUCENE FILTER|PRIMARY_OUTPUT|0.698408|0.669765|0.999987|0.834876|0.999973|0.000027|
+|3|SNOWBALL RUSSIAN DIRECT|PRIMARY_OUTPUT|0.698563|0.669747|0.999987|0.834867|0.999973|0.000027|
+|4|RUSSIAN LUCENE RUSSIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.905597|0.235385|0.999999|0.617692|0.999965|0.000035|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.986313|0.983806|0.981311|0.968128|0.983815|0.983814|
+|2|SNOWBALL RUSSIAN LUCENE FILTER|PRIMARY_OUTPUT|0.692485|0.683786|0.675304|0.519510|0.683936|0.683923|
+|3|SNOWBALL RUSSIAN DIRECT|PRIMARY_OUTPUT|0.692603|0.683851|0.675318|0.519585|0.684003|0.683989|
+|4|RUSSIAN LUCENE RUSSIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.577011|0.373649|0.276278|0.229747|0.461696|0.461687|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.983805|0.997699|0.997274|0.997487|0.997487|
+|2|SNOWBALL RUSSIAN LUCENE FILTER|PRIMARY_OUTPUT|0.683773|0.974131|0.953674|0.963794|0.963794|
+|3|SNOWBALL RUSSIAN DIRECT|PRIMARY_OUTPUT|0.683838|0.974180|0.953661|0.963811|0.963811|
+|4|RUSSIAN LUCENE RUSSIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.373638|0.994311|0.870888|0.928516|0.928516|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|12823203|155850|266302|295576135166|155850 / 295576291016|266302 / 13089505|
+|2|SNOWBALL RUSSIAN LUCENE FILTER|PRIMARY_OUTPUT|8766889|3785790|4322616|295572505226|3785790 / 295576291016|4322616 / 13089505|
+|3|SNOWBALL RUSSIAN DIRECT|PRIMARY_OUTPUT|8766656|3782908|4322849|295572508108|3782908 / 295576291016|4322849 / 13089505|
+|4|RUSSIAN LUCENE RUSSIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|3081067|321183|10008438|295575969833|321183 / 295576291016|10008438 / 13089505|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 295576291016 (0.000000%)|13 / 13089505 (0.000099%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0.999999|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|0.999999|0.999999|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|13089492|0|13|295576291016|0 / 295576291016|13 / 13089505|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999999|434710 / 295576291016 (0.000147%)|13 / 13089505 (0.000099%)|0.974119|0.983665|0.983796|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.967857|0.999999|0.999999|0.999999|0.999999|0.000001|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.974119|0.983665|0.993401|0.967856|0.983797|0.983796|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|13089492|434710|13|295575856306|434710 / 295576291016|13 / 13089505|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|266289|155850|278860|19162|2.492190%|4|788492|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **6 result rows**, **4 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.989852|155850 / 295000681652 (0.000053%)|265613 / 13087126 (2.029575%)|0.986322|0.983830|0.983838|
+|2|SNOWBALL RUSSIAN DIRECT|PRIMARY_OUTPUT|0.834854|3782908 / 295000681652 (0.001282%)|4322407 / 13087126 (33.027931%)|0.692561|0.683815|0.683953|
+|3|SNOWBALL RUSSIAN LUCENE FILTER|PRIMARY_OUTPUT|0.834854|3782908 / 295000681652 (0.001282%)|4322407 / 13087126 (33.027931%)|0.692561|0.683815|0.683953|
+|4|RUSSIAN LUCENE RUSSIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.617630|318921 / 295000681652 (0.000108%)|10008238 / 13087126 (76.473918%)|0.577038|0.373540|0.461703|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.987991|0.979704|0.999999|0.989852|0.999999|0.000001|
+|2|SNOWBALL RUSSIAN DIRECT|PRIMARY_OUTPUT|0.698516|0.669721|0.999987|0.834854|0.999973|0.000027|
+|3|SNOWBALL RUSSIAN LUCENE FILTER|PRIMARY_OUTPUT|0.698516|0.669721|0.999987|0.834854|0.999973|0.000027|
+|4|RUSSIAN LUCENE RUSSIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.906139|0.235261|0.999999|0.617630|0.999965|0.000035|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.986322|0.983830|0.981350|0.968175|0.983839|0.983838|
+|2|SNOWBALL RUSSIAN DIRECT|PRIMARY_OUTPUT|0.692561|0.683815|0.675288|0.519544|0.683967|0.683953|
+|3|SNOWBALL RUSSIAN LUCENE FILTER|PRIMARY_OUTPUT|0.692561|0.683815|0.675288|0.519544|0.683967|0.683953|
+|4|RUSSIAN LUCENE RUSSIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.577038|0.373540|0.276152|0.229664|0.461713|0.461703|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.983829|0.997697|0.997321|0.997509|0.997509|
+|2|SNOWBALL RUSSIAN DIRECT|PRIMARY_OUTPUT|0.683802|0.974149|0.953634|0.963782|0.963782|
+|3|SNOWBALL RUSSIAN LUCENE FILTER|PRIMARY_OUTPUT|0.683802|0.974149|0.953634|0.963782|0.963782|
+|4|RUSSIAN LUCENE RUSSIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|0.373528|0.994350|0.870767|0.928464|0.928464|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|12821513|155850|265613|295000525802|155850 / 295000681652|265613 / 13087126|
+|2|SNOWBALL RUSSIAN DIRECT|PRIMARY_OUTPUT|8764719|3782908|4322407|294996898744|3782908 / 295000681652|4322407 / 13087126|
+|3|SNOWBALL RUSSIAN LUCENE FILTER|PRIMARY_OUTPUT|8764719|3782908|4322407|294996898744|3782908 / 295000681652|4322407 / 13087126|
+|4|RUSSIAN LUCENE RUSSIAN LIGHT STEM FILTER|PRIMARY_OUTPUT|3078888|318921|10008238|295000362731|318921 / 295000681652|10008238 / 13087126|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 295000681652 (0.000000%)|0 / 13087126 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|13087126|0|0|295000681652|0 / 295000681652|0 / 13087126|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999999|434710 / 295000681652 (0.000147%)|0 / 13087126 (0.000000%)|0.974115|0.983663|0.983794|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.967851|1.000000|0.999999|0.999999|0.999999|0.000001|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.974115|0.983663|0.993401|0.967851|0.983794|0.983794|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|13087126|434710|0|295000246942|434710 / 295000681652|0 / 13087126|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|265613|155850|278860|18991|2.472358%|4|787549|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `RU_RU`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/spanish.md
+++ b/docs/benchmarks/languages/spanish.md
@@ -0,0 +1,466 @@
+# Spanish Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Spanish. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `ES_ES` | 65,059 | 926,393 | 120,121 | 806,272 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **926,393**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 5,367 | 0.579% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 524,682 | 56.637% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 240,872 | 26.001% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 130,089 | 14.043% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 25,383 | 2.740% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 97.459% | 97.544% | 96.891% | Full Radixor dictionary patch-command stemmer. |
+| Lucene HunspellStemFilter | 49.074% | 42.656% | 92.154% | Benchmark-only Spanish Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Lucene SpanishMinimalStemFilter | 17.284% | 5.347% | 97.403% | Minimal suffix reducer; narrow baseline, not a full stemmer. |
+| Lucene SpanishPluralStemFilter | 15.140% | 5.802% | 77.820% | Plural-focused suffix reducer; narrow baseline. |
+| Lucene SpanishLightStemFilter | 9.577% | 7.088% | 26.279% | Light suffix stemmer; intentionally narrower than a dictionary-derived stemmer. |
+| Lucene SnowballFilter | 4.889% | 4.287% | 8.932% | Lucene TokenFilter integration path around the Snowball algorithm. |
+| Official Snowball direct | 4.889% | 4.287% | 8.930% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `spanishRadixor` | 78.919 | 7.253 | 97.9 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene HunspellStemFilter | `luceneHunspellStemFilter` | 2079.041 | 193.548 | 2578.6 | 26.344 | Benchmark-only Spanish Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Lucene SpanishMinimalStemFilter | `spanishLuceneSpanishMinimalStemFilter` | 45.596 | 4.639 | 56.6 | 0.578 | Minimal Spanish suffix reducer; narrow baseline. |
+| Lucene SpanishLightStemFilter | `spanishLuceneSpanishLightStemFilter` | 42.003 | 1.683 | 52.1 | 0.532 | Light Spanish suffix stemmer. |
+| Lucene SpanishPluralStemFilter | `spanishLuceneSpanishPluralStemFilter` | 93.734 | 6.247 | 116.3 | 1.188 | Plural-oriented Spanish suffix reducer. |
+| Official Snowball direct | `snowballDirect[SPANISH]` | 171.995 | 11.035 | 213.3 | 2.179 | Official Snowball generated Java stemmer; direct API. |
+| Lucene SnowballFilter | `luceneSnowballFilter[SPANISH]` | 211.138 | 17.940 | 261.9 | 2.675 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `ES_ES` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/es_es/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.989295** among 7 deterministic stemmers. The runner-up is `SNOWBALL SPANISH LUCENE FILTER` at 0.652614, a difference of 0.336680. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.989429** among 7 deterministic stemmers. The runner-up is `SNOWBALL SPANISH DIRECT` at 0.652720, a difference of 0.336709. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **11 result rows**, **7 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.989295|288483 / 379567318110 (0.000076%)|898652 / 41973336 (2.141007%)|0.990105|0.985755|0.985780|
+|2|SNOWBALL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.652614|2230481 / 379567318110 (0.000588%)|29161643 / 41973336 (69.476591%)|0.627151|0.449411|0.509848|
+|3|SNOWBALL SPANISH DIRECT|PRIMARY_OUTPUT|0.652614|2228819 / 379567318110 (0.000587%)|29161649 / 41973336 (69.476605%)|0.627192|0.449424|0.509876|
+|4|HUNSPELL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.615102|536192 / 379567318110 (0.000141%)|32310860 / 41973336 (76.979490%)|0.583708|0.370408|0.466992|
+|5|SPANISH LUCENE SPANISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.514823|147956 / 379567318110 (0.000039%)|40729019 / 41973336 (97.035458%)|0.130864|0.057387|0.162762|
+|6|SPANISH LUCENE SPANISH PLURAL STEM FILTER|PRIMARY_OUTPUT|0.503874|58578 / 379567318110 (0.000015%)|41648091 / 41973336 (99.225115%)|0.037377|0.015357|0.081026|
+|7|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.501768|47859 / 379567318110 (0.000013%)|41824873 / 41973336 (99.646292%)|0.017361|0.007041|0.051714|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.993026|0.978590|0.999999|0.989295|0.999997|0.000003|
+|2|SNOWBALL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.851718|0.305234|0.999994|0.652614|0.999917|0.000083|
+|3|SNOWBALL SPANISH DIRECT|PRIMARY_OUTPUT|0.851812|0.305234|0.999994|0.652614|0.999917|0.000083|
+|4|HUNSPELL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.947425|0.230205|0.999999|0.615102|0.999913|0.000087|
+|5|SPANISH LUCENE SPANISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.893731|0.029645|1.000000|0.514823|0.999892|0.000108|
+|6|SPANISH LUCENE SPANISH PLURAL STEM FILTER|PRIMARY_OUTPUT|0.847383|0.007749|1.000000|0.503874|0.999890|0.000110|
+|7|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.756222|0.003537|1.000000|0.501768|0.999890|0.000110|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.990105|0.985755|0.981443|0.971910|0.985781|0.985780|
+|2|SNOWBALL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.627151|0.449411|0.350170|0.289832|0.509876|0.509848|
+|3|SNOWBALL SPANISH DIRECT|PRIMARY_OUTPUT|0.627192|0.449424|0.350173|0.289843|0.509904|0.509876|
+|4|HUNSPELL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.583708|0.370408|0.271278|0.227301|0.467014|0.466992|
+|5|SPANISH LUCENE SPANISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.130864|0.057387|0.036752|0.029541|0.162773|0.162762|
+|6|SPANISH LUCENE SPANISH PLURAL STEM FILTER|PRIMARY_OUTPUT|0.037377|0.015357|0.009664|0.007738|0.081032|0.081026|
+|7|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.017361|0.007041|0.004416|0.003533|0.051719|0.051714|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.985753|0.995418|0.993266|0.994341|0.994341|
+|2|SNOWBALL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.449379|0.981386|0.852461|0.912391|0.912391|
+|3|SNOWBALL SPANISH DIRECT|PRIMARY_OUTPUT|0.449392|0.981406|0.852463|0.912401|0.912401|
+|4|HUNSPELL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.370381|0.993314|0.790558|0.880414|0.880414|
+|5|SPANISH LUCENE SPANISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.057381|0.993824|0.756690|0.859195|0.859195|
+|6|SPANISH LUCENE SPANISH PLURAL STEM FILTER|PRIMARY_OUTPUT|0.015355|0.995442|0.723731|0.838115|0.838115|
+|7|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.007040|0.995635|0.710610|0.829316|0.829316|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|41074684|288483|898652|379567029627|288483 / 379567318110|898652 / 41973336|
+|2|SNOWBALL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|12811693|2230481|29161643|379565087629|2230481 / 379567318110|29161643 / 41973336|
+|3|SNOWBALL SPANISH DIRECT|PRIMARY_OUTPUT|12811687|2228819|29161649|379565089291|2228819 / 379567318110|29161649 / 41973336|
+|4|HUNSPELL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|9662476|536192|32310860|379566781918|536192 / 379567318110|32310860 / 41973336|
+|5|SPANISH LUCENE SPANISH LIGHT STEM FILTER|PRIMARY_OUTPUT|1244317|147956|40729019|379567170154|147956 / 379567318110|40729019 / 41973336|
+|6|SPANISH LUCENE SPANISH PLURAL STEM FILTER|PRIMARY_OUTPUT|325245|58578|41648091|379567259532|58578 / 379567318110|41648091 / 41973336|
+|7|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|148463|47859|41824873|379567270251|47859 / 379567318110|41824873 / 41973336|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|0.999993|2 / 379567318110 (0.000000%)|626 / 41973336 (0.001491%)|0.999997|0.999993|0.999993|
+|2|HUNSPELL SPANISH LUCENE FILTER|ANY_CANDIDATE|0.620065|416345 / 379567318110 (0.000110%)|31894218 / 41973336 (75.986855%)|0.600268|0.384195|0.480192|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0.999985|1.000000|0.999993|1.000000|0.000000|
+|2|HUNSPELL SPANISH LUCENE FILTER|ANY_CANDIDATE|0.960331|0.240131|0.999999|0.620065|0.999915|0.000085|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|0.999997|0.999993|0.999988|0.999985|0.999993|0.999993|
+|2|HUNSPELL SPANISH LUCENE FILTER|ANY_CANDIDATE|0.600268|0.384195|0.282504|0.237773|0.480214|0.480192|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL SPANISH LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|41972710|2|626|379567318108|2 / 379567318110|626 / 41973336|
+|2|HUNSPELL SPANISH LUCENE FILTER|ANY_CANDIDATE|10079118|416345|31894218|379566901765|416345 / 379567318110|31894218 / 41973336|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999991|1349800 / 379567318110 (0.000356%)|626 / 41973336 (0.001491%)|0.974915|0.984168|0.984289|
+|2|HUNSPELL SPANISH LUCENE FILTER|ALL_CANDIDATES|0.620065|888077 / 379567318110 (0.000234%)|31894218 / 41973336 (75.986855%)|0.587073|0.380771|0.469749|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.968843|0.999985|0.999996|0.999991|0.999996|0.000004|
+|2|HUNSPELL SPANISH LUCENE FILTER|ALL_CANDIDATES|0.919024|0.240131|0.999998|0.620065|0.999914|0.000086|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.974915|0.984168|0.993598|0.968829|0.984291|0.984289|
+|2|HUNSPELL SPANISH LUCENE FILTER|ALL_CANDIDATES|0.587073|0.380771|0.281759|0.235156|0.469773|0.469749|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL SPANISH LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|41972710|1349800|626|379565968310|1349800 / 379567318110|626 / 41973336|
+|2|HUNSPELL SPANISH LUCENE FILTER|ALL_CANDIDATES|10079118|888077|31894218|379566430033|888077 / 379567318110|31894218 / 41973336|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|HUNSPELL SPANISH LUCENE FILTER|416642|119847|351885|17877|2.051686%|5|890999|
+|Radixor|898026|288481|1061317|42637|4.893313%|21|916797|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **11 result rows**, **7 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.989429|276044 / 377860669765 (0.000073%)|885033 / 41863370 (2.114099%)|0.990385|0.986031|0.986056|
+|2|SNOWBALL SPANISH DIRECT|PRIMARY_OUTPUT|0.652720|2201196 / 377860669765 (0.000583%)|29076352 / 41863370 (69.455354%)|0.627946|0.449839|0.510450|
+|3|SNOWBALL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.652720|2201196 / 377860669765 (0.000583%)|29076352 / 41863370 (69.455354%)|0.627946|0.449839|0.510450|
+|4|HUNSPELL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.614999|531181 / 377860669765 (0.000141%)|32234855 / 41863370 (77.000144%)|0.583531|0.370163|0.466854|
+|5|SPANISH LUCENE SPANISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.514832|146613 / 377860669765 (0.000039%)|40621522 / 41863370 (97.033569%)|0.130949|0.057424|0.162875|
+|6|SPANISH LUCENE SPANISH PLURAL STEM FILTER|PRIMARY_OUTPUT|0.503877|57716 / 377860669765 (0.000015%)|41538714 / 41863370 (99.224487%)|0.037409|0.015370|0.081139|
+|7|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.501770|47148 / 377860669765 (0.000012%)|41715144 / 41863370 (99.645929%)|0.017379|0.007049|0.051824|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.993309|0.978859|0.999999|0.989429|0.999997|0.000003|
+|2|SNOWBALL SPANISH DIRECT|PRIMARY_OUTPUT|0.853138|0.305446|0.999994|0.652720|0.999917|0.000083|
+|3|SNOWBALL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.853138|0.305446|0.999994|0.652720|0.999917|0.000083|
+|4|HUNSPELL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.947717|0.229999|0.999999|0.614999|0.999913|0.000087|
+|5|SPANISH LUCENE SPANISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.894406|0.029664|1.000000|0.514832|0.999892|0.000108|
+|6|SPANISH LUCENE SPANISH PLURAL STEM FILTER|PRIMARY_OUTPUT|0.849058|0.007755|1.000000|0.503877|0.999890|0.000110|
+|7|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.758678|0.003541|1.000000|0.501770|0.999889|0.000111|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.990385|0.986031|0.981715|0.972447|0.986057|0.986056|
+|2|SNOWBALL SPANISH DIRECT|PRIMARY_OUTPUT|0.627946|0.449839|0.350441|0.290188|0.510478|0.510450|
+|3|SNOWBALL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.627946|0.449839|0.350441|0.290188|0.510478|0.510450|
+|4|HUNSPELL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.583531|0.370163|0.271053|0.227117|0.466876|0.466854|
+|5|SPANISH LUCENE SPANISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.130949|0.057424|0.036775|0.029561|0.162886|0.162875|
+|6|SPANISH LUCENE SPANISH PLURAL STEM FILTER|PRIMARY_OUTPUT|0.037409|0.015370|0.009672|0.007744|0.081145|0.081139|
+|7|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.017379|0.007049|0.004421|0.003537|0.051829|0.051824|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.986029|0.995464|0.993323|0.994392|0.994392|
+|2|SNOWBALL SPANISH DIRECT|PRIMARY_OUTPUT|0.449806|0.981469|0.852556|0.912482|0.912482|
+|3|SNOWBALL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.449806|0.981469|0.852556|0.912482|0.912482|
+|4|HUNSPELL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|0.370136|0.993362|0.790500|0.880396|0.880396|
+|5|SPANISH LUCENE SPANISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.057417|0.993866|0.756725|0.859234|0.859234|
+|6|SPANISH LUCENE SPANISH PLURAL STEM FILTER|PRIMARY_OUTPUT|0.015368|0.995484|0.723753|0.838145|0.838145|
+|7|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.007048|0.995676|0.710626|0.829341|0.829341|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|40978337|276044|885033|377860393721|276044 / 377860669765|885033 / 41863370|
+|2|SNOWBALL SPANISH DIRECT|PRIMARY_OUTPUT|12787018|2201196|29076352|377858468569|2201196 / 377860669765|29076352 / 41863370|
+|3|SNOWBALL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|12787018|2201196|29076352|377858468569|2201196 / 377860669765|29076352 / 41863370|
+|4|HUNSPELL SPANISH LUCENE FILTER|PRIMARY_OUTPUT|9628515|531181|32234855|377860138584|531181 / 377860669765|32234855 / 41863370|
+|5|SPANISH LUCENE SPANISH LIGHT STEM FILTER|PRIMARY_OUTPUT|1241848|146613|40621522|377860523152|146613 / 377860669765|40621522 / 41863370|
+|6|SPANISH LUCENE SPANISH PLURAL STEM FILTER|PRIMARY_OUTPUT|324656|57716|41538714|377860612049|57716 / 377860669765|41538714 / 41863370|
+|7|SPANISH LUCENE SPANISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|148226|47148|41715144|377860622617|47148 / 377860669765|41715144 / 41863370|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 377860669765 (0.000000%)|0 / 41863370 (0.000000%)|1.000000|1.000000|1.000000|
+|2|HUNSPELL SPANISH LUCENE FILTER|ANY_CANDIDATE|0.619928|412198 / 377860669765 (0.000109%)|31822108 / 41863370 (76.014205%)|0.600000|0.383864|0.479978|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+|2|HUNSPELL SPANISH LUCENE FILTER|ANY_CANDIDATE|0.960568|0.239858|0.999999|0.619928|0.999915|0.000085|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+|2|HUNSPELL SPANISH LUCENE FILTER|ANY_CANDIDATE|0.600000|0.383864|0.282205|0.237519|0.480000|0.479978|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL SPANISH LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|41863370|0|0|377860669765|0 / 377860669765|0 / 41863370|
+|2|HUNSPELL SPANISH LUCENE FILTER|ANY_CANDIDATE|10041262|412198|31822108|377860257567|412198 / 377860669765|31822108 / 41863370|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999998|1255381 / 377860669765 (0.000332%)|0 / 41863370 (0.000000%)|0.976572|0.985228|0.985334|
+|2|HUNSPELL SPANISH LUCENE FILTER|ALL_CANDIDATES|0.619928|878949 / 377860669765 (0.000233%)|31822108 / 41863370 (76.014205%)|0.586905|0.380469|0.469606|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.970885|1.000000|0.999997|0.999998|0.999997|0.000003|
+|2|HUNSPELL SPANISH LUCENE FILTER|ALL_CANDIDATES|0.919512|0.239858|0.999998|0.619928|0.999913|0.000087|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.976572|0.985228|0.994038|0.970885|0.985335|0.985334|
+|2|HUNSPELL SPANISH LUCENE FILTER|ALL_CANDIDATES|0.586905|0.380469|0.281467|0.234926|0.469630|0.469606|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|HUNSPELL SPANISH LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|41863370|1255381|0|377859414384|1255381 / 377860669765|0 / 41863370|
+|2|HUNSPELL SPANISH LUCENE FILTER|ALL_CANDIDATES|10041262|878949|31822108|377859790816|878949 / 377860669765|31822108 / 41863370|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|HUNSPELL SPANISH LUCENE FILTER|412747|118983|347768|17807|2.048262%|5|888962|
+|Radixor|885033|276044|979337|42403|4.877434%|21|914127|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `ES_ES`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/swedish.md
+++ b/docs/benchmarks/languages/swedish.md
@@ -0,0 +1,420 @@
+# Swedish Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Swedish. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `SV_SE` | 12,371 | 110,468 | 24,731 | 85,737 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **110,468**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 502 | 0.454% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 14,268 | 12.916% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 66,796 | 60.466% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 25,745 | 23.305% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 3,157 | 2.858% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 96.713% | 97.407% | 94.307% | Full Radixor dictionary patch-command stemmer. |
+| Lucene SwedishMinimalStemFilter | 49.532% | 49.186% | 50.730% | Minimal suffix reducer; narrow baseline, not a full stemmer. |
+| Lucene SwedishLightStemFilter | 45.672% | 46.383% | 43.209% | Light suffix stemmer; intentionally narrower than a dictionary-derived stemmer. |
+| Official Snowball direct | 40.068% | 37.512% | 48.926% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+| Lucene SnowballFilter | 38.785% | 35.839% | 48.999% | Lucene TokenFilter integration path around the Snowball algorithm. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `swedishRadixor` | 5.489 | 0.355 | 64.0 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene SwedishMinimalStemFilter | `swedishLuceneSwedishMinimalStemFilter` | 4.630 | 0.130 | 54.0 | 0.843 | Minimal Swedish suffix reducer. |
+| Lucene SwedishLightStemFilter | `swedishLuceneSwedishLightStemFilter` | 4.876 | 0.328 | 56.9 | 0.888 | Light Swedish suffix stemmer. |
+| Official Snowball direct | `snowballDirect[SWEDISH]` | 7.517 | 0.072 | 87.7 | 1.370 | Official Snowball generated Java stemmer; direct API. |
+| Lucene SnowballFilter | `luceneSnowballFilter[SWEDISH]` | 9.793 | 0.338 | 114.2 | 1.784 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `SV_SE` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/sv_se/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.974636** among 5 deterministic stemmers. The runner-up is `SNOWBALL SWEDISH DIRECT` at 0.807534, a difference of 0.167101. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.974584** among 5 deterministic stemmers. The runner-up is `SNOWBALL SWEDISH DIRECT` at 0.807599, a difference of 0.166985. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **7 result rows**, **5 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.974636|24473 / 4812155436 (0.000509%)|19546 / 385342 (5.072377%)|0.939665|0.943246|0.943260|
+|2|SNOWBALL SWEDISH DIRECT|PRIMARY_OUTPUT|0.807534|67105 / 4812155436 (0.001394%)|148325 / 385342 (38.491781%)|0.739832|0.687540|0.692339|
+|3|SNOWBALL SWEDISH LUCENE FILTER|PRIMARY_OUTPUT|0.799307|64262 / 4812155436 (0.001335%)|154666 / 385342 (40.137333%)|0.736940|0.678180|0.684227|
+|4|SWEDISH LUCENE SWEDISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.796072|40227 / 4812155436 (0.000836%)|157161 / 385342 (40.784809%)|0.781991|0.698068|0.709491|
+|5|SWEDISH LUCENE SWEDISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.783685|45941 / 4812155436 (0.000955%)|166707 / 385342 (43.262089%)|0.757232|0.672808|0.684713|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.937292|0.949276|0.999995|0.974636|0.999991|0.000009|
+|2|SNOWBALL SWEDISH DIRECT|PRIMARY_OUTPUT|0.779348|0.615082|0.999986|0.807534|0.999955|0.000045|
+|3|SNOWBALL SWEDISH LUCENE FILTER|PRIMARY_OUTPUT|0.782117|0.598627|0.999987|0.799307|0.999955|0.000045|
+|4|SWEDISH LUCENE SWEDISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.850127|0.592152|0.999992|0.796072|0.999959|0.000041|
+|5|SWEDISH LUCENE SWEDISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.826360|0.567379|0.999990|0.783685|0.999956|0.000044|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.939665|0.943246|0.946855|0.892588|0.943265|0.943260|
+|2|SNOWBALL SWEDISH DIRECT|PRIMARY_OUTPUT|0.739832|0.687540|0.642152|0.523856|0.692361|0.692339|
+|3|SNOWBALL SWEDISH LUCENE FILTER|PRIMARY_OUTPUT|0.736940|0.678180|0.628098|0.513065|0.684249|0.684227|
+|4|SWEDISH LUCENE SWEDISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.781991|0.698068|0.630412|0.536179|0.709510|0.709491|
+|5|SWEDISH LUCENE SWEDISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.757232|0.672808|0.605321|0.506941|0.684733|0.684713|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.943241|0.992631|0.993395|0.993013|0.993013|
+|2|SNOWBALL SWEDISH DIRECT|PRIMARY_OUTPUT|0.687518|0.984860|0.942685|0.963311|0.963311|
+|3|SNOWBALL SWEDISH LUCENE FILTER|PRIMARY_OUTPUT|0.678157|0.985207|0.939659|0.961894|0.961894|
+|4|SWEDISH LUCENE SWEDISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.698048|0.988493|0.944582|0.966038|0.966038|
+|5|SWEDISH LUCENE SWEDISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.672787|0.986795|0.942303|0.964036|0.964036|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|365796|24473|19546|4812130963|24473 / 4812155436|19546 / 385342|
+|2|SNOWBALL SWEDISH DIRECT|PRIMARY_OUTPUT|237017|67105|148325|4812088331|67105 / 4812155436|148325 / 385342|
+|3|SNOWBALL SWEDISH LUCENE FILTER|PRIMARY_OUTPUT|230676|64262|154666|4812091174|64262 / 4812155436|154666 / 385342|
+|4|SWEDISH LUCENE SWEDISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|228181|40227|157161|4812115209|40227 / 4812155436|157161 / 385342|
+|5|SWEDISH LUCENE SWEDISH LIGHT STEM FILTER|PRIMARY_OUTPUT|218635|45941|166707|4812109495|45941 / 4812155436|166707 / 385342|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 4812155436 (0.000000%)|0 / 385342 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|385342|0|0|4812155436|0 / 4812155436|0 / 385342|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999995|47848 / 4812155436 (0.000994%)|0 / 385342 (0.000000%)|0.909640|0.941544|0.943152|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.889545|1.000000|0.999990|0.999995|0.999990|0.000010|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.909640|0.941544|0.975768|0.889545|0.943157|0.943152|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|385342|47848|0|4812107588|47848 / 4812155436|0 / 385342|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|19546|24473|23375|5767|5.878216%|5|104148|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **7 result rows**, **5 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.974584|24473 / 4789911577 (0.000511%)|19546 / 384563 (5.082652%)|0.939544|0.943132|0.943146|
+|2|SNOWBALL SWEDISH DIRECT|PRIMARY_OUTPUT|0.807599|67105 / 4789911577 (0.001401%)|147975 / 384563 (38.478741%)|0.739645|0.687500|0.692274|
+|3|SNOWBALL SWEDISH LUCENE FILTER|PRIMARY_OUTPUT|0.799355|64262 / 4789911577 (0.001342%)|154316 / 384563 (40.127625%)|0.736744|0.678122|0.684143|
+|4|SWEDISH LUCENE SWEDISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.795947|40227 / 4789911577 (0.000840%)|156939 / 384563 (40.809698%)|0.781694|0.697790|0.709212|
+|5|SWEDISH LUCENE SWEDISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.783598|45941 / 4789911577 (0.000959%)|166437 / 384563 (43.279515%)|0.756945|0.672575|0.684469|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.937167|0.949173|0.999995|0.974584|0.999991|0.000009|
+|2|SNOWBALL SWEDISH DIRECT|PRIMARY_OUTPUT|0.779037|0.615213|0.999986|0.807599|0.999955|0.000045|
+|3|SNOWBALL SWEDISH LUCENE FILTER|PRIMARY_OUTPUT|0.781800|0.598724|0.999987|0.799355|0.999954|0.000046|
+|4|SWEDISH LUCENE SWEDISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.849816|0.591903|0.999992|0.795947|0.999959|0.000041|
+|5|SWEDISH LUCENE SWEDISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.826025|0.567205|0.999990|0.783598|0.999956|0.000044|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.939544|0.943132|0.946748|0.892384|0.943151|0.943146|
+|2|SNOWBALL SWEDISH DIRECT|PRIMARY_OUTPUT|0.739645|0.687500|0.642223|0.523810|0.692296|0.692274|
+|3|SNOWBALL SWEDISH LUCENE FILTER|PRIMARY_OUTPUT|0.736744|0.678122|0.628142|0.512999|0.684165|0.684143|
+|4|SWEDISH LUCENE SWEDISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.781694|0.697790|0.630152|0.535851|0.709231|0.709212|
+|5|SWEDISH LUCENE SWEDISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.756945|0.672575|0.605126|0.506676|0.684489|0.684469|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.943127|0.992612|0.993378|0.992995|0.992995|
+|2|SNOWBALL SWEDISH DIRECT|PRIMARY_OUTPUT|0.687478|0.984821|0.942695|0.963298|0.963298|
+|3|SNOWBALL SWEDISH LUCENE FILTER|PRIMARY_OUTPUT|0.678100|0.985169|0.939661|0.961877|0.961877|
+|4|SWEDISH LUCENE SWEDISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|0.697770|0.988463|0.944528|0.965996|0.965996|
+|5|SWEDISH LUCENE SWEDISH LIGHT STEM FILTER|PRIMARY_OUTPUT|0.672553|0.986761|0.942265|0.964000|0.964000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|365017|24473|19546|4789887104|24473 / 4789911577|19546 / 384563|
+|2|SNOWBALL SWEDISH DIRECT|PRIMARY_OUTPUT|236588|67105|147975|4789844472|67105 / 4789911577|147975 / 384563|
+|3|SNOWBALL SWEDISH LUCENE FILTER|PRIMARY_OUTPUT|230247|64262|154316|4789847315|64262 / 4789911577|154316 / 384563|
+|4|SWEDISH LUCENE SWEDISH MINIMAL STEM FILTER|PRIMARY_OUTPUT|227624|40227|156939|4789871350|40227 / 4789911577|156939 / 384563|
+|5|SWEDISH LUCENE SWEDISH LIGHT STEM FILTER|PRIMARY_OUTPUT|218126|45941|166437|4789865636|45941 / 4789911577|166437 / 384563|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 4789911577 (0.000000%)|0 / 384563 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|384563|0|0|4789911577|0 / 4789911577|0 / 384563|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999995|47848 / 4789911577 (0.000999%)|0 / 384563 (0.000000%)|0.909473|0.941433|0.943047|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.889346|1.000000|0.999990|0.999995|0.999990|0.000010|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.909473|0.941433|0.975720|0.889346|0.943051|0.943047|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|384563|47848|0|4789863729|47848 / 4789911577|0 / 384563|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|19546|24473|23375|5767|5.891848%|5|103921|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `SV_SE`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/ukrainian.md
+++ b/docs/benchmarks/languages/ukrainian.md
@@ -0,0 +1,474 @@
+# Ukrainian Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Ukrainian. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `UK_UA` | 1,493 | 15,737 | 2,985 | 12,752 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **15,737**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `AppendCharacterCommand` | Appends one character to the end of the word form. | 249 | 1.582% |
+| `BackwardCompoundCommand` | Applies a multi-step backward patch made from skip, delete, insert, and replace operations. | 4,160 | 26.435% |
+| `DeleteSuffixCommand` | Deletes one or more trailing characters from the word form. | 5,859 | 37.231% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 3,004 | 19.089% |
+| `ReplaceLastCharacterCommand` | Replaces the final character of the word form. | 2,465 | 15.664% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 99.307% | 99.365% | 99.062% | Full Radixor dictionary patch-command stemmer. |
+| Lucene HunspellStemFilter | 86.815% | 83.759% | 99.866% | Benchmark-only Ukrainian Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Lucene MorfologikFilter | 92.362% | 90.637% | 99.732% | Dictionary-based path; Morfologik can emit multiple terms. |
+| Morfologik direct | 92.362% | 90.637% | 99.732% | Direct dictionary lookup; first returned stem is used for quality when no ranking weight is exposed. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `ukrainianRadixor` | 0.682 | 0.057 | 53.5 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Lucene HunspellStemFilter | `luceneHunspellStemFilter` | 43.527 | 1.207 | 3413.3 | 63.799 | Benchmark-only Ukrainian Hunspell dictionary compared via Lucene HunspellStemFilter. |
+| Morfologik direct | `ukrainianMorfologikDirect` | 8.680 | 0.073 | 680.7 | 12.723 | Direct Morfologik dictionary lookup; first returned stem is used for quality. |
+| Lucene MorfologikFilter | `ukrainianLuceneMorfologikFilter` | 14.575 | 0.248 | 1143.0 | 21.364 | Dictionary-based Morfologik TokenFilter; may emit multiple terms. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `UK_UA` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/uk_ua/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.995343** among 4 deterministic stemmers. The runner-up is `UKRAINIAN LUCENE MORFOLOGIK FILTER` at 0.928768, a difference of 0.066575. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.995342** among 4 deterministic stemmers. The runner-up is `UKRAINIAN LUCENE MORFOLOGIK FILTER` at 0.928751, a difference of 0.066591. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **12 result rows**, **4 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.995343|880 / 101387550 (0.000868%)|608 / 65340 (0.930517%)|0.987406|0.988637|0.988632|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.928768|828 / 101387550 (0.000817%)|9308 / 65340 (14.245485%)|0.956896|0.917054|0.919223|
+|3|UKRAINIAN MORFOLOGIK DIRECT|PRIMARY_OUTPUT|0.928646|828 / 101387550 (0.000817%)|9324 / 65340 (14.269972%)|0.956832|0.916912|0.919090|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|PRIMARY_OUTPUT|0.885793|794 / 101387550 (0.000783%)|14924 / 65340 (22.840526%)|0.933008|0.865139|0.871499|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.986588|0.990695|0.999991|0.995343|0.999985|0.000015|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.985438|0.857545|0.999992|0.928768|0.999900|0.000100|
+|3|UKRAINIAN MORFOLOGIK DIRECT|PRIMARY_OUTPUT|0.985434|0.857300|0.999992|0.928646|0.999900|0.000100|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|PRIMARY_OUTPUT|0.984495|0.771595|0.999992|0.885793|0.999845|0.000155|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.987406|0.988637|0.989871|0.977529|0.988639|0.988632|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.956896|0.917054|0.880397|0.846814|0.919270|0.919223|
+|3|UKRAINIAN MORFOLOGIK DIRECT|PRIMARY_OUTPUT|0.956832|0.916912|0.880190|0.846572|0.919137|0.919090|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|PRIMARY_OUTPUT|0.933008|0.865139|0.806475|0.762331|0.871568|0.871499|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.988630|0.997994|0.998266|0.998130|0.998130|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.917004|0.997990|0.971000|0.984310|0.984310|
+|3|UKRAINIAN MORFOLOGIK DIRECT|PRIMARY_OUTPUT|0.916862|0.997990|0.970876|0.984246|0.984246|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|PRIMARY_OUTPUT|0.865063|0.998114|0.949804|0.973360|0.973360|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|64732|880|608|101386670|880 / 101387550|608 / 65340|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|56032|828|9308|101386722|828 / 101387550|9308 / 65340|
+|3|UKRAINIAN MORFOLOGIK DIRECT|PRIMARY_OUTPUT|56016|828|9324|101386722|828 / 101387550|9324 / 65340|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|PRIMARY_OUTPUT|50416|794|14924|101386756|794 / 101387550|14924 / 65340|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 101387550 (0.000000%)|0 / 65340 (0.000000%)|1.000000|1.000000|1.000000|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|0.962151|122 / 101387550 (0.000120%)|4946 / 65340 (7.569636%)|0.982323|0.959732|0.960413|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ANY_CANDIDATE|0.962029|122 / 101387550 (0.000120%)|4962 / 65340 (7.594123%)|0.982267|0.959599|0.960286|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ANY_CANDIDATE|0.927570|326 / 101387550 (0.000322%)|9465 / 65340 (14.485767%)|0.962884|0.919443|0.922008|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|0.997984|0.924304|0.999999|0.962151|0.999950|0.000050|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ANY_CANDIDATE|0.997983|0.924059|0.999999|0.962029|0.999950|0.000050|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ANY_CANDIDATE|0.994199|0.855142|0.999997|0.927570|0.999903|0.000097|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|0.982323|0.959732|0.938156|0.922581|0.960438|0.960413|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ANY_CANDIDATE|0.982267|0.959599|0.937954|0.922337|0.960310|0.960286|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ANY_CANDIDATE|0.962884|0.919443|0.879752|0.850897|0.922053|0.922008|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|65340|0|0|101387550|0 / 101387550|0 / 65340|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|60394|122|4946|101387428|122 / 101387550|4946 / 65340|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ANY_CANDIDATE|60378|122|4962|101387428|122 / 101387550|4962 / 65340|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ANY_CANDIDATE|55875|326|9465|101387224|326 / 101387550|9465 / 65340|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999993|1490 / 101387550 (0.001470%)|0 / 65340 (0.000000%)|0.982084|0.988727|0.988782|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|0.962145|1368 / 101387550 (0.001349%)|4946 / 65340 (7.569636%)|0.966650|0.950323|0.950669|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ALL_CANDIDATES|0.962023|1368 / 101387550 (0.001349%)|4962 / 65340 (7.594123%)|0.966592|0.950191|0.950541|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ALL_CANDIDATES|0.927565|1271 / 101387550 (0.001254%)|9465 / 65340 (14.485767%)|0.950501|0.912349|0.914347|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.977705|1.000000|0.999985|0.999993|0.999985|0.000015|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|0.977850|0.924304|0.999987|0.962145|0.999938|0.000062|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ALL_CANDIDATES|0.977845|0.924059|0.999987|0.962023|0.999938|0.000062|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ALL_CANDIDATES|0.977759|0.855142|0.999987|0.927565|0.999894|0.000106|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.982084|0.988727|0.995460|0.977705|0.988789|0.988782|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|0.966650|0.950323|0.934539|0.905349|0.950700|0.950669|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ALL_CANDIDATES|0.966592|0.950191|0.934337|0.905109|0.950571|0.950541|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ALL_CANDIDATES|0.950501|0.912349|0.877142|0.838825|0.914398|0.914347|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|65340|1490|0|101386060|1490 / 101387550|0 / 65340|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|60394|1368|4946|101386182|1368 / 101387550|4946 / 65340|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ALL_CANDIDATES|60378|1368|4962|101386182|1368 / 101387550|4962 / 65340|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ALL_CANDIDATES|55875|1271|9465|101386279|1271 / 101387550|9465 / 65340|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|HUNSPELL UKRAINIAN LUCENE FILTER|5459|468|477|1322|9.280449%|6|15740|
+|UKRAINIAN LUCENE MORFOLOGIK FILTER|4362|706|540|2207|15.493155%|6|16937|
+|UKRAINIAN MORFOLOGIK DIRECT|4362|706|540|2207|15.493155%|6|16937|
+|Radixor|608|880|610|190|1.333801%|2|14435|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **12 result rows**, **4 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.995342|880 / 101259406 (0.000869%)|608 / 65324 (0.930745%)|0.987403|0.988634|0.988629|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.928751|828 / 101259406 (0.000818%)|9308 / 65324 (14.248974%)|0.956884|0.917032|0.919202|
+|3|UKRAINIAN MORFOLOGIK DIRECT|PRIMARY_OUTPUT|0.928751|828 / 101259406 (0.000818%)|9308 / 65324 (14.248974%)|0.956884|0.917032|0.919202|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|PRIMARY_OUTPUT|0.885796|794 / 101259406 (0.000784%)|14920 / 65324 (22.839998%)|0.933007|0.865141|0.871500|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.986585|0.990693|0.999991|0.995342|0.999985|0.000015|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.985434|0.857510|0.999992|0.928751|0.999900|0.000100|
+|3|UKRAINIAN MORFOLOGIK DIRECT|PRIMARY_OUTPUT|0.985434|0.857510|0.999992|0.928751|0.999900|0.000100|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|PRIMARY_OUTPUT|0.984492|0.771600|0.999992|0.885796|0.999845|0.000155|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.987403|0.988634|0.989868|0.977524|0.988636|0.988629|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.956884|0.917032|0.880367|0.846777|0.919249|0.919202|
+|3|UKRAINIAN MORFOLOGIK DIRECT|PRIMARY_OUTPUT|0.956884|0.917032|0.880367|0.846777|0.919249|0.919202|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|PRIMARY_OUTPUT|0.933007|0.865141|0.806479|0.762334|0.871570|0.871500|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.988627|0.997992|0.998264|0.998128|0.998128|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|0.916982|0.997988|0.970978|0.984298|0.984298|
+|3|UKRAINIAN MORFOLOGIK DIRECT|PRIMARY_OUTPUT|0.916982|0.997988|0.970978|0.984298|0.984298|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|PRIMARY_OUTPUT|0.865065|0.998113|0.949788|0.973351|0.973351|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|64716|880|608|101258526|880 / 101259406|608 / 65324|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|PRIMARY_OUTPUT|56016|828|9308|101258578|828 / 101259406|9308 / 65324|
+|3|UKRAINIAN MORFOLOGIK DIRECT|PRIMARY_OUTPUT|56016|828|9308|101258578|828 / 101259406|9308 / 65324|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|PRIMARY_OUTPUT|50404|794|14920|101258612|794 / 101259406|14920 / 65324|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 101259406 (0.000000%)|0 / 65324 (0.000000%)|1.000000|1.000000|1.000000|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|0.962142|122 / 101259406 (0.000120%)|4946 / 65324 (7.571490%)|0.982318|0.959722|0.960404|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ANY_CANDIDATE|0.962142|122 / 101259406 (0.000120%)|4946 / 65324 (7.571490%)|0.982318|0.959722|0.960404|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ANY_CANDIDATE|0.927552|326 / 101259406 (0.000322%)|9465 / 65324 (14.489315%)|0.962874|0.919422|0.921988|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|0.997983|0.924285|0.999999|0.962142|0.999950|0.000050|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ANY_CANDIDATE|0.997983|0.924285|0.999999|0.962142|0.999950|0.000050|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ANY_CANDIDATE|0.994198|0.855107|0.999997|0.927552|0.999903|0.000097|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|0.982318|0.959722|0.938141|0.922562|0.960428|0.960404|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ANY_CANDIDATE|0.982318|0.959722|0.938141|0.922562|0.960428|0.960404|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ANY_CANDIDATE|0.962874|0.919422|0.879722|0.850861|0.922033|0.921988|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|65324|0|0|101259406|0 / 101259406|0 / 65324|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ANY_CANDIDATE|60378|122|4946|101259284|122 / 101259406|4946 / 65324|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ANY_CANDIDATE|60378|122|4946|101259284|122 / 101259406|4946 / 65324|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ANY_CANDIDATE|55859|326|9465|101259080|326 / 101259406|9465 / 65324|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999993|1490 / 101259406 (0.001471%)|0 / 65324 (0.000000%)|0.982079|0.988724|0.988779|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|0.962136|1368 / 101259406 (0.001351%)|4946 / 65324 (7.571490%)|0.966642|0.950311|0.950657|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ALL_CANDIDATES|0.962136|1368 / 101259406 (0.001351%)|4946 / 65324 (7.571490%)|0.966642|0.950311|0.950657|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ALL_CANDIDATES|0.927547|1271 / 101259406 (0.001255%)|9465 / 65324 (14.489315%)|0.950487|0.912326|0.914325|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.977699|1.000000|0.999985|0.999993|0.999985|0.000015|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|0.977845|0.924285|0.999986|0.962136|0.999938|0.000062|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ALL_CANDIDATES|0.977845|0.924285|0.999986|0.962136|0.999938|0.000062|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ALL_CANDIDATES|0.977752|0.855107|0.999987|0.927547|0.999894|0.000106|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.982079|0.988724|0.995459|0.977699|0.988787|0.988779|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|0.966642|0.950311|0.934522|0.905326|0.950688|0.950657|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ALL_CANDIDATES|0.966642|0.950311|0.934522|0.905326|0.950688|0.950657|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ALL_CANDIDATES|0.950487|0.912326|0.877111|0.838787|0.914376|0.914325|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|65324|1490|0|101257916|1490 / 101259406|0 / 65324|
+|2|UKRAINIAN LUCENE MORFOLOGIK FILTER|ALL_CANDIDATES|60378|1368|4946|101258038|1368 / 101259406|4946 / 65324|
+|3|UKRAINIAN MORFOLOGIK DIRECT|ALL_CANDIDATES|60378|1368|4946|101258038|1368 / 101259406|4946 / 65324|
+|4|HUNSPELL UKRAINIAN LUCENE FILTER|ALL_CANDIDATES|55859|1271|9465|101258135|1271 / 101259406|9465 / 65324|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|HUNSPELL UKRAINIAN LUCENE FILTER|5455|468|477|1321|9.279292%|6|15730|
+|UKRAINIAN LUCENE MORFOLOGIK FILTER|4362|706|540|2207|15.502950%|6|16928|
+|UKRAINIAN MORFOLOGIK DIRECT|4362|706|540|2207|15.502950%|6|16928|
+|Radixor|608|880|610|190|1.334645%|2|14426|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `UK_UA`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/languages/yiddish.md
+++ b/docs/benchmarks/languages/yiddish.md
@@ -0,0 +1,395 @@
+# Yiddish Stemmer Benchmarks
+
+This page reports same-language stemming benchmarks for Yiddish. Accuracy is listed first because speed without root agreement is not enough to interpret stemmer quality.
+
+All speed values are environment-specific and were measured on the hardware and JVM listed in the [benchmark overview](../index.md). Speed benchmark operations process changed dictionary tokens only. Accuracy uses the complete Radixor dictionary for the language.
+
+Radixor must not be read as simply "slower" when a narrow competitor has a lower timing row. In these tables Radixor is the quality-oriented baseline: its exact-root accuracy is typically close to 100%, while many faster rule-based, light, minimal, or possessive filters reach that speed by doing much less linguistic work and often score far lower in `All exact` and `Changed exact`. The Radixor rows in this benchmark refresh use the contracted compiled patch trie: compilation collapses uniform patch-command subtrees into accepting leaves, reducing hot lookup depth while preserving the preferred stemming result measured by the accuracy pass. The [EnglishRadixorDictionaryCoverageBenchmark](../reference/english-coverage.md) table shows the resulting quality/speed envelope explicitly. The same interpretation applies to this language page: speed rows must be read together with the accuracy table above them.
+
+## Dictionary Corpus
+
+| Resource | Dictionary rows | Complete quality tokens | Already-root tokens | Changed speed tokens |
+| --- | ---: | ---: | ---: | ---: |
+| `YI` | 802 | 4,300 | 1,524 | 2,776 |
+
+## Radixor Patch Command Distribution
+
+Radixor stores the preferred transformation for each normalized dictionary word form as a compiled patch command. This distribution shows which runtime command class is selected by the trained trie for the complete language dictionary. The total number of preferred patch commands analyzed for this language is **4,300**.
+
+| Command class | Meaning | Word forms | Share |
+| --- | --- | ---: | ---: |
+| `DeletePrefixCommand` | Deletes one or more leading characters from the word form in forward traversal. | 25 | 0.581% |
+| `ForwardCompoundCommand` | Applies a multi-step forward patch made from skip, delete, insert, and replace operations. | 2,721 | 63.279% |
+| `PreserveCommand` | Returns the word form unchanged because it already matches the preferred root. | 1,551 | 36.070% |
+| `ReplaceFirstCharacterCommand` | Replaces the first character of the word form in forward traversal. | 3 | 0.070% |
+
+## Accuracy
+
+Accuracy is computed from JMH auxiliary counters in the current report. The counters are deterministic for a fixed corpus and stemmer; percentages divide matching counters by evaluated counters from the same report and are not timing metrics.
+
+| Stemmer | All exact | Changed exact | Root preserved | Note |
+| --- | ---: | ---: | ---: | --- |
+| Radixor | 98.930% | 98.343% | 100.000% | Full Radixor dictionary patch-command stemmer. |
+| Lucene SnowballFilter | 2.837% | 2.558% | 3.346% | Lucene TokenFilter integration path around the Snowball algorithm. |
+| Official Snowball direct | 2.837% | 2.558% | 3.346% | Official Snowball generated Java stemmer; rule-based suffix algorithm. |
+
+## Speed
+
+Speed uses JMH average time, 3 warmup iterations, 5 measurement iterations, 1 fork, and 1 thread. Relative factor is computed against the single Radixor row on this language page. Values below 1.000 are faster than that Radixor baseline; values above 1.000 are slower.
+
+| Stemmer | Benchmark method | Score ms/op | Error ms | ns/token | Relative vs Radixor | Note |
+| --- | --- | ---: | ---: | ---: | ---: | --- |
+| Radixor | `radixor[YIDDISH]` | 0.254 | 0.004 | 50.7 | 1.000 | Full Radixor dictionary patch-command stemmer. |
+| Official Snowball direct | `snowballDirect[YIDDISH]` | 1.537 | 0.220 | 307.3 | 6.058 | Official Snowball generated Java stemmer; direct API. |
+| Lucene SnowballFilter | `luceneSnowballFilter[YIDDISH]` | 1.714 | 0.120 | 342.8 | 6.756 | Lucene TokenFilter path around Snowball; includes TokenStream overhead. |
+
+## Interpretation Notes
+
+- Radixor is a dictionary-derived patch-command stemmer. Its quality depends on the language resource used to train the compiled trie.
+- Light, minimal, plural, and possessive filters are narrow baselines. They can be fast because they intentionally perform less linguistic work.
+- Lucene TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+- Morfologik rows are dictionary-based and can emit multiple terms for one input token. Quality rows use the first returned term when no ranking weight is available.
+- Snowball rows are rule-based generated suffix stemmers; they are useful algorithmic baselines, not dictionary-root equivalence guarantees.
+
+<!-- STEMMING-QUALITY:START -->
+
+## Stemming Quality
+
+Runtime performance and linguistic grouping quality are independent dimensions. This section evaluates language `YI` using the complete validated stemming-quality result matrix. Every usable dictionary row is one gold-standard group of forms expected to share a morphological family or lemma. Exact equality with a predetermined lemma is not required. Same-row pairs are positive pairs; pairs from different rows are negative pairs.
+
+`ALL_WORDS` includes every valid group and its original forms. `LOWERCASE_GROUPS_ONLY` excludes an entire group when any Unicode code point is uppercase or titlecase; retained words are not lowercased or otherwise rewritten. This isolates case-handling effects without changing retained inputs. [Download the complete machine-readable result snapshot](../data/stemming-quality.csv).
+
+### Evaluation Scope and Key Findings
+
+The dictionary resource is `src/main/resources/yi/stemmer.gz`. The following findings compare only deterministic `PRIMARY_OUTPUT` rows over identical included groups; candidate policies are reported separately as capability analyses.
+
+- **ALL_WORDS:** `Radixor` ranks first by balanced accuracy at **0.988241** among 3 deterministic stemmers. The runner-up is `SNOWBALL YIDDISH DIRECT` at 0.890988, a difference of 0.097253. This rank does not imply leadership in throughput or every secondary metric.
+- **LOWERCASE_GROUPS_ONLY:** `Radixor` ranks first by balanced accuracy at **0.988241** among 3 deterministic stemmers. The runner-up is `SNOWBALL YIDDISH DIRECT` at 0.890988, a difference of 0.097253. This rank does not imply leadership in throughput or every secondary metric.
+### `ALL_WORDS`
+
+This mode contains **5 result rows**, **3 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.988241|195 / 6392909 (0.003050%)|149 / 6344 (2.348676%)|0.970881|0.972986|0.972965|
+|2|SNOWBALL YIDDISH DIRECT|PRIMARY_OUTPUT|0.890988|1151 / 6392909 (0.018004%)|1382 / 6344 (21.784363%)|0.805624|0.796661|0.796600|
+|3|SNOWBALL YIDDISH LUCENE FILTER|PRIMARY_OUTPUT|0.890988|1151 / 6392909 (0.018004%)|1382 / 6344 (21.784363%)|0.805624|0.796661|0.796600|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.969484|0.976513|0.999969|0.988241|0.999946|0.000054|
+|2|SNOWBALL YIDDISH DIRECT|PRIMARY_OUTPUT|0.811713|0.782156|0.999820|0.890988|0.999604|0.000396|
+|3|SNOWBALL YIDDISH LUCENE FILTER|PRIMARY_OUTPUT|0.811713|0.782156|0.999820|0.890988|0.999604|0.000396|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.970881|0.972986|0.975099|0.947393|0.972992|0.972965|
+|2|SNOWBALL YIDDISH DIRECT|PRIMARY_OUTPUT|0.805624|0.796661|0.787894|0.662041|0.796798|0.796600|
+|3|SNOWBALL YIDDISH LUCENE FILTER|PRIMARY_OUTPUT|0.805624|0.796661|0.787894|0.662041|0.796798|0.796600|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.972959|0.995691|0.996142|0.995917|0.995917|
+|2|SNOWBALL YIDDISH DIRECT|PRIMARY_OUTPUT|0.796462|0.982919|0.962014|0.972354|0.972354|
+|3|SNOWBALL YIDDISH LUCENE FILTER|PRIMARY_OUTPUT|0.796462|0.982919|0.962014|0.972354|0.972354|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|6195|195|149|6392714|195 / 6392909|149 / 6344|
+|2|SNOWBALL YIDDISH DIRECT|PRIMARY_OUTPUT|4962|1151|1382|6391758|1151 / 6392909|1382 / 6344|
+|3|SNOWBALL YIDDISH LUCENE FILTER|PRIMARY_OUTPUT|4962|1151|1382|6391758|1151 / 6392909|1382 / 6344|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 6392909 (0.000000%)|0 / 6344 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|6344|0|0|6392909|0 / 6392909|0 / 6344|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999970|389 / 6392909 (0.006085%)|0 / 6344 (0.000000%)|0.953240|0.970253|0.970653|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.942225|1.000000|0.999939|0.999970|0.999939|0.000061|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.953240|0.970253|0.987885|0.942225|0.970683|0.970653|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|6344|389|0|6392520|389 / 6392909|0 / 6344|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|149|195|194|89|2.487423%|3|3676|
+
+### `LOWERCASE_GROUPS_ONLY`
+
+This mode contains **5 result rows**, **3 evaluated stemmers**, and **3 output policies**. Applied-row and form counts are shown per row because adapters share the language corpus but policy rows remain independently auditable. Rankings are separated by output policy and ordered by unrounded balanced accuracy, followed by MCC, F1, over-stemming rate, over-stemming count, under-stemming rate, and stemmer. Balanced accuracy is a navigation metric, not a universally authoritative quality score.
+
+#### `PRIMARY_OUTPUT` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.988241|195 / 6392909 (0.003050%)|149 / 6344 (2.348676%)|0.970881|0.972986|0.972965|
+|2|SNOWBALL YIDDISH DIRECT|PRIMARY_OUTPUT|0.890988|1151 / 6392909 (0.018004%)|1382 / 6344 (21.784363%)|0.805624|0.796661|0.796600|
+|3|SNOWBALL YIDDISH LUCENE FILTER|PRIMARY_OUTPUT|0.890988|1151 / 6392909 (0.018004%)|1382 / 6344 (21.784363%)|0.805624|0.796661|0.796600|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.969484|0.976513|0.999969|0.988241|0.999946|0.000054|
+|2|SNOWBALL YIDDISH DIRECT|PRIMARY_OUTPUT|0.811713|0.782156|0.999820|0.890988|0.999604|0.000396|
+|3|SNOWBALL YIDDISH LUCENE FILTER|PRIMARY_OUTPUT|0.811713|0.782156|0.999820|0.890988|0.999604|0.000396|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.970881|0.972986|0.975099|0.947393|0.972992|0.972965|
+|2|SNOWBALL YIDDISH DIRECT|PRIMARY_OUTPUT|0.805624|0.796661|0.787894|0.662041|0.796798|0.796600|
+|3|SNOWBALL YIDDISH LUCENE FILTER|PRIMARY_OUTPUT|0.805624|0.796661|0.787894|0.662041|0.796798|0.796600|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|0.972959|0.995691|0.996142|0.995917|0.995917|
+|2|SNOWBALL YIDDISH DIRECT|PRIMARY_OUTPUT|0.796462|0.982919|0.962014|0.972354|0.972354|
+|3|SNOWBALL YIDDISH LUCENE FILTER|PRIMARY_OUTPUT|0.796462|0.982919|0.962014|0.972354|0.972354|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|PRIMARY_OUTPUT|6195|195|149|6392714|195 / 6392909|149 / 6344|
+|2|SNOWBALL YIDDISH DIRECT|PRIMARY_OUTPUT|4962|1151|1382|6391758|1151 / 6392909|1382 / 6344|
+|3|SNOWBALL YIDDISH LUCENE FILTER|PRIMARY_OUTPUT|4962|1151|1382|6391758|1151 / 6392909|1382 / 6344|
+
+</details>
+
+#### `ANY_CANDIDATE` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|0 / 6392909 (0.000000%)|0 / 6344 (0.000000%)|1.000000|1.000000|1.000000|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|0.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|1.000000|1.000000|1.000000|1.000000|1.000000|1.000000|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ANY_CANDIDATE|6344|0|0|6392909|0 / 6392909|0 / 6344|
+
+</details>
+
+#### `ALL_CANDIDATES` ranking
+
+<div class="quality-table quality-table--compact" role="region" aria-label="Compact stemming-quality ranking; scroll horizontally for additional columns" tabindex="0" markdown="1">
+
+| Rank | Stemmer | Output policy | Balanced accuracy | Over-stemming | Under-stemming | F0.5 | F1 | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.999970|389 / 6392909 (0.006085%)|0 / 6344 (0.000000%)|0.953240|0.970253|0.970653|
+
+</div>
+
+<details class="quality-details" markdown="1"><summary>Classification metrics</summary>
+
+| Rank | Stemmer | Output policy | Precision | Recall | Specificity | Balanced accuracy | Pairwise accuracy | Error rate |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.942225|1.000000|0.999939|0.999970|0.999939|0.000061|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Pair-relation metrics</summary>
+
+| Rank | Stemmer | Output policy | F0.5 | F1 | F2 | Jaccard | Fowlkes–Mallows | MCC |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|0.953240|0.970253|0.987885|0.942225|0.970683|0.970653|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Partition metrics (PRIMARY_OUTPUT only)</summary>
+
+| Rank | Stemmer | Output policy | Adjusted Rand Index | Homogeneity | Completeness | V-measure | Normalized mutual information |
+|---:|---|---|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|n/a|n/a|n/a|n/a|n/a|
+
+</details>
+
+<details class="quality-details" markdown="1"><summary>Raw pair counts</summary>
+
+| Rank | Stemmer | Output policy | TP | FP | FN | TN | Over error / possible | Under error / possible |
+|---:|---|---|---:|---:|---:|---:|---:|---:|
+|1|Radixor|ALL_CANDIDATES|6344|389|0|6392520|389 / 6392909|0 / 6344|
+
+</details>
+
+#### Multi-output analysis
+
+Alternative candidates are capability analyses, not replacements for the deterministic comparison.
+
+| Stemmer | Under pairs repaired | Best-case over pairs avoided | All-candidate collisions added | Multi-candidate forms | Multi-candidate share | Maximum candidates | Total candidate assignments |
+|---|---:|---:|---:|---:|---:|---:|---:|
+|Radixor|149|195|194|89|2.487423%|3|3676|
+
+### Output Policies and Metric Definitions
+
+`PRIMARY_OUTPUT` uses one deterministic stem per form and therefore defines a strict partition. `ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound: a same-group pair succeeds when candidates intersect, while a different-group pair succeeds when a non-colliding selection exists. Candidate choices may differ between pairs, so this is not deterministic runtime behaviour and need not represent one globally consistent assignment. `ALL_CANDIDATES` activates every returned candidate; forms are related when candidate sets intersect. Alternatives can reduce under-stemming but can introduce cross-group collisions, and the resulting relation can overlap and need not be a partition.
+
+For each row, `TP = underPossiblePairs - underErrorPairs`, `FN = underErrorPairs`, `FP = overErrorPairs`, and `TN = overPossiblePairs - overErrorPairs`. TP and FN concern same-group pairs; FP and TN concern different-group pairs. Consequently, under-stemming and over-stemming use different denominators. Undefined values are rendered as `n/a`.
+
+- Under-stemming rate: `FN / (TP + FN)`, the false-negative rate over same-group pairs.
+- Over-stemming rate: `FP / (TN + FP)`, the false-positive rate over different-group pairs.
+- Pairwise precision: `TP / (TP + FP)`, the fraction of predicted conflations that are gold-standard positive pairs.
+- Pairwise recall: `TP / (TP + FN)`, the fraction of gold-standard positive pairs successfully connected.
+- Pairwise specificity: `TN / (TN + FP)`, the fraction of different-group pairs correctly separated.
+- Balanced accuracy: `(recall + specificity) / 2`. It gives equal weight to positive and negative pair classes and is less dominated by the large true-negative class than ordinary accuracy. It does not replace the raw errors or other metrics.
+- Pairwise F-beta: `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`. F0.5 emphasizes precision and penalizes over-stemming more; F1 weights precision and recall equally; F2 emphasizes recall and penalizes under-stemming more.
+- MCC: `(TP * TN - FP * FN) / sqrt((TP + FP) * (TP + FN) * (TN + FP) * (TN + FN))`. It uses all confusion counts and remains useful under class imbalance, except when its denominator is degenerate.
+- Jaccard index: `TP / (TP + FP + FN)`.
+- Fowlkes–Mallows index: `sqrt(precision * recall)`.
+- Pairwise accuracy: `(TP + TN) / (TP + TN + FP + FN)`. It can be dominated by true-negative cross-group pairs.
+- Pairwise error rate: `(FP + FN) / (TP + TN + FP + FN)`.
+
+Adjusted Rand Index uses the gold/predicted contingency table and chance correction. Homogeneity is `1 - H(gold | predicted) / H(gold)`; completeness is `1 - H(predicted | gold) / H(predicted)`; V-measure is their harmonic mean; normalized mutual information uses the arithmetic-mean entropy normalization `MI / ((H(gold) + H(predicted)) / 2)`. These partition-only metrics apply to `PRIMARY_OUTPUT`; candidate-relation rows show `n/a`.
+
+### Provenance
+
+- Authoritative source: `docs/benchmarks/data/stemming-quality.csv`
+- Source SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Evaluation command: `./gradlew stemmingQuality`
+- Dictionary language: `YI`
+- Processing modes: `ALL_WORDS`, `LOWERCASE_GROUPS_ONLY`
+- Stemmer versions and transitive artifacts: resolved by the repository's JMH Gradle configuration and `gradle.lockfile`
+- Radixor version, Git revision, generation date, JDK version, operating system, and dictionary revision: not recorded in the authoritative CSV
+
+<!-- STEMMING-QUALITY:END -->
--- a/docs/benchmarks/reference/candidates.md
+++ b/docs/benchmarks/reference/candidates.md
@@ -0,0 +1,31 @@
+# Benchmark Candidate Evaluation
+
+Implemented benchmark methods are documented on the per-language pages under [Language Benchmark Pages](../languages/index.md). This keeps the exact method names, implementation descriptions, accuracy tables, and speed tables close to the language where they are valid.
+
+## Included Candidate Families
+
+The current benchmark pages include Radixor baselines, Lucene language filters where the language matches a bundled Radixor resource, Lucene Stempel and Morfologik paths where applicable, official Snowball Java stemmers where same-language comparison is available, benchmark-only CISTEM German stemmer evaluation, benchmark-only Hunspell comparisons, and selected English-specific non-Lucene baselines such as OpenNLP Porter and Paice/Husk Lancaster.
+
+Benchmark-only Hunspell comparisons use bundled benchmark dictionaries and the Lucene HunspellStemFilter adapter over the selected language token streams.
+The CISTEM candidate is implemented in `src/jmh/java/org/egothor/stemmer/benchmark/Cistem.java` and follows the original MIT-licensed upstream implementation from Leonie Weissweiler's CISTEM project.
+CISTEM German gold-standard files are not vendored in this repository. The Gradle JMH resource preparation tasks download `goldstandard1.txt` and `goldstandard2.txt` from the upstream CISTEM repository into generated build resources.
+
+Direct stemmer APIs and Lucene TokenFilter paths are documented separately on language pages. TokenFilter rows include TokenStream, attribute, and required normalization overhead. Direct rows measure exposed direct APIs.
+
+For the benchmark refresh used in this documentation build:
+
+- Hunspell families are included in `HunspellStemmerComparisonBenchmark` (speed) and `HunspellStemmerComparisonBenchmarkQuality` (quality for all benchmark languages in this corpus). The legacy
+  `EnglishHunspellStemmerComparisonBenchmarkQuality` result is retained for continuity.
+- CISTEM quality is present in the published per-language results under `GERMAN_CISTEM`. CISTEM speed is present as `germanCistem` in `MultiLanguageStemmerComparisonBenchmark`.
+
+## Evaluated But Skipped Candidates
+
+| Candidate | Language | Link/source | Reason skipped |
+| --- | --- | --- | --- |
+| Lucene Arabic, Bulgarian, Bengali, Sorani, Greek, Galician, Hindi, Indonesian, Latvian, Telugu filters | Various | `lucene-analysis-common` | No bundled same-language Radixor resource in this repository snapshot. |
+| Lucene analyzer-only paths | Multiple | Lucene analyzers | Full analyzers mix tokenization, stop-word handling, and other behavior; direct filters are used where available. |
+| Lucene StemmerOverrideFilter | Multiple | `lucene-analysis-common` | Override map facility, not a stemmer algorithm. |
+| Additional Snowball Lovins | English | Official Snowball Java distribution | No Lovins Java stemmer was present in the selected Snowball Java distribution. |
+| Lemur Project Krovetz Stemmer | English | Lemur Project | Lucene KStem represents the Krovetz-style path without adding separate dependency and license risk. |
+| Smile Lancaster / Paice-Husk | English | Smile NLP | Smile is large for one stemmer; Paice/Husk is included through a smaller benchmark-only generated path. |
+| `stemmerEval` reference repository | Multiple | `https://github.com/endredy/stemmerEval` | Used only as a candidate reference; no code or data copied. |
--- a/docs/benchmarks/reference/corpora.md
+++ b/docs/benchmarks/reference/corpora.md
@@ -0,0 +1,28 @@
+# Benchmark Corpora
+
+The table below describes the Radixor resources used to build speed and quality corpora. `Total tokens` is the complete dictionary token count used by quality benchmarks. `Already-root tokens` counts fields where the token is already equal to the line root. `Changed tokens` is the speed workload before the minimum-size repeat rule.
+
+| Language resource | Dictionary rows | Total tokens | Already-root tokens | Changed tokens | Speed timing tokens |
+| --- | ---: | ---: | ---: | ---: | ---: |
+| `cs_cz` | 5,113 | 56,612 | 10,049 | 46,563 | 46,563 |
+| `da_dk` | 4,179 | 32,256 | 8,356 | 23,900 | 23,900 |
+| `de_de` | 39,315 | 213,440 | 73,799 | 139,641 | 139,641 |
+| `es_es` | 65,059 | 926,393 | 120,121 | 806,272 | 806,272 |
+| `fa_ir` | 69 | 3,770 | 138 | 3,632 | 5,000 |
+| `fi_fi` | 57,027 | 1,865,215 | 110,525 | 1,754,690 | 1,754,690 |
+| `fr_fr` | 59,240 | 474,110 | 108,141 | 365,969 | 365,969 |
+| `he_il` | 2,358 | 61,071 | 4,715 | 56,356 | 56,356 |
+| `hu_hu` | 19,406 | 935,713 | 38,775 | 896,938 | 896,938 |
+| `it_it` | 10,009 | 337,546 | 20,004 | 317,542 | 317,542 |
+| `nb_no` | 17,929 | 90,757 | 33,376 | 57,381 | 57,381 |
+| `nl_nl` | 4,992 | 31,466 | 9,981 | 21,485 | 21,485 |
+| `nn_no` | 4,688 | 19,651 | 6,089 | 13,562 | 13,562 |
+| `pl_pl` | 9,990 | 132,308 | 19,957 | 112,351 | 112,351 |
+| `pt_pt` | 4,001 | 215,490 | 8,002 | 207,488 | 207,488 |
+| `ru_ru` | 37,410 | 806,279 | 74,808 | 731,471 | 731,471 |
+| `sv_se` | 12,371 | 110,468 | 24,731 | 85,737 | 85,737 |
+| `uk_ua` | 1,493 | 15,737 | 2,985 | 12,752 | 12,752 |
+| `us_uk` | 396,939 | 1,004,374 | 793,874 | 210,500 | 210,500 |
+| `yi` | 802 | 4,300 | 1,524 | 2,776 | 5,000 |
+
+Speed benchmarks process the complete changed-token dictionary sequence for the language. Only resources with fewer than 5,000 changed tokens are repeated to reach the minimum timing size; larger resources are not sampled or truncated.
--- a/docs/benchmarks/reference/english-coverage.md
+++ b/docs/benchmarks/reference/english-coverage.md
@@ -0,0 +1,43 @@
+# English Dictionary Coverage Benchmark
+
+`EnglishRadixorDictionaryCoverageBenchmark` builds Radixor from deterministic slices of the English dictionary rows and evaluates accuracy against the complete dictionary. The speed method then stems the full changed-token English timing corpus.
+
+This benchmark is the clearest demonstration of the Radixor quality/speed envelope after contracted-trie compilation. More dictionary knowledge still gives the strongest changed-form precision, but uniform-subtree contraction removes much of the historical lookup-depth penalty. The table should therefore be read as a measured operating curve rather than as a strictly monotonic function of dictionary size.
+
+| Used rows | Actual row ratio | All exact | Changed exact | Root preserved | Speed ms/op | Error ms | ns/token |
+| ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
+| 100% | 100.000% | 97.478% | 97.197% | 97.552% | 28.578 | 7.571 | 135.8 |
+| 90% | 90.000% | 97.047% | 94.913% | 97.613% | 26.612 | 9.227 | 126.4 |
+| 80% | 80.000% | 96.635% | 92.768% | 97.661% | 23.331 | 8.106 | 110.8 |
+| 70% | 70.000% | 96.209% | 90.565% | 97.705% | 22.362 | 1.957 | 106.2 |
+| 60% | 60.000% | 95.750% | 88.384% | 97.703% | 16.497 | 2.026 | 78.4 |
+| 50% | 50.000% | 95.262% | 86.107% | 97.690% | 16.035 | 0.986 | 76.2 |
+| 40% | 40.000% | 94.753% | 83.855% | 97.643% | 16.459 | 0.664 | 78.2 |
+| 30% | 30.000% | 94.208% | 81.651% | 97.537% | 19.566 | 0.758 | 92.9 |
+| 20% | 20.000% | 93.633% | 79.366% | 97.416% | 14.616 | 0.487 | 69.4 |
+| 10% | 10.000% | 92.868% | 76.516% | 97.204% | 18.093 | 3.147 | 86.0 |
+
+## Column Meanings
+
+- `Used rows`: requested deterministic percentage of English dictionary rows used to build the trie.
+- `Actual row ratio`: selected rows divided by all parsed English dictionary rows.
+- `All exact`: exact agreement over the complete dictionary.
+- `Changed exact`: exact agreement over dictionary tokens where `token != expectedRoot`.
+- `Root preserved`: percentage of already-root dictionary tokens that are left unchanged.
+- `Speed ms/op`: JMH average time for one full changed-token English operation.
+- `Error ms`: JMH score error converted to milliseconds.
+- `ns/token`: `Speed ms/op` divided by 210,500 changed English tokens.
+
+For non-English languages, the same principle applies: dictionary-driven Radixor quality depends on the amount and consistency of the language resource, while contracted tries reduce the cost of uniform regions in the compiled lookup graph. The English table is the clearest because the English resource is large and the benchmark can show gradual deterministic reductions from 100% to 10%.
+
+## Why The Historical Porter Ratio Changed
+
+The historical English benchmark in `HEAD` used synthetic lexical families. Its `familyCount=5000` parameter generated roughly 70,000 artificial tokens rather than measuring the complete real English dictionary resource. That older workload was useful as a low-level stress test, but it was not a dictionary-quality comparison. Many synthetic tokens were not present in the Radixor dictionary, so Radixor often executed a fast miss path where lookup returned `null` and no patch command was applied.
+
+The current benchmark is intentionally based on real Radixor dictionary data. For English, the speed workload processes 210,500 changed token/root pairs where the dictionary token differs from the expected root, and the quality workload evaluates the complete 1,004,374-token dictionary. This is a hit-heavy workload that measures real lookup plus compiled patch-command application against known expected roots. It is therefore a different and more linguistically meaningful workload than the historical synthetic benchmark.
+
+The result must be interpreted in Radixor's favor through both speed and exact-root quality. Non-Radixor stemmers can look faster because many of them perform narrower or more aggressive transformations and do not attempt to match the dictionary root with the same precision. The English result table shows that this speed often comes with substantially lower `All exact` and `Changed exact` accuracy.
+
+Radixor uses the dictionary as training data for transformation rules. With the full English dictionary, it reaches much higher exact-root agreement than the Porter-family and other narrow baselines. Higher speed is still possible by reducing the amount or complexity of the input dictionary used to build the stemmer, but that is an explicit quality/speed trade-off rather than an accidental benchmark artifact.
+
+The coverage table shows that contracted tries substantially improve the operating point. Reducing dictionary knowledge still primarily damages changed-form exactness, while root preservation remains high. Even when Radixor is trained from only 10% of the English dictionary rows, the complete-dictionary `All exact` score remains above 92%. This is why Radixor performance should be discussed as a configurable quality/speed point, not as a single fixed ratio against Porter.
--- a/docs/benchmarks/reference/environment.md
+++ b/docs/benchmarks/reference/environment.md
@@ -0,0 +1,56 @@
+# Benchmark Environment And Reports
+
+The values below are environment-specific and must not be read as universal performance claims.
+
+| Item | Value |
+| --- | --- |
+| Benchmark date | 2026-07-06 (Europe/Prague) |
+| Focused comparison command family | `./gradlew jmh -Pjmh.includes='.*StemmerComparisonBenchmark.*' --no-daemon` |
+| English coverage command | `./gradlew jmh -Pjmh.includes='.*EnglishRadixorDictionaryCoverageBenchmark.*' --no-daemon` |
+| Speed result reports | `build/reports/jmh/stemmer-comparison-2026-07-06.csv`, `build/reports/jmh/stemmer-comparison-2026-07-06.txt`, `build/reports/jmh/english-coverage-2026-07-06.csv`, and `build/reports/jmh/english-coverage-2026-07-06.txt` |
+| Accuracy result reports | `build/reports/jmh/stemmer-comparison-2026-07-06.csv`, `build/reports/jmh/english-coverage-2026-07-06.csv`, and deterministic Radixor exact-root accounting over the same bundled language corpora |
+| Final comparison JMH scope | Stemmer comparison benchmarks only; internal `FrequencyTrie*` microbenchmarks were not run |
+| Coverage JMH scope | English Radixor dictionary coverage benchmark only |
+| JMH version | 1.37 |
+| Speed benchmark mode | Average time, `time/op` |
+| Score unit | `ns/op` |
+| Speed warmup | 3 iterations, 1 s each |
+| Speed measurement | 5 iterations, 1 s each |
+| Accuracy warmup | 3 JMH warmup iterations were applied by the Gradle invocation; timing scores from quality methods are not interpreted |
+| Accuracy measurement | 5 JMH measurement samples; documentation uses deterministic auxiliary counter ratios from the same report |
+| Fork count in generated report files | 1 |
+| Default fork policy for accuracy-only benchmark classes | `@Fork(0)` for future default runs because accuracy counters are deterministic and not interpreted as speed |
+| Thread count | 1 |
+| JVM reported by JMH | JDK 25.0.3, OpenJDK 64-Bit Server VM, 25.0.3+9 |
+| Java runtime | OpenJDK Runtime Environment, Red Hat build 25.0.3+9 |
+| JVM invoker | `/usr/lib/jvm/java-25-openjdk/bin/java` |
+| Operating system | Fedora Linux 44 (MATE-Compiz) |
+| Kernel | Linux 7.0.12-201.fc44.x86_64 |
+| Architecture | x86_64 |
+| CPU | AMD Ryzen 5 8600G w/ Radeon 760M Graphics |
+| Physical cores | 6 |
+| Logical CPUs | 12 |
+
+## Contracted Trie Baseline
+
+All Radixor rows in the refreshed benchmark tables use contracted compiled patch tries. During compilation, a subtree whose reachable entries all resolve to the same preferred patch command is represented as an accepting leaf. Runtime lookup can therefore stop as soon as that leaf is reached, which reduces depth in uniform regions while preserving the preferred result used by `get()`.
+
+## Report Files
+
+Generated local report files for this benchmark update:
+
+- `build/reports/jmh/stemmer-comparison-2026-07-06.csv`
+- `build/reports/jmh/stemmer-comparison-2026-07-06.txt`
+- `build/reports/jmh/english-coverage-2026-07-06.csv`
+- `build/reports/jmh/english-coverage-2026-07-06.txt`
+
+JMH TXT and CSV reports are still published as benchmark artifacts. They are not converted into a Porter speed badge.
+
+## Published Metrics
+
+The historical English Radixor versus Porter performance badge is no longer generated. `tools/generate-pages-badges.py` now produces only coverage and mutation badge endpoint JSON files:
+
+- `coverage-badge.json`
+- `pitest-badge.json`
+
+The README therefore no longer presents a single Porter speed ratio. Benchmark interpretation now uses both speed and quality, because a narrow or aggressive stemmer can be fast while disagreeing with the dictionary root much more often than Radixor.
--- a/docs/benchmarks/reference/linguistic-quality.md
+++ b/docs/benchmarks/reference/linguistic-quality.md
@@ -0,0 +1,82 @@
+# Linguistic Quality Methodology
+
+This evaluation measures agreement between the relation predicted by a stemmer and the gold-standard relation represented by Radixor dictionary groups. It does not require a generated stem to equal one predetermined lemma string. Runtime performance and linguistic quality are separate measurements.
+
+## Scope and fair-comparison rules
+
+The authoritative Radixor language universe is the reconciled set of `stemmer.gz` resources under `src/main/resources` and `StemmerPatchTrieLoader.Language`. Radixor is evaluated for every reconciled language. A third-party adapter is evaluated only for languages supported by its tested implementation and having a compatible Radixor dictionary; unsupported combinations are absent rather than assigned zero quality.
+
+Within one language and dictionary mode, every adapter receives the same original included forms. Exact duplicates are removed only within one dictionary row. Identical surface forms in different rows remain distinct entries. Candidate strings use exact `String.equals`, with no evaluation-only lowercasing, normalization, accent removal, or gold-label-aware selection. Adapter preprocessing and lifecycle match the JMH comparison path.
+
+## Gold-standard pairs
+
+Every usable dictionary row is a gold-standard equivalence group. An unordered pair from the same row is positive; a pair from different rows is negative. For group size `n`, `C2(n) = n * (n - 1) / 2`.
+
+- `TP = underPossiblePairs - underErrorPairs`: same-group pairs correctly related.
+- `FN = underErrorPairs`: same-group pairs incorrectly separated.
+- `FP = overErrorPairs`: different-group pairs incorrectly related.
+- `TN = overPossiblePairs - overErrorPairs`: different-group pairs correctly separated.
+
+Under-stemming is the false-negative relation among same-group pairs. Over-stemming is the false-positive relation among different-group pairs. Their percentages use different denominators and must not be added or averaged without an explicitly defined composite.
+
+## Dictionary-processing modes
+
+- `ALL_WORDS` includes every valid group and preserves every original form.
+- `LOWERCASE_GROUPS_ONLY` excludes an entire group if any Unicode code point is uppercase or titlecase. Retained forms are not converted to lowercase. Digits, punctuation, combining marks, and characters without case distinctions do not exclude a group by themselves.
+
+## Output policies
+
+`PRIMARY_OUTPUT` uses the adapter's deterministic primary stem. It defines a strict predicted partition and is the principal direct comparison between implementations.
+
+`ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound. Same-group pairs succeed when candidate sets intersect. Different-group pairs avoid an error whenever a non-colliding candidate selection exists. Selection may differ between pairs, so this policy is not deterministic runtime behaviour and may not correspond to one globally realizable assignment.
+
+`ALL_CANDIDATES` treats every returned candidate as active. Two forms are related when their candidate sets intersect. Alternatives can recover same-group relationships while introducing cross-group collisions. This overlapping relation need not be transitive or form a partition.
+
+Candidate-aware policies are reported as capability analyses. They are not mixed into the principal `PRIMARY_OUTPUT` ranking.
+
+## Relation metrics
+
+Undefined denominators produce `n/a`, never zero, `NaN`, or infinity. Metrics are calculated from unrounded raw counts and displayed with six decimals.
+
+| Metric | Formula | Range and interpretation | Sensitivity and applicability |
+| --- | --- | --- | --- |
+| Under-stemming rate | `FN / (TP + FN)` | `[0, 1]`; lower is better. False-negative rate over same-group pairs. | Sensitive to splitting large gold groups. All policies. |
+| Over-stemming rate | `FP / (TN + FP)` | `[0, 1]`; lower is better. False-positive rate over different-group pairs. | The denominator is usually very large. All policies. |
+| Precision | `TP / (TP + FP)` | `[0, 1]`; higher is better. Fraction of predicted relations that are gold-positive. | Penalizes over-stemming. All policies, with oracle-assisted interpretation for `ANY_CANDIDATE`. |
+| Recall | `TP / (TP + FN)` | `[0, 1]`; higher is better. Fraction of gold-positive pairs recovered. | Equivalent to one minus the under-stemming rate. All policies. |
+| Specificity | `TN / (TN + FP)` | `[0, 1]`; higher is better. Fraction of negative pairs separated. | Sensitive to cross-group collisions. All policies. |
+| Balanced accuracy | `(recall + specificity) / 2` | `[0, 1]`; higher is better. Equal weight for positive and negative classes. | Primary navigation metric; less dominated by TN than ordinary accuracy, but not uniquely authoritative. |
+| Pairwise accuracy | `(TP + TN) / (TP + TN + FP + FN)` | `[0, 1]`; higher is better. | Can be dominated by the very large TN class and is not the default ranking metric. |
+| Pairwise error rate | `(FP + FN) / (TP + TN + FP + FN)` | `[0, 1]`; lower is better. | Also sensitive to the number of negative pairs. |
+| F0.5 | `1.25 TP / (1.25 TP + 0.25 FN + FP)` | `[0, 1]`; higher is better. | Gives greater weight to precision and over-stemming avoidance. |
+| F1 | `2 TP / (2 TP + FN + FP)` | `[0, 1]`; higher is better. | Equal precision/recall emphasis. |
+| F2 | `5 TP / (5 TP + 4 FN + FP)` | `[0, 1]`; higher is better. | Gives greater weight to recall and under-stemming avoidance. |
+| Jaccard | `TP / (TP + FP + FN)` | `[0, 1]`; higher is better. | Excludes TN. All policies. |
+| Fowlkes–Mallows | `sqrt(precision * recall)` | `[0, 1]`; higher is better. | Geometric balance of precision and recall. All policies. |
+| MCC | `(TP TN - FP FN) / sqrt((TP+FP)(TP+FN)(TN+FP)(TN+FN))` | `[-1, 1]`; higher is better. Uses all four counts. | Informative under imbalance; undefined for a zero product denominator. All policies with policy-specific interpretation. |
+
+The general F-beta formula is `((1 + betaSquared) * TP) / (((1 + betaSquared) * TP) + (betaSquared * FN) + FP)`.
+
+## Partition-only metrics
+
+These metrics apply only to `PRIMARY_OUTPUT`. Candidate relations are not forced into artificial partitions.
+
+- Adjusted Rand Index is the Rand agreement corrected for agreement expected from the gold/predicted contingency-table marginals. Its usual range is `[-1, 1]`, with `1` indicating identical partitions.
+- Homogeneity is `1 - H(gold | predicted) / H(gold)`, in `[0, 1]`; each predicted cluster ideally contains one gold group.
+- Completeness is `1 - H(predicted | gold) / H(predicted)`, in `[0, 1]`; each gold group ideally maps to one predicted cluster.
+- V-measure is the harmonic mean of homogeneity and completeness, in `[0, 1]`.
+- Normalized mutual information uses arithmetic-mean entropy normalization: `MI / ((H(gold) + H(predicted)) / 2)`, in `[0, 1]` under this implementation.
+
+Entropy zero cases follow the evaluator's explicit perfect/undefined conventions. Language tables render inapplicable candidate-policy values as `n/a`.
+
+## Aggregation and ranking
+
+Macro metrics average defined per-language values, giving each language equal weight. Micro metrics sum TP, FP, FN, and TN before calculating a metric. Cross-stemmer aggregate comparisons require the exact common supported-language intersection; unsupported languages are not zero-filled.
+
+Language tables sort by unrounded balanced accuracy, then MCC, F1, over-stemming rate, over-stemming error count, under-stemming rate, stemmer name, and stable policy order. Display rounding never controls rank.
+
+Multiple metrics and Pearson/Spearman correlation datasets are published because metric suitability and correlation remain analytical questions. Strong correlation does not establish equivalence.
+
+## Limitations
+
+Dictionary groups encode the available annotation, not every linguistic distinction. Homographs may occur in different groups, singleton rows contribute no under-stemming pair, and group size affects pair counts. `ANY_CANDIDATE` is optimistic; `ALL_CANDIDATES` measures an overlapping graph; neither is a deterministic global assignment. Results characterize the tested versions, adapters, dictionaries, and preprocessing, not every deployment or domain.
--- a/docs/benchmarks/reference/methodology.md
+++ b/docs/benchmarks/reference/methodology.md
@@ -0,0 +1,59 @@
+# Benchmark Methodology
+
+The stemmer comparison suite measures Radixor and Java stemmers on the same language and deterministic Radixor dictionary-derived data. Published Radixor rows in this refresh use contracted compiled patch tries, where uniform preferred-command subtrees are collapsed into accepting leaves before the trie is frozen for lookup. For each language, the bundled dictionary resource stores the expected root as the first tab-separated field on a line and its surface forms on the same line. Every single-token field on that line can therefore be paired with the same expected root.
+
+Published stemmer comparison results must come only from benchmark classes matching `.*StemmerComparisonBenchmark.*`. Internal `FrequencyTrie*` microbenchmarks are not part of those results.
+
+## Benchmark Passes
+
+There are two distinct benchmark passes:
+
+- Speed benchmarks process only changed dictionary pairs where `token != expectedRoot`. This removes already-root tokens from timing so a stemmer is measured on words that actually require a transformation. If a language has fewer than 5,000 changed pairs, the complete changed-pair sequence is repeated in stable order until the timing corpus has at least 5,000 tokens. Larger changed-pair corpora are not sampled or truncated.
+- Quality benchmarks process the complete dictionary for the language. They report exact agreement over all tokens, exact agreement over changed tokens only, and preservation of tokens that are already roots.
+
+Timing corpora are generated once per JMH JVM and kept in memory as shared `{token, expectedRoot}` arrays. Corpus construction, dictionary loading, trie loading, table loading, and analyzer construction are setup work and are not included in measured benchmark methods.
+
+Performance is interpreted as average time per input token:
+
+```text
+timePerChangedTokenNs = JMH score ns/op / changedTimingTokenCount
+```
+
+This is necessary because Radixor dictionaries have different token counts by language.
+
+## Exact-root quality and interpretation
+
+Runtime and exact-root agreement must be interpreted separately. Light, minimal, possessive, and aggressive rule-based implementations deliberately address different scopes and may achieve lower latency by performing fewer transformations. A throughput advantage does not establish higher linguistic quality, and higher dictionary agreement does not establish lower operational cost.
+
+The [English dictionary coverage benchmark](english-coverage.md) shows this operating curve explicitly: contracted tries reduce lookup cost in uniform regions, while reduced dictionary coverage still lowers changed-form precision.
+
+## Normalization Policy
+
+Radixor is measured over dictionary tokens from its own resources: lower-case with `Locale.ROOT`, diacritics preserved. The corpus is normalized during setup, so the Radixor benchmark path uses `FrequencyTrie.getNormalized(CharSequence)` and does not measure redundant lookup-time lowercasing or diacritic normalization.
+
+Lucene TokenFilter paths include required normalization in the measured pipeline. Examples include lower-case normalization for filters requiring lower-case input, German normalization before German light/minimal stemming, and Persian decimal, Arabic, and Persian normalization before Persian stemming. No ASCII folding is applied to Czech or Polish paths, because those Lucene stemmers are diacritic-aware or dictionary/table-backed for those languages. TokenFilter throughput methods materialize each emitted `CharTermAttribute` as a `String` before passing it to the JMH `Blackhole`, so output consumption is easier to inspect and closer to the direct stemmer methods.
+
+For right-to-left Radixor languages, patch application uses the traversal direction stored in trie metadata. This is required because static backward patch application is not correct for all bundled languages.
+
+## Quality Metric
+
+The quality pass reports exact-root agreement against the expected root from the Radixor dictionary line. It writes to the normal JMH report files:
+
+- `build/reports/jmh/jmh-results.csv`
+- `build/reports/jmh/jmh-results.txt`
+
+Accuracy is computed from standard JMH secondary rows:
+
+```text
+allExactPercent = correctMatches / evaluatedTokens * 100
+changedExactPercent = changedCorrectMatches / changedEvaluatedTokens * 100
+rootPreservedPercent = rootPreservedMatches / rootEvaluatedTokens * 100
+```
+
+`allExactPercent` uses the complete dictionary. `changedExactPercent` uses only tokens where `token != expectedRoot`. `rootPreservedPercent` measures whether a stemmer leaves already-root dictionary entries unchanged.
+
+Morfologik can emit multiple terms for one input token. The quality benchmark uses the first emitted term for exact-root accounting when no ranking weight is exposed. Throughput benchmarks for Morfologik TokenFilter paths consume all emitted terms.
+
+Quality reports use JMH auxiliary counter rows. Exact-root accounting is deterministic for a fixed corpus and stemmer, so repeated measurement samples duplicate the same counters; documentation uses the counter ratios and does not interpret quality benchmark timing scores.
+
+Pairwise over-stemming, under-stemming, candidate-aware policies, balanced accuracy, and partition comparison are a separate analytical evaluation. See [Linguistic Quality Methodology](linguistic-quality.md); exact-root accuracy must not be interpreted as the complement of pairwise under-stemming.
--- a/docs/benchmarks/reference/reproducibility.md
+++ b/docs/benchmarks/reference/reproducibility.md
@@ -0,0 +1,61 @@
+# Reproducibility and Raw Data
+
+## Published quality snapshot
+
+- Machine-readable CSV: [stemming-quality.csv](../data/stemming-quality.csv)
+- SHA-256 record: [stemming-quality.sha256](../data/stemming-quality.sha256)
+- SHA-256: `5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28`
+- Complete scenarios: 308
+- Authoritative language universe: 20 languages
+- Language-page scenarios: 302 across 19 existing benchmark pages
+
+The six remaining scenarios are the three Radixor policies in two modes for `HE_IL`. Hebrew is present in the complete result snapshot but has no existing language benchmark page.
+
+The CSV contains raw TP, FP, FN, and TN counts; raw over/under numerators and denominators; candidate statistics; relation metrics; and partition-only metrics. Documentation is regenerated from this file rather than manually transcribed.
+
+## Commands
+
+```bash
+./gradlew stemmingQuality
+./gradlew publishStemmingQualityDocumentation
+./gradlew verifyStemmingQualityDocumentation
+./gradlew test
+mkdocs build --strict
+```
+
+`stemmingQuality` performs the expensive complete evaluation and is intentionally not attached to `test` or `check`. It prepares JMH third-party dependencies automatically and writes:
+
+- `build/reports/stemming-quality/stemming-quality.csv`
+- `build/reports/stemming-quality/stemming-quality.md`
+- `build/reports/stemming-quality/metric-correlations-pearson.csv`
+- `build/reports/stemming-quality/metric-correlations-spearman.csv`
+
+Audit mode is enabled with `-PstemmingQualityAudit=true`. Language, stemmer, dictionary-mode, output-policy, and ranking filters are documented on the central [stemming-quality page](../../stemming-quality.md). Filtered reports use separate filenames and cannot be accepted as publication sources.
+
+`publishStemmingQualityDocumentation` validates the complete build CSV, copies a versioned documentation snapshot, and replaces only marked generated sections. `verifyStemmingQualityDocumentation` re-renders from the checked-in snapshot and fails on changed values, ordering, missing pages, duplicate keys, arithmetic inconsistencies, policy violations, or stale sections.
+
+## Performance benchmark reproduction
+
+The JMH comparison command family is:
+
+```bash
+./gradlew jmh -Pjmh.includes='.*StemmerComparisonBenchmark.*' --no-daemon
+```
+
+The exact JMH configuration, hardware, operating system, and JDK captured for the published performance tables are listed in [Environment and reports](environment.md). Quality and performance reports are separate datasets and are not combined into an undocumented scalar.
+
+## Recorded and unavailable provenance
+
+The performance documentation records its 2026-07-06 environment, JDK 25.0.3, operating system, and hardware. The quality CSV records the evaluated identifiers and counts but does not embed the Radixor Git revision, generation date, JDK, operating system, dictionary content hash, or immutable upstream revisions for every downloaded source. These fields are explicitly unavailable for this snapshot and are not reconstructed from filesystem timestamps.
+
+Dependency versions that are reproducible from repository configuration include Apache Lucene 10.5.0, Morfologik 2.1.9, the Ukrainian dictionary artifact 4.9.1, and JMH 1.37. Other upstream branches or downloaded dictionary revisions should be pinned and embedded in a future result schema.
+
+## Correlation and audit data
+
+Pearson and Spearman files are generated from unrounded metric values in cohorts separated by dictionary mode and output policy. A missing coefficient means too few observations, undefined input, or zero variance. Correlation is descriptive and does not demonstrate that two metrics are scientifically interchangeable.
+
+Audit reports preserve original multilingual forms and identify high-contributing dictionary groups. They are build artifacts rather than checked-in publication data because of their size. No documentation value is manually altered after generation.
+
+## JMH badge compatibility
+
+The quality documentation generator does not invoke JMH, change JMH result formats, or modify badge tooling. Existing JMH result paths and historical badge-compatible inputs remain independent. The repository currently publishes coverage and mutation badge metadata and retains JMH TXT/CSV artifacts as documented in [Environment and reports](environment.md).
--- a/docs/benchmarks/reference/tested-stemmers.md
+++ b/docs/benchmarks/reference/tested-stemmers.md
@@ -0,0 +1,29 @@
+# Tested Stemmer Inventory
+
+The JMH adapter registry is authoritative for evaluated implementations and language mappings. Names below describe the implementation actually invoked, not an abstract algorithm in every possible implementation. Unsupported language combinations are omitted rather than scored as failures.
+
+| Family or implementation | Upstream / attribution | Tested version or revision | Evaluated scope | Output capability and adapter behaviour | Interpretation notes |
+| --- | --- | --- | --- | --- | --- |
+| Radixor | Egothor / Radixor project | Current repository revision; exact revision was not embedded in the quality CSV | All 20 reconciled Radixor dictionary languages; 19 have benchmark pages | Deterministic preferred patch via `get`; ranked distinct alternatives via `getAll`; primary is always included | Dictionary-derived compiled patch trie. Quality depends on dictionary coverage and annotation. |
+| Apache Lucene language stem filters | Apache Lucene project | 10.5.0 | Adapter-declared language-specific subsets | TokenFilter lifecycle and language normalization match JMH; normally single-output | Light, minimal, possessive, and language stem filters deliberately implement different scopes. Narrow scope is not a defect. |
+| Apache Lucene SnowballFilter | Apache Lucene project using Snowball algorithms | Lucene 10.5.0 | Snowball-supported subset of Radixor languages | Single primary token emitted through the Lucene TokenFilter path | Includes TokenStream overhead and required normalization. |
+| Official Snowball Java | Snowball project | Repository preparation downloads the configured upstream Java distribution; an immutable revision was not recorded in the quality CSV | Same-language adapter subset | Direct generated Java API; single output | Rule-based suffix algorithms provide broad baselines rather than dictionary-root guarantees. |
+| Lucene Stempel | Apache Lucene / Polish stemming tables | Lucene 10.5.0 | Polish | Direct and TokenFilter paths where registered; single primary output | Table-driven Polish implementation. |
+| Morfologik | Morfologik project; Lucene integration by Apache Lucene | Morfologik 2.1.9, Lucene integration 10.5.0; Ukrainian dictionary artifact 4.9.1 | Registered Polish and Ukrainian paths | Deterministic first lemma for primary comparison; all distinct lemma strings for candidate policies | Several analyses may share a lemma and are deduplicated by exact string equality. |
+| Hunspell via Lucene | Hunspell dictionaries from the `wooorm/dictionaries` repository; adapter by Apache Lucene | Lucene 10.5.0; dictionary repository revision was not recorded | Configured German, English, Spanish, French, Dutch, Polish, and Ukrainian dictionaries | First emitted stem is primary; all distinct stems at the token position are candidates | Dictionary content and affix rules differ by language. |
+| CISTEM | Leonie Weissweiler, CISTEM project | Upstream `master` source path used by preparation; immutable commit not recorded | German | Single output | German stemming algorithm; benchmark-only implementation and gold-standard preparation remain under JMH infrastructure. |
+| OpenNLP Porter | Apache OpenNLP project | Version resolved by `gradle/opennlp-benchmarks.gradle` and `gradle.lockfile` | English | Direct single output | Porter-family English baseline. |
+| Lucene Porter source copy | Apache Lucene project | 10.5.0 source artifact | English | Package-isolated benchmark-only generated source; single output | Generated into the JMH build tree, never production code. |
+| Paice/Husk Lancaster | Upstream Java implementation from `Hopper262/paice-husk-stemmer` | Configured upstream branch/revision in `gradle/paicehusk-benchmarks.gradle`; immutable commit not recorded | English | Direct single output | Aggressive rule-based English baseline; benchmark-only generated source. |
+
+## Preprocessing and lifecycle
+
+The quality evaluator calls the same adapter matrix used by JMH. Each language mapping is explicit. Retained dictionary forms are not evaluation-lowercased or normalized. Where an implementation requires preprocessing, such as Lucene German or Persian normalization, that operation is part of its documented adapter path. Stateful TokenFilters are reset through the same sequential lifecycle used by the benchmark and are not invoked concurrently.
+
+Candidate sets are non-null, non-empty, contain the deterministic primary output, contain no null strings, and are deduplicated using exact Java string equality. Gold-standard group identity never selects, removes, or ranks a candidate.
+
+## Coverage fairness
+
+Radixor coverage is derived independently from its resources and language enumeration. Third-party coverage is the intersection of that universe with actual adapter support. Absence therefore means “not supported or not configured for this language,” not “zero quality.” Consult each language page for the exact evaluated rows.
+
+Project authors and organizations are named only where repository configuration or source notices establish attribution. No broader authorship or license claim is inferred when metadata was not captured.
--- a/docs/built-in-languages.md
+++ b/docs/built-in-languages.md
@@ -10,7 +10,7 @@ Bundled dictionaries are exposed through:
 org.egothor.stemmer.StemmerPatchTrieLoader.Language
 ```

-Each bundled dictionary is packaged with the library as a compressed UTF-8 text resource. When loaded, the resource is parsed by `StemmerDictionaryParser`, transformed into patch-command mappings, and compiled into a read-only `FrequencyTrie<String>` by `StemmerPatchTrieLoader`.
+Each bundled dictionary is packaged with the library as a compressed UTF-8 text resource. When loaded through the runtime API, the resource is parsed by `StemmerDictionaryParser`, transformed into patch-command mappings, and compiled into a read-only `FrequencyTrie<CompiledPatchCommand>` by `StemmerPatchTrieLoader`.

 The bundled language definition also carries a language-level right-to-left flag. That flag is used by the loader to derive the `WordTraversalDirection` used for both trie-key construction and patch-command generation. In practice, left-to-right bundled languages use historical backward Egothor traversal, while right-to-left bundled languages use forward traversal over the stored form.

@@ -18,28 +18,28 @@ The bundled language definition also carries a language-level right-to-left flag

 The following bundled language identifiers are currently available:

-| Language | Enum constant | Writing direction | Notes |
-|---|---|---:|---|
-| Czech | `CS_CZ` | LTR | Bundled general-purpose dictionary |
-| Danish | `DA_DK` | LTR | Bundled general-purpose dictionary |
-| German | `DE_DE` | LTR | Bundled general-purpose dictionary |
-| Spanish | `ES_ES` | LTR | Bundled general-purpose dictionary |
-| Persian | `FA_IR` | RTL | Bundled dictionary uses forward traversal over the stored form |
-| Finnish | `FI_FI` | LTR | Bundled general-purpose dictionary |
-| French | `FR_FR` | LTR | Bundled general-purpose dictionary |
-| Hebrew | `HE_IL` | RTL | Bundled dictionary uses forward traversal over the stored form |
-| Hungarian | `HU_HU` | LTR | Bundled general-purpose dictionary |
-| Italian | `IT_IT` | LTR | Bundled general-purpose dictionary |
-| Norwegian Bokmål | `NB_NO` | LTR | Bundled general-purpose dictionary |
-| Dutch | `NL_NL` | LTR | Bundled general-purpose dictionary |
-| Norwegian Nynorsk | `NN_NO` | LTR | Bundled general-purpose dictionary |
-| Polish | `PL_PL` | LTR | Bundled general-purpose dictionary |
-| Portuguese | `PT_PT` | LTR | Bundled general-purpose dictionary |
-| Russian | `RU_RU` | LTR | Bundled general-purpose dictionary |
-| Swedish | `SV_SE` | LTR | Bundled general-purpose dictionary |
-| Ukrainian | `UK_UA` | LTR | Bundled general-purpose dictionary |
-| English | `US_UK` | LTR | Bundled general-purpose dictionary |
-| Yiddish | `YI` | RTL | Bundled dictionary uses forward traversal over the stored form |
+| Language | Enum constant | Writing direction | Notes | Benchmark page |
+|---|---|---:|---|---|
+| Czech | `CS_CZ` | LTR | Bundled general-purpose dictionary | [Czech](benchmarks/languages/czech.md) |
+| Danish | `DA_DK` | LTR | Bundled general-purpose dictionary | [Danish](benchmarks/languages/danish.md) |
+| German | `DE_DE` | LTR | Bundled general-purpose dictionary | [German](benchmarks/languages/german.md) |
+| Spanish | `ES_ES` | LTR | Bundled general-purpose dictionary | [Spanish](benchmarks/languages/spanish.md) |
+| Persian | `FA_IR` | RTL | Bundled dictionary uses forward traversal over the stored form | [Persian](benchmarks/languages/persian.md) |
+| Finnish | `FI_FI` | LTR | Bundled general-purpose dictionary | [Finnish](benchmarks/languages/finnish.md) |
+| French | `FR_FR` | LTR | Bundled general-purpose dictionary | [French](benchmarks/languages/french.md) |
+| Hebrew | `HE_IL` | RTL | Bundled dictionary uses forward traversal over the stored form | No same-language external benchmark in this run |
+| Hungarian | `HU_HU` | LTR | Bundled general-purpose dictionary | [Hungarian](benchmarks/languages/hungarian.md) |
+| Italian | `IT_IT` | LTR | Bundled general-purpose dictionary | [Italian](benchmarks/languages/italian.md) |
+| Norwegian Bokmål | `NB_NO` | LTR | Bundled general-purpose dictionary | [Norwegian Bokmal](benchmarks/languages/norwegian-bokmal.md) |
+| Dutch | `NL_NL` | LTR | Bundled general-purpose dictionary | [Dutch](benchmarks/languages/dutch.md) |
+| Norwegian Nynorsk | `NN_NO` | LTR | Bundled general-purpose dictionary | [Norwegian Nynorsk](benchmarks/languages/norwegian-nynorsk.md) |
+| Polish | `PL_PL` | LTR | Bundled general-purpose dictionary | [Polish](benchmarks/languages/polish.md) |
+| Portuguese | `PT_PT` | LTR | Bundled general-purpose dictionary | [Portuguese](benchmarks/languages/portuguese.md) |
+| Russian | `RU_RU` | LTR | Bundled general-purpose dictionary | [Russian](benchmarks/languages/russian.md) |
+| Swedish | `SV_SE` | LTR | Bundled general-purpose dictionary | [Swedish](benchmarks/languages/swedish.md) |
+| Ukrainian | `UK_UA` | LTR | Bundled general-purpose dictionary | [Ukrainian](benchmarks/languages/ukrainian.md) |
+| English | `US_UK` | LTR | Bundled general-purpose dictionary | [English](benchmarks/languages/english.md) |
+| Yiddish | `YI` | RTL | Bundled dictionary uses forward traversal over the stored form | [Yiddish](benchmarks/languages/yiddish.md) |

 ## Basic usage

@@ -48,6 +48,7 @@ Load a bundled dictionary like this:
 ```java
 import java.io.IOException;

+import org.egothor.stemmer.CompiledPatchCommand;
 import org.egothor.stemmer.FrequencyTrie;
 import org.egothor.stemmer.ReductionMode;
 import org.egothor.stemmer.StemmerPatchTrieLoader;
@@ -59,7 +60,7 @@ public final class BuiltInExample {
    }

    public static void main(final String[] arguments) throws IOException {
-        final FrequencyTrie<String> trie = StemmerPatchTrieLoader.load(
+        final FrequencyTrie<CompiledPatchCommand> trie = StemmerPatchTrieLoader.loadCompiled(
                StemmerPatchTrieLoader.Language.US_UK,
                true,
                ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS);
@@ -76,8 +77,8 @@ This call loads the bundled dictionary resource for the selected language, parse
 ```java
 import java.io.IOException;

+import org.egothor.stemmer.CompiledPatchCommand;
 import org.egothor.stemmer.FrequencyTrie;
-import org.egothor.stemmer.PatchCommandEncoder;
 import org.egothor.stemmer.ReductionMode;
 import org.egothor.stemmer.StemmerPatchTrieLoader;

@@ -88,21 +89,21 @@ public final class EnglishExample {
    }

    public static void main(final String[] arguments) throws IOException {
-        final FrequencyTrie<String> trie = StemmerPatchTrieLoader.load(
+        final FrequencyTrie<CompiledPatchCommand> trie = StemmerPatchTrieLoader.loadCompiled(
                StemmerPatchTrieLoader.Language.US_UK,
                true,
                ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS);

        final String word = "running";
-        final String patch = trie.get(word);
-        final String stem = PatchCommandEncoder.apply(word, patch, trie.traversalDirection());
+        final CompiledPatchCommand patch = trie.get(word);
+        final String stem = patch == null ? word : patch.apply(word);

        System.out.println(word + " -> " + stem);
    }
 }
 ```

-Passing `trie.traversalDirection()` to `PatchCommandEncoder.apply(...)` is the correct general contract. It ensures that the patch is applied using the same logical traversal model that was used when the trie and its patch commands were produced.
+`CompiledPatchCommand` values are compiled with the traversal direction used when the trie and its patch commands were produced.

 ## Traversal behavior and right-to-left languages

@@ -114,7 +115,7 @@ For bundled right-to-left languages such as Persian, Hebrew, and Yiddish, Radixo

 - trie keys are traversed from the logical beginning of the stored form,
 - patch commands are generated in that same forward direction,
- patch application must use `WordTraversalDirection.FORWARD`, which is naturally obtained from `trie.traversalDirection()`.
+- compiled patch-command application uses `WordTraversalDirection.FORWARD`, which is naturally captured when `loadCompiled(...)` creates `CompiledPatchCommand` values.

 This design keeps the traversal policy explicit and consistent across dictionary loading, trie lookup, binary persistence, builder reconstruction, and patch application.

@@ -124,11 +125,16 @@ Bundled dictionaries can be compiled using any supported `ReductionMode`. The re

 Typical entry points are:

- `StemmerPatchTrieLoader.load(language, storeOriginal, reductionMode)`
- `StemmerPatchTrieLoader.load(language, storeOriginal, reductionSettings)`
+- `StemmerPatchTrieLoader.loadCompiled(language, storeOriginal, reductionMode)`
+- `StemmerPatchTrieLoader.loadCompiled(language, storeOriginal, reductionSettings)`

 For most users, `ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS` is the most conservative general-purpose choice because it preserves ranked `getAll(...)` behavior.

+Compiled bundled dictionaries also use internal uniform-subtree contraction. If a whole subtree
+would return the same preferred patch command, Radixor stores that subtree as an accepting leaf and
+removes the deeper branches. This is the contracted trie representation used by the published
+benchmark tables and is independent of the public reduction mode selected by the caller.
+
 ## Intended role of bundled dictionaries

 Bundled dictionaries should be understood as practical default resources.
@@ -178,7 +184,7 @@ For production systems, the most robust workflow is usually:
 2. extend it with domain-specific forms if needed,
 3. rebuild it into a binary artifact,
 4. deploy that compiled binary artifact,
-5. load it at runtime through `loadBinary(...)`.
+5. load it at runtime through `loadBinaryCompiled(...)`.

 This avoids repeated startup parsing and makes the deployed stemming behavior explicit, reproducible, and versionable.

@@ -190,6 +196,7 @@ import java.nio.file.Path;

 import org.egothor.stemmer.FrequencyTrie;
 import org.egothor.stemmer.FrequencyTrieBuilders;
+import org.egothor.stemmer.PatchCommandEncoder;
 import org.egothor.stemmer.ReductionMode;
 import org.egothor.stemmer.ReductionSettings;
 import org.egothor.stemmer.StemmerPatchTrieBinaryIO;
@@ -213,7 +220,11 @@ public final class BundledRefinementExample {
                ReductionSettings.withDefaults(
                        ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS));

-        builder.put("microservices", "Na");
+        final PatchCommandEncoder encoder = PatchCommandEncoder.builder()
+                .traversalDirection(base.traversalDirection())
+                .build();
+
+        builder.put("microservices", encoder.encode("microservices", "microservice"));

        final FrequencyTrie<String> compiled = builder.build();

--- a/docs/cli-compilation.md
+++ b/docs/cli-compilation.md
@@ -171,6 +171,11 @@ The short form `-h` is also supported.

 Reduction mode is not only a storage decision. It also influences what semantics are preserved when the mutable trie is compiled into its canonical read-only form.

+Before the selected public reduction mode is applied, compilation performs uniform-subtree
+contraction. If all reachable entries below a subtree select the same preferred patch command, the
+compiler stores that subtree as an accepting leaf and removes the deeper branches. This reduces
+runtime lookup depth without changing the preferred result returned by the standard stemming path.
+
 ### Ranked `getAll()` equivalence

 `MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS` merges subtrees whose `getAll()` results remain equivalent for every reachable key suffix and whose local result ordering is the same.
@@ -203,9 +208,9 @@ The CLI is best used as a preparation step during packaging, deployment, or cont

 ### Treat compiled files as versioned assets

-A `.radixor.gz` file should be handled as a versioned output artifact. It represents a specific dictionary state, a specific reduction mode, and, where relevant, specific dominant-result thresholds.
+A `.radixor.gz` file should be handled as a versioned output artifact. It represents a specific dictionary state, a specific reduction mode, whether uniform-subtree contraction was used, and, where relevant, specific dominant-result thresholds.

-Compiled tries also persist a human-readable metadata block (`key=value` lines) that includes format version, traversal direction, RTL indicator, reduction mode, dominant thresholds, diacritic-processing mode, and case-processing mode. After decompression, you can inspect this block directly to identify what dictionary/trie configuration the artifact contains. The current CLI uses `DiacriticProcessingMode.AS_IS`; custom diacritic stripping is available through the programmatic builder and loader APIs rather than through a CLI flag.
+Compiled tries also persist a human-readable metadata block (`key=value` lines) that includes format version, traversal direction, RTL indicator, reduction mode, contraction flag, dominant thresholds, diacritic-processing mode, and case-processing mode. After decompression, you can inspect this block directly to identify what dictionary/trie configuration the artifact contains. The current CLI uses `DiacriticProcessingMode.AS_IS`; custom diacritic stripping is available through the programmatic builder and loader APIs rather than through a CLI flag.

 ### Choose reduction mode deliberately

@@ -237,11 +242,12 @@ java org.egothor.stemmer.Compile \
 ### 3. Load it in an application

 ```java
+import org.egothor.stemmer.CompiledPatchCommand;
 import org.egothor.stemmer.FrequencyTrie;
 import org.egothor.stemmer.StemmerPatchTrieLoader;

-final FrequencyTrie<String> trie =
-        StemmerPatchTrieLoader.loadBinary("english.radixor.gz");
+final FrequencyTrie<CompiledPatchCommand> trie =
+        StemmerPatchTrieLoader.loadBinaryCompiled("english.radixor.gz");
 ```

 ## Exit codes and error handling
--- a/docs/compatibility-and-guarantees.md
+++ b/docs/compatibility-and-guarantees.md
@@ -81,7 +81,11 @@ Compiled `FrequencyTrie` instances are immutable and thread-safe for concurrent

 ### Stable patch application behavior

-`PatchCommandEncoder.apply(...)` remains the compatibility API for string-returning patch application. Buffer-oriented `applyTo(...)` overloads are additive APIs for caller-owned output storage. They do not retain output arrays, report insufficient capacity with `APPLY_INSUFFICIENT_CAPACITY`, and preserve the existing malformed-patch compatibility behavior where `apply(...)` preserves the source.
+Serialized patch-command strings remain the stable stored representation used by textual dictionaries and binary artifacts. Runtime stemming should use `CompiledPatchCommand` values produced by `StemmerPatchTrieLoader.loadCompiled(...)`, `StemmerPatchTrieLoader.loadBinaryCompiled(...)`, or `PatchCommandEncoder.compile(...)`.
+
+The historical `PatchCommandEncoder.apply(...)` and String-based `applyTo(...)` overloads remain compatibility APIs during the 2.x transition, but they are deprecated because they reparse the patch-command string on each application. See [Migration and Backward Compatibility](migration-and-backward-compatibility.md) for old and new code examples.
+
+Compiled buffer-oriented `CompiledPatchCommand.applyTo(...)` overloads use caller-owned output storage. They do not retain output arrays and report insufficient capacity with `CompiledPatchCommand.APPLY_INSUFFICIENT_CAPACITY`.

 ### Stable reduction-mode intent

@@ -89,6 +93,11 @@ Each public `ReductionMode` constant carries a semantic contract that should rem

 In other words, the implementation may evolve, but the intended meaning of modes such as ranked `getAll()` equivalence, unordered `getAll()` equivalence, and dominant `get()` equivalence should not drift casually.

+Internal pre-reduction optimizations may still change the physical compiled trie shape when they
+preserve the documented lookup contract. Uniform-subtree contraction is one such optimization: it
+can replace a subtree with an accepting leaf when all reachable entries choose the same preferred
+patch command.
+
 ### Stable binary artifact purpose

 Compiled `.radixor.gz` artifacts are a first-class project output. Loading and persisting compiled stemmer artifacts is part of the intended usage model, not an incidental implementation side effect.
--- a/docs/fast-track.md
+++ b/docs/fast-track.md
@@ -0,0 +1,130 @@
+# Fast Track
+
+This page is the shortest path from an empty Java project to a working Radixor stemmer.
+It deliberately uses a bundled dictionary and the preferred compiled-command runtime API, so the
+first result does not require writing a dictionary, running the CLI compiler, or understanding
+reduction internals.
+
+Use this page when the goal is:
+
+- add the dependency,
+- load a bundled language resource,
+- stem a token,
+- know where to go next.
+
+For deeper production guidance, see [Integration Deep Dive](integration-deep-dive.md).
+
+## 1. Add The Dependency
+
+Radixor is published as:
+
+```text
+groupId:    org.egothor
+artifactId: radixor
+```
+
+Use the current published version from Maven Central. The snippets below use `3.0.0`; replace it
+with the version you deploy if a newer release is available.
+
+For a Gradle project:
+
+```kotlin
+dependencies {
+    implementation("org.egothor:radixor:3.0.0")
+}
+```
+
+For a Maven project:
+
+```xml
+<dependency>
+    <groupId>org.egothor</groupId>
+    <artifactId>radixor</artifactId>
+    <version>3.0.0</version>
+</dependency>
+```
+
+Radixor targets modern Java and has a dependency-light runtime core. The project documentation and
+benchmarks assume a current JDK; Java 21 or newer is the practical baseline for current releases.
+
+## 2. Load A Bundled Dictionary
+
+The fastest path is to use a bundled dictionary through `StemmerPatchTrieLoader.Language`.
+This example uses the bundled English resource, `US_UK`.
+
+```java
+import java.io.IOException;
+
+import org.egothor.stemmer.CompiledPatchCommand;
+import org.egothor.stemmer.FrequencyTrie;
+import org.egothor.stemmer.ReductionMode;
+import org.egothor.stemmer.StemmerPatchTrieLoader;
+
+public final class RadixorFirstStem {
+
+    private RadixorFirstStem() {
+        throw new AssertionError("No instances.");
+    }
+
+    public static void main(final String[] arguments) throws IOException {
+        final FrequencyTrie<CompiledPatchCommand> stemmer = StemmerPatchTrieLoader.loadCompiled(
+                StemmerPatchTrieLoader.Language.US_UK,
+                true,
+                ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS);
+
+        final String token = "running";
+        final CompiledPatchCommand command = stemmer.get(token);
+        final String stem = command == null ? token : command.apply(token);
+
+        System.out.println(token + " -> " + stem);
+    }
+}
+```
+
+The loaded `FrequencyTrie<CompiledPatchCommand>` is immutable and can be shared across request
+threads. Load it once during application startup and reuse it for indexing and query processing.
+
+## 3. Choose A Language Resource
+
+Bundled dictionaries are exposed as enum constants. Common examples:
+
+| Language | Enum constant |
+| --- | --- |
+| English | `US_UK` |
+| German | `DE_DE` |
+| French | `FR_FR` |
+| Spanish | `ES_ES` |
+| Italian | `IT_IT` |
+| Polish | `PL_PL` |
+| Russian | `RU_RU` |
+| Czech | `CS_CZ` |
+
+The full list, writing-direction notes, and benchmark links are in
+[Built-in Languages](built-in-languages.md).
+
+## 4. Use The Same Stemmer On Both Sides
+
+For search, use the same Radixor configuration during indexing and query processing. A typical
+minimal integration flow is:
+
+1. tokenize text with your application or search platform,
+2. normalize tokens consistently,
+3. call `stemmer.get(token)`,
+4. apply the returned `CompiledPatchCommand`,
+5. index or query with the resulting stem.
+
+Do not load the trie per token. The compiled trie is the runtime artifact; per-token work should be
+limited to lookup and patch application.
+
+## 5. Next Step For Production
+
+The fast path compiles a bundled dictionary during startup. That is convenient for evaluation and
+small services. For larger deployments, compile once, persist a `.radixor.gz` artifact, and load
+that binary artifact at runtime.
+
+Continue with:
+
+- [Integration Deep Dive](integration-deep-dive.md) for production lifecycle guidance.
+- [Loading and Building Stemmers](programmatic-loading-and-building.md) for all loading APIs.
+- [Built-in Languages](built-in-languages.md) for bundled resources and dictionary locations.
+- [Benchmarking](benchmarking.md) for speed and quality interpretation.
--- a/docs/index.md
+++ b/docs/index.md
@@ -30,8 +30,10 @@ Radixor is intended for teams that require consistent stemming quality at scale,

 ## Start here

- Read [Quick Start](quick-start.md) for immediate implementation guidance.
- Use [Programmatic Usage](programmatic-usage.md) for application integration patterns.
- Review [Benchmarking](benchmarking.md) for reproducible performance methodology.
+- Read [Fast Track](fast-track.md) when you want the shortest path to a working bundled stemmer.
+- Use [Integration Deep Dive](integration-deep-dive.md) when you are wiring Radixor into a real application or search pipeline.
+- Read [Quick Start](quick-start.md) for the broader developer walkthrough after the first result works.
+- Use [Built-in Languages](built-in-languages.md) to find the bundled dictionaries exposed by Radixor.
+- Review [Benchmarking](benchmarking.md) and [Benchmark Results](benchmarks/index.md) for reproducible performance and quality methodology.
 - Open [CI Reports](reports.md) to inspect published build artifacts and quality metrics.
 - See the historical paper: [*Lemmatizer for Document Information Retrieval Systems in JAVA*](https://www.researchgate.net/publication/221512865_Lemmatizer_for_Document_Information_Retrieval_Systems_in_JAVA).
--- a/docs/integration-deep-dive.md
+++ b/docs/integration-deep-dive.md
@@ -0,0 +1,235 @@
+# Integration Deep Dive
+
+This page explains how to integrate Radixor into a real Java application after the first
+fast-track experiment works. It covers dependencies, bundled dictionaries, runtime lifecycle,
+deployment artifacts, and the decisions that matter in search or text-processing systems.
+
+## Integration Model
+
+Radixor has two separate phases:
+
+| Phase | Work | Typical location |
+| --- | --- | --- |
+| Preparation | Parse dictionaries, derive patch commands, reduce and contract the trie, optionally persist a binary artifact. | Build pipeline, packaging job, admin tool, or startup for small services. |
+| Runtime | Load an immutable compiled trie, look up patch commands, apply them to tokens. | Search indexing, query processing, text normalization, enrichment pipelines. |
+
+The practical rule is simple: compile rarely, stem often.
+
+For production systems, prefer a startup-owned or dependency-injected singleton
+`FrequencyTrie<CompiledPatchCommand>` per language/configuration. The trie is immutable after
+construction and is suitable for concurrent reads.
+
+## Dependency Coordinates
+
+The Maven coordinates are:
+
+```text
+org.egothor:radixor
+```
+
+Gradle:
+
+```kotlin
+dependencies {
+    implementation("org.egothor:radixor:3.0.0")
+}
+```
+
+Maven:
+
+```xml
+<dependency>
+    <groupId>org.egothor</groupId>
+    <artifactId>radixor</artifactId>
+    <version>3.0.0</version>
+</dependency>
+```
+
+Replace `3.0.0` with the current release selected for your deployment.
+
+The core Java module is:
+
+```java
+module org.egothor.radixor;
+```
+
+A named consuming module declares:
+
+```java
+module example.search {
+    requires org.egothor.radixor;
+}
+```
+
+## Bundled Dictionaries
+
+Radixor ships bundled dictionaries inside the library artifact. The public API exposes them through:
+
+```java
+StemmerPatchTrieLoader.Language
+```
+
+The physical resources are packaged as compressed UTF-8 dictionaries under resource directories
+such as:
+
+```text
+us_uk/stemmer.gz
+de_de/stemmer.gz
+fr_fr/stemmer.gz
+pl_pl/stemmer.gz
+```
+
+Treat those resource paths as implementation details. Application code should load bundled
+dictionaries through `StemmerPatchTrieLoader.Language`, because the enum also carries the language
+metadata needed for correct traversal.
+
+See [Built-in Languages](built-in-languages.md) for the complete language list, writing-direction
+notes, and links to per-language benchmark pages.
+
+## Minimal Service Wrapper
+
+A small service wrapper keeps loading, null handling, and fallback behavior in one place.
+
+```java
+import java.io.IOException;
+import java.util.Objects;
+
+import org.egothor.stemmer.CompiledPatchCommand;
+import org.egothor.stemmer.FrequencyTrie;
+import org.egothor.stemmer.ReductionMode;
+import org.egothor.stemmer.StemmerPatchTrieLoader;
+
+public final class RadixorStemmerService {
+
+    private final FrequencyTrie<CompiledPatchCommand> trie;
+
+    public RadixorStemmerService(final StemmerPatchTrieLoader.Language language) throws IOException {
+        this.trie = StemmerPatchTrieLoader.loadCompiled(
+                Objects.requireNonNull(language, "language"),
+                true,
+                ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS);
+    }
+
+    public String stem(final String token) {
+        final String checkedToken = Objects.requireNonNull(token, "token");
+        final CompiledPatchCommand command = trie.get(checkedToken);
+        return command == null ? checkedToken : command.apply(checkedToken);
+    }
+}
+```
+
+The fallback behavior preserves the original token when the trie has no patch command for it. That
+is usually the right default for search normalization, because unknown tokens should remain
+searchable.
+
+## Production Artifact Workflow
+
+For a controlled deployment, compile once and deploy the binary artifact:
+
+1. choose a bundled or custom dictionary,
+2. optionally extend it with domain vocabulary,
+3. compile a contracted trie,
+4. persist it as `.radixor.gz`,
+5. deploy that artifact with the application,
+6. load it with `StemmerPatchTrieLoader.loadBinaryCompiled(...)`.
+
+Runtime loading then avoids dictionary parsing and preparation-time memory pressure.
+
+```java
+import java.io.IOException;
+import java.nio.file.Path;
+
+import org.egothor.stemmer.CompiledPatchCommand;
+import org.egothor.stemmer.FrequencyTrie;
+import org.egothor.stemmer.StemmerPatchTrieLoader;
+
+public final class BinaryStemmerLoader {
+
+    private BinaryStemmerLoader() {
+        throw new AssertionError("No instances.");
+    }
+
+    public static FrequencyTrie<CompiledPatchCommand> loadEnglish() throws IOException {
+        return StemmerPatchTrieLoader.loadBinaryCompiled(Path.of("stemmers", "english.radixor.gz"));
+    }
+}
+```
+
+Use [CLI Compilation](cli-compilation.md) for command-line artifact creation, or
+[Extending and Persisting Compiled Tries](programmatic-extending-and-persistence.md) for
+programmatic artifact generation.
+
+## Search Pipeline Guidance
+
+Use Radixor consistently across indexing and querying:
+
+- choose one language dictionary per field or per analysis chain,
+- apply the same token normalization before stemming on both sides,
+- keep the compiled trie in memory and reuse it,
+- use `get(...)` for a single preferred stem,
+- use `getAll(...)` when a retrieval model benefits from preserving alternatives,
+- version custom `.radixor.gz` artifacts with the application or index schema.
+
+For multilingual content, do not run every token through every language. Route text by field,
+document metadata, or language detection before stemming.
+
+## Choosing Bundled Versus Custom Dictionaries
+
+Start with bundled dictionaries when:
+
+- the language is supported,
+- the application needs a strong baseline quickly,
+- the vocabulary is general-purpose,
+- the team is evaluating Radixor or building an initial integration.
+
+Use custom or extended dictionaries when:
+
+- domain vocabulary changes search quality,
+- product names, technical terms, legal terms, or biomedical terms must be preserved consistently,
+- stemming behavior must be curated and reviewed,
+- a release process needs a versioned lexical artifact.
+
+The dictionary format is intentionally simple and documented in
+[Dictionary Format](dictionary-format.md). Contribution standards are described in
+[Contributing Dictionaries](contributing-dictionaries.md).
+
+## Performance Practices
+
+The hot path should be only:
+
+```text
+token -> trie lookup -> compiled command application -> stem
+```
+
+Avoid these patterns in production request paths:
+
+- loading or compiling dictionaries per request,
+- applying serialized patch strings repeatedly instead of `CompiledPatchCommand`,
+- rebuilding tries for short-lived batches,
+- mixing different stemmer configurations between indexing and querying,
+- interpreting speed without checking exact-root quality.
+
+The current benchmark documentation separates methodology, corpora, environment, and language
+results so performance claims remain auditable. Start with [Benchmarking](benchmarking.md), then
+use [Benchmark Results](benchmarks/index.md) for the detailed reference tree.
+
+## Operational Checklist
+
+Before production rollout:
+
+- dependency version is pinned,
+- language resource and reduction mode are documented,
+- indexing and query pipelines use the same stemming configuration,
+- custom artifacts are versioned and reproducible,
+- fallback behavior for unknown tokens is explicit,
+- benchmark expectations are read together with quality metrics,
+- CI includes at least a smoke test that stems representative project vocabulary.
+
+## Related Pages
+
+- [Fast Track](fast-track.md)
+- [Quick Start](quick-start.md)
+- [Built-in Languages](built-in-languages.md)
+- [Programmatic Usage](programmatic-usage.md)
+- [CLI Compilation](cli-compilation.md)
+- [Benchmarking](benchmarking.md)
--- a/docs/lookup-edge-optimization.md
+++ b/docs/lookup-edge-optimization.md
@@ -6,7 +6,10 @@ Compiled trie nodes (`CompiledNode`) use three lookup strategies when resolving
 2. linear scan for very small child counts,
 3. binary search over sorted edge labels.

-This page explains the dense path, what `maxExpandedIndex` controls, and how to tune it.
+This page explains the dense path, what `maxExpandedIndex` controls, and how to tune it. These
+edge lookup strategies operate after trie reduction and uniform-subtree contraction. If lookup
+reaches an accepting contracted leaf, no child edge search is needed for the remaining input
+characters.

 ## Runtime model of one node

@@ -75,7 +78,7 @@ The value never changes lookup semantics. It only changes the in-memory structur
 This threshold is **not** stored in `TrieMetadata`.

 - The binary format stores only trie payload and semantic metadata (`reduction`, `traversal`,
-  case/diacritic settings, and stream version).
+  case/diacritic settings, contraction settings, and stream version).
 - `maxExpandedIndex` is chosen when materializing nodes in memory.
 - You can therefore keep one persisted artifact and load it with different in-memory
  trade-offs depending on deployment constraints.
@@ -135,15 +138,17 @@ At artifact load time, you can tune the same trade-off independently of persiste
 ```java
 import java.nio.file.Path;

+import org.egothor.stemmer.CompiledPatchCommand;
+import org.egothor.stemmer.FrequencyTrie;
 import org.egothor.stemmer.StemmerPatchTrieLoader;

-var defaultLookup = StemmerPatchTrieLoader.loadBinary(
+final FrequencyTrie<CompiledPatchCommand> defaultLookup = StemmerPatchTrieLoader.loadBinaryCompiled(
        Path.of("stemmers", "english.radixor.gz"));

-var fastLookup = StemmerPatchTrieLoader.loadBinary(
+final FrequencyTrie<CompiledPatchCommand> fastLookup = StemmerPatchTrieLoader.loadBinaryCompiled(
        Path.of("stemmers", "english.radixor.gz"), 1024);

-var compactLookup = StemmerPatchTrieLoader.loadBinary(
+final FrequencyTrie<CompiledPatchCommand> compactLookup = StemmerPatchTrieLoader.loadBinaryCompiled(
        Path.of("stemmers", "english.radixor.gz"), 0);
 ```

--- a/docs/migration-and-backward-compatibility.md
+++ b/docs/migration-and-backward-compatibility.md
@@ -0,0 +1,155 @@
+# Migration and Backward Compatibility
+
+This page describes the migration from repeated serialized patch-command application to compiled patch commands.
+
+## Summary
+
+Radixor patch commands are still encoded as compact strings when dictionaries are built and persisted. That serialized form remains the interchange format used by textual dictionaries, binary artifacts, and compilation tooling.
+
+Runtime stemming should no longer repeatedly apply those serialized strings directly. Since 2.3.0, the String-based patch application API is deprecated. Code that stems live input should load or create `CompiledPatchCommand` values and reuse them. The deprecated API remains available for compatibility during the transition, but applications should migrate before 3.0.0.
+
+The reason is performance. The old API parses the serialized P-command every time it is applied. `CompiledPatchCommand` parses it once and stores a concrete immutable command object, so repeated stemming avoids the same analysis work.
+
+## Deprecated Runtime APIs
+
+The following API family is kept for source compatibility but is no longer the preferred runtime path:
+
+- `PatchCommandEncoder.apply(String, String)`
+- `PatchCommandEncoder.apply(String, String, WordTraversalDirection)`
+- `PatchCommandEncoder.applyTo(..., String, WordTraversalDirection, ...)`
+- `PatchCommandEncoder.applyWithConfiguredDirection(String, String)`
+- `StemmerPatchTrieLoader.load(...)` overloads returning `FrequencyTrie<String>`
+- `StemmerPatchTrieLoader.loadBinary(...)` overloads returning `FrequencyTrie<String>`
+
+Use the compiled equivalents for runtime stemming:
+
+- `CompiledPatchCommand.compile(String, WordTraversalDirection)`
+- `PatchCommandEncoder.compile(String)`
+- `PatchCommandEncoder.compile(String, WordTraversalDirection)`
+- `StemmerPatchTrieLoader.loadCompiled(...)`
+- `StemmerPatchTrieLoader.loadBinaryCompiled(...)`
+
+## Loading A Text Dictionary
+
+Old runtime code:
+
+```java
+Path dictionary = Path.of("dictionary.txt");
+ReductionSettings settings = ReductionSettings.withDefaults(
+        ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS);
+FrequencyTrie<String> trie = StemmerPatchTrieLoader.load(dictionary, true, settings);
+
+String word = "running";
+String patch = trie.get(word);
+String stem = patch == null
+        ? word
+        : PatchCommandEncoder.apply(word, patch, trie.traversalDirection());
+```
+
+New runtime code:
+
+```java
+Path dictionary = Path.of("dictionary.txt");
+ReductionSettings settings = ReductionSettings.withDefaults(
+        ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS);
+FrequencyTrie<CompiledPatchCommand> trie = StemmerPatchTrieLoader.loadCompiled(dictionary, true, settings);
+
+String word = "running";
+CompiledPatchCommand patch = trie.get(word);
+String stem = patch == null ? word : patch.apply(word);
+```
+
+## Loading A Binary Artifact
+
+Old runtime code:
+
+```java
+FrequencyTrie<String> trie = StemmerPatchTrieLoader.loadBinary(Path.of("us-uk.radixor.gz"));
+
+String word = "studies";
+String patch = trie.get(word);
+String stem = patch == null
+        ? word
+        : PatchCommandEncoder.apply(word, patch, trie.traversalDirection());
+```
+
+New runtime code:
+
+```java
+FrequencyTrie<CompiledPatchCommand> trie =
+        StemmerPatchTrieLoader.loadBinaryCompiled(Path.of("us-uk.radixor.gz"));
+
+String word = "studies";
+CompiledPatchCommand patch = trie.get(word);
+String stem = patch == null ? word : patch.apply(word);
+```
+
+Existing binary artifacts remain readable. `loadBinaryCompiled(...)` reads the stored serialized patch strings and compiles them during load setup, before live stemming begins.
+
+## Manual Patch Encoding
+
+Encoding still produces a serialized patch command because that is the compact stored representation:
+
+```java
+PatchCommandEncoder encoder = PatchCommandEncoder.builder().build();
+String patch = encoder.encode("running", "run");
+```
+
+Old repeated application:
+
+```java
+String stem = PatchCommandEncoder.apply("running", patch);
+```
+
+New repeated application:
+
+```java
+CompiledPatchCommand compiled = encoder.compile(patch);
+String stem = compiled.apply("running");
+```
+
+## Caller-Owned Output Buffers
+
+Old buffer-oriented code:
+
+```java
+char[] output = new char[32];
+int length = PatchCommandEncoder.applyTo(
+        "running",
+        patch,
+        WordTraversalDirection.BACKWARD,
+        output,
+        0,
+        output.length);
+```
+
+New buffer-oriented code:
+
+```java
+CompiledPatchCommand compiled = CompiledPatchCommand.compile(patch, WordTraversalDirection.BACKWARD);
+char[] output = new char[32];
+int length = compiled.applyTo("running", output, 0, output.length);
+```
+
+Both APIs return `CompiledPatchCommand.APPLY_INSUFFICIENT_CAPACITY` when the caller-owned output range is too small.
+
+## Compatibility Rules
+
+Serialized patch strings remain part of the dictionary and artifact format. The deprecation is about repeated runtime application of serialized strings, not about the stored representation itself.
+
+Compatibility tests may continue to exercise the deprecated API to prove that old artifacts and source code still work during the transition. New production code, examples, and benchmark runtime paths should use `CompiledPatchCommand`.
+
+The command-line compiler still writes artifacts containing serialized patch commands. Runtime loaders can expose those commands as compiled immutable objects through `loadCompiled(...)` and `loadBinaryCompiled(...)`.
+
+## Contracted Trie Artifacts
+
+Current compiled loaders and freshly written binary artifacts can use contracted compiled tries.
+Contraction replaces a subtree with an accepting leaf when every reachable entry below that subtree
+selects the same preferred patch command. This changes the physical trie shape and the binary
+stream version, but it does not change the serialized patch-command language.
+
+Existing binary artifacts remain readable through the compatibility reader. To obtain the
+contracted runtime representation, rebuild the artifact with the current compiler or load the
+source dictionary through the current `loadCompiled(...)` APIs. Applications that only consume
+`CompiledPatchCommand` values through `get()` and `apply(...)` do not need code changes for this
+optimization.
--- a/docs/programmatic-extending-and-persistence.md
+++ b/docs/programmatic-extending-and-persistence.md
@@ -94,7 +94,8 @@ This model works especially well when domain-specific extensions are added in la
 After loading a compiled artifact, applications can inspect the persisted build descriptor directly:

 ```java
-final FrequencyTrie<String> trie = StemmerPatchTrieLoader.loadBinary("build/stemmers/cs_cz.dat.gz");
+final FrequencyTrie<CompiledPatchCommand> trie =
+        StemmerPatchTrieLoader.loadBinaryCompiled("build/stemmers/cs_cz.dat.gz");
 final TrieMetadata metadata = trie.metadata();

 System.out.println(metadata.formatVersion());
--- a/docs/programmatic-loading-and-building.md
+++ b/docs/programmatic-loading-and-building.md
@@ -4,11 +4,12 @@ This document explains how to acquire a compiled Radixor stemmer in Java.

 ## Load a bundled language dictionary

-Bundled language resources are simple to use and compile directly into a `FrequencyTrie<String>` during loading.
+Bundled language resources are simple to use and compile directly into a `FrequencyTrie<CompiledPatchCommand>` during loading.

 ```java
 import java.io.IOException;

+import org.egothor.stemmer.CompiledPatchCommand;
 import org.egothor.stemmer.FrequencyTrie;
 import org.egothor.stemmer.ReductionMode;
 import org.egothor.stemmer.StemmerPatchTrieLoader;
@@ -20,7 +21,7 @@ public final class BundledLanguageExample {
    }

    public static void main(final String[] arguments) throws IOException {
-        final FrequencyTrie<String> trie = StemmerPatchTrieLoader.load(
+        final FrequencyTrie<CompiledPatchCommand> trie = StemmerPatchTrieLoader.loadCompiled(
                StemmerPatchTrieLoader.Language.US_UK,
                true,
                ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS);
@@ -30,6 +31,11 @@ public final class BundledLanguageExample {

 The `storeOriginal` flag controls whether the canonical stem is inserted as a no-op patch entry for the stem itself.

+Bundled `loadCompiled(...)` entry points build the runtime trie with the same contracted
+representation used by the published benchmarks. During compilation, uniform preferred-command
+subtrees are collapsed into accepting leaves, so lookup can stop before consuming the entire input
+when the remaining characters cannot change the selected patch command.
+
 ## Load a textual dictionary

 Loading from a dictionary file follows the same preparation model as bundled resources, but the source comes from your own file or path. The input may be plain UTF-8 text or GZip-compressed UTF-8 text; the loader detects GZip data from the stream header. The textual format is tab-separated values, meaning that columns are separated by the tab character. Each non-empty logical line starts with the stem column and may contain zero or more variant columns. Input case normalization is controlled by `CaseProcessingMode` (default: `LOWERCASE_WITH_LOCALE_ROOT`), trailing remarks introduced by `#` or `//` are ignored, and dictionary items containing embedded whitespace are currently ignored with warning-level diagnostics.
@@ -38,6 +44,7 @@ Loading from a dictionary file follows the same preparation model as bundled res
 import java.io.IOException;
 import java.nio.file.Path;

+import org.egothor.stemmer.CompiledPatchCommand;
 import org.egothor.stemmer.FrequencyTrie;
 import org.egothor.stemmer.ReductionMode;
 import org.egothor.stemmer.ReductionSettings;
@@ -50,7 +57,7 @@ public final class LoadTextDictionaryExample {
    }

    public static void main(final String[] arguments) throws IOException {
-        final FrequencyTrie<String> trie = StemmerPatchTrieLoader.load(
+        final FrequencyTrie<CompiledPatchCommand> trie = StemmerPatchTrieLoader.loadCompiled(
                Path.of("data", "stemmer.tsv"),
                true,
                ReductionSettings.withDefaults(
@@ -59,7 +66,12 @@ public final class LoadTextDictionaryExample {
 }
 ```

-Additional `StemmerPatchTrieLoader.load(...)` overloads let callers provide explicit `WordTraversalDirection`, `CaseProcessingMode`, `DiacriticProcessingMode`, or a complete `TrieMetadata` instance. Use those overloads when a custom dictionary must be compiled with forward traversal for right-to-left languages, case-sensitive keys, or diacritic stripping.
+Additional `StemmerPatchTrieLoader.loadCompiled(...)` overloads let callers provide explicit `WordTraversalDirection`, `CaseProcessingMode`, `DiacriticProcessingMode`, or a complete `TrieMetadata` instance. Use those overloads when a custom dictionary must be compiled with forward traversal for right-to-left languages, case-sensitive keys, or diacritic stripping.
+
+When `ReductionSettings` are supplied through these compiled loader APIs, uniform-subtree
+contraction is still enabled as an internal pre-reduction step. The public `ReductionMode` remains
+the semantic policy for subtree equivalence after that contraction has removed regions whose
+preferred command is already uniform.

 ## Load a compiled binary artifact

@@ -69,6 +81,7 @@ Binary loading is typically the preferred runtime path because it avoids reparsi
 import java.io.IOException;
 import java.nio.file.Path;

+import org.egothor.stemmer.CompiledPatchCommand;
 import org.egothor.stemmer.FrequencyTrie;
 import org.egothor.stemmer.StemmerPatchTrieLoader;

@@ -79,7 +92,7 @@ public final class LoadBinaryExample {
    }

    public static void main(final String[] arguments) throws IOException {
-        final FrequencyTrie<String> trie = StemmerPatchTrieLoader.loadBinary(
+        final FrequencyTrie<CompiledPatchCommand> trie = StemmerPatchTrieLoader.loadBinaryCompiled(
                Path.of("stemmers", "english.radixor.gz"));
    }
 }
@@ -96,6 +109,7 @@ at load time. This does not change persisted metadata, only the materialized in-
 import java.io.IOException;
 import java.nio.file.Path;

+import org.egothor.stemmer.CompiledPatchCommand;
 import org.egothor.stemmer.FrequencyTrie;
 import org.egothor.stemmer.StemmerPatchTrieLoader;

@@ -106,14 +120,14 @@ public final class LoadBinaryWithDenseLookupExample {
    }

    public static void main(final String[] arguments) throws IOException {
-        final FrequencyTrie<String> balanced = StemmerPatchTrieLoader.loadBinary(
+        final FrequencyTrie<CompiledPatchCommand> balanced = StemmerPatchTrieLoader.loadBinaryCompiled(
                Path.of("stemmers", "english.radixor.gz"));

-        final FrequencyTrie<String> fast = StemmerPatchTrieLoader.loadBinary(
+        final FrequencyTrie<CompiledPatchCommand> fast = StemmerPatchTrieLoader.loadBinaryCompiled(
                Path.of("stemmers", "english.radixor.gz"),
                1024);

-        final FrequencyTrie<String> compact = StemmerPatchTrieLoader.loadBinary(
+        final FrequencyTrie<CompiledPatchCommand> compact = StemmerPatchTrieLoader.loadBinaryCompiled(
                Path.of("stemmers", "english.radixor.gz"),
                0);
    }
--- a/docs/programmatic-querying-and-ambiguity.md
+++ b/docs/programmatic-querying-and-ambiguity.md
@@ -10,7 +10,7 @@ This document explains how a compiled Radixor trie is queried and how ambiguity

 ```java
 final String word = "running";
-final String patch = trie.get(word);
+final CompiledPatchCommand patch = trie.get(word);
 ```

 ### `getAll(...)`: ordered local values
@@ -18,7 +18,7 @@ final String patch = trie.get(word);
 `FrequencyTrie.getAll(String)` returns all values stored at the addressed node, ordered by descending frequency using the same deterministic tie-breaking rules. The returned array is a defensive copy. If the key is missing or has no local values, an empty array is returned.

 ```java
-final String[] patches = trie.getAll("axes");
+final CompiledPatchCommand[] patches = trie.getAll("axes");
 ```

 ### `getEntries(...)`: values with counts
@@ -28,9 +28,10 @@ final String[] patches = trie.getAll("axes");
 ```java
 import java.util.List;

+import org.egothor.stemmer.CompiledPatchCommand;
 import org.egothor.stemmer.ValueCount;

-final List<ValueCount<String>> entries = trie.getEntries("axes");
+final List<ValueCount<CompiledPatchCommand>> entries = trie.getEntries("axes");
 ```

 ### Visitor lookup for hot paths
@@ -55,31 +56,29 @@ trie.getAllNormalized(token, 0, token.length, (patch, count, rank) -> {

 `getAllNormalized(...)` bypasses `caseProcessingMode` and `diacriticProcessingMode`; callers are responsible for supplying canonical input. `maxResults == 0` visits nothing, negative values are rejected, and a sink returning `false` stops iteration after the current callback.

-## Apply patch commands
+## Apply compiled patch commands

-A patch command is not the final stem. It must be applied to the original input token. `PatchCommandEncoder.apply(source, patchCommand)` performs that transformation directly on the serialized command format. If the source is `null`, the method returns `null`. If the patch is `null`, empty, or malformed in compatibility-relevant ways, the original source word is preserved. Equal source and target words are represented by the canonical no-op patch.
+A patch command is not the final stem. It must be applied to the original input token. Runtime code should use `CompiledPatchCommand`, which parses the stored patch-command representation once during setup and then applies the concrete immutable command repeatedly.

 ```java
-import org.egothor.stemmer.PatchCommandEncoder;
+import org.egothor.stemmer.CompiledPatchCommand;

 final String word = "running";
-final String patch = trie.get(word);
-final String stem = PatchCommandEncoder.apply(word, patch);
+final CompiledPatchCommand patch = trie.get(word);
+final String stem = patch == null ? word : patch.apply(word);
 ```

 Hot paths can apply a patch into caller-owned character storage:

 ```java
 final char[] output = new char[32];
-final int produced = PatchCommandEncoder.applyTo(
+final int produced = patch.applyTo(
        word,
-        patch,
-        trie.traversalDirection(),
        output,
        0,
        output.length);

-if (produced != PatchCommandEncoder.APPLY_INSUFFICIENT_CAPACITY) {
+if (produced != CompiledPatchCommand.APPLY_INSUFFICIENT_CAPACITY) {
    final String stem = new String(output, 0, produced);
 }
 ```
@@ -90,12 +89,14 @@ For multiple candidates:

 ```java
 final String word = "axes";
-for (final String patch : trie.getAll(word)) {
-    final String stem = PatchCommandEncoder.apply(word, patch);
+for (final CompiledPatchCommand patch : trie.getAll(word)) {
+    final String stem = patch.apply(word);
    System.out.println(word + " -> " + stem + " (" + patch + ")");
 }
 ```

+The historical `PatchCommandEncoder.apply(...)` API still exists for compatibility with code that directly handles serialized patch-command strings, but it is deprecated because it reparses the command on every call. See [Migration and Backward Compatibility](migration-and-backward-compatibility.md) for the old and new forms side by side.
+
 ## Understand reduction modes

 Reduction mode determines how mutable subtrees are merged during compilation. All modes operate on full subtree semantics rather than only on local node content.
--- a/docs/programmatic-usage.md
+++ b/docs/programmatic-usage.md
@@ -24,6 +24,8 @@ This is why Radixor can generalize beyond explicitly listed forms and why compil

 The programmatic API is easier to understand when split by developer task:

+- [Fast Track](fast-track.md) gives the shortest dependency-to-first-stem path for a new Java project.
+- [Integration Deep Dive](integration-deep-dive.md) explains production integration, deployment artifacts, search-pipeline usage, and operational decisions.
 - [Loading and Building Stemmers](programmatic-loading-and-building.md) explains how to acquire a compiled stemmer from bundled resources, textual dictionaries, binary artifacts, or direct builder usage.
 - [Lookup Edge Optimization](lookup-edge-optimization.md) explains dense child lookup tuning and the speed/memory trade-off when materializing compiled tries.
 - [Querying and Ambiguity Handling](programmatic-querying-and-ambiguity.md) explains `get(...)`, `getAll(...)`, `getEntries(...)`, patch application, and the practical meaning of reduction modes.
@@ -35,19 +37,40 @@ The main types involved in programmatic usage are:

 - `FrequencyTrie.Builder<V>` for mutable construction and extension,
 - `FrequencyTrie<V>` for the compiled read-only trie,
- `PatchCommandEncoder` for creating and applying patch commands,
+- `PatchCommandEncoder` for creating serialized patch commands,
+- `CompiledPatchCommand` for repeated runtime patch application,
 - `StemmerPatchTrieLoader` for loading bundled or textual dictionaries,
 - `StemmerPatchTrieBinaryIO` for reading and writing compressed binary artifacts,
 - `FrequencyTrieBuilders` for reconstructing a mutable builder from a compiled trie,
 - `ReductionMode` and `ReductionSettings` for controlling compilation semantics.

+## Java module system (JPMS)
+
+The core artifact is published as an explicit JPMS module:
+
+```java
+module org.egothor.radixor;
+```
+
+A named consuming module uses:
+
+```java
+module example.consumer {
+    requires org.egothor.radixor;
+}
+```
+
+The core module is standalone and can be consumed directly as a normal Java module.
+
 ## Recommended reading order

 For most developers, the best order is:

-1. [Loading and Building Stemmers](programmatic-loading-and-building.md)
-2. [Querying and Ambiguity Handling](programmatic-querying-and-ambiguity.md)
-3. [Extending and Persisting Compiled Tries](programmatic-extending-and-persistence.md)
+1. [Fast Track](fast-track.md)
+2. [Integration Deep Dive](integration-deep-dive.md)
+3. [Loading and Building Stemmers](programmatic-loading-and-building.md)
+4. [Querying and Ambiguity Handling](programmatic-querying-and-ambiguity.md)
+5. [Extending and Persisting Compiled Tries](programmatic-extending-and-persistence.md)

 ## Next steps

--- a/docs/quick-start.md
+++ b/docs/quick-start.md
@@ -2,6 +2,10 @@

 This guide introduces the fastest practical path to using **Radixor**.

+If you are new to Radixor and want the shortest possible path to a first working stem, start with
+[Fast Track](fast-track.md). This Quick Start is a broader developer walkthrough: it introduces the
+main loading options, query methods, artifact workflow, and metadata model.
+
 Radixor separates preparation from runtime usage. Source dictionaries are used to derive patch commands and reduce them into a compact read-only trie. Runtime stemming then operates on that compiled structure rather than on the original dictionary text. A richer dictionary usually improves the quality and coverage of inferred transformations, including transformations that are applicable to words not explicitly present in the source material. The reduction step also removes a large amount of redundant lexical information, which is why very large dictionaries can still produce compact runtime artifacts. These artifacts can be persisted and loaded directly when needed.

 A practical workflow usually consists of two independent phases:
@@ -15,11 +19,12 @@ A compiled stemmer can be obtained in three common ways.

 ### Use a bundled language dictionary

-Radixor ships with bundled dictionaries for a set of supported languages. These resources are line-oriented dictionaries stored with the library and compiled into a `FrequencyTrie<String>` when loaded. The loader can also store the canonical stem itself as a no-op patch command. Compiled trie artifacts now persist self-describing metadata, including the traversal direction and compilation reduction settings used to build the artifact.
+Radixor ships with bundled dictionaries for a set of supported languages. These resources are line-oriented dictionaries stored with the library and compiled into a `FrequencyTrie<CompiledPatchCommand>` when loaded through the runtime API. The loader can also store the canonical stem itself as a no-op patch command. Compiled trie artifacts now persist self-describing metadata, including the traversal direction and compilation reduction settings used to build the artifact.

 ```java
 import java.io.IOException;

+import org.egothor.stemmer.CompiledPatchCommand;
 import org.egothor.stemmer.FrequencyTrie;
 import org.egothor.stemmer.ReductionMode;
 import org.egothor.stemmer.StemmerPatchTrieLoader;
@@ -31,7 +36,7 @@ public final class BundledStemmerExample {
    }

    public static void main(final String[] arguments) throws IOException {
-        final FrequencyTrie<String> trie = StemmerPatchTrieLoader.load(
+        final FrequencyTrie<CompiledPatchCommand> trie = StemmerPatchTrieLoader.loadCompiled(
                StemmerPatchTrieLoader.Language.US_UK,
                true,
                ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS);
@@ -49,6 +54,7 @@ Compiled stemmers can be stored as GZip-compressed binary artifacts and loaded d
 import java.io.IOException;
 import java.nio.file.Path;

+import org.egothor.stemmer.CompiledPatchCommand;
 import org.egothor.stemmer.FrequencyTrie;
 import org.egothor.stemmer.StemmerPatchTrieLoader;

@@ -59,7 +65,7 @@ public final class LoadBinaryStemmerExample {
    }

    public static void main(final String[] arguments) throws IOException {
-        final FrequencyTrie<String> trie = StemmerPatchTrieLoader.loadBinary(
+        final FrequencyTrie<CompiledPatchCommand> trie = StemmerPatchTrieLoader.loadBinaryCompiled(
                Path.of("stemmers", "english.radixor.gz"));

        System.out.println("Canonical node count: " + trie.size());
@@ -73,6 +79,7 @@ You can tune in-memory child lookup density at load time without changing the ar
 import java.io.IOException;
 import java.nio.file.Path;

+import org.egothor.stemmer.CompiledPatchCommand;
 import org.egothor.stemmer.FrequencyTrie;
 import org.egothor.stemmer.StemmerPatchTrieLoader;

@@ -83,10 +90,10 @@ public final class LoadBinaryStemmerExampleTuned {
    }

    public static void main(final String[] arguments) throws IOException {
-        final FrequencyTrie<String> fast = StemmerPatchTrieLoader.loadBinary(
+        final FrequencyTrie<CompiledPatchCommand> fast = StemmerPatchTrieLoader.loadBinaryCompiled(
                Path.of("stemmers", "english.radixor.gz"),
                1024);
-        final FrequencyTrie<String> compact = StemmerPatchTrieLoader.loadBinary(
+        final FrequencyTrie<CompiledPatchCommand> compact = StemmerPatchTrieLoader.loadBinaryCompiled(
                Path.of("stemmers", "english.radixor.gz"),
                128);

@@ -107,7 +114,7 @@ A dedicated CLI compilation workflow deserves its own focused page and should re

 ## 2. Use the compiled stemmer

-A compiled `FrequencyTrie<String>` stores patch commands, not final stems. Querying therefore has two steps:
+A compiled `FrequencyTrie<CompiledPatchCommand>` stores patch commands, not final stems. Querying therefore has two steps:

 1. retrieve one or more patch commands from the trie,
 2. apply each patch command to the original input word.
@@ -121,8 +128,8 @@ Use `get(...)` when the application needs a single preferred transformation.
 ```java
 import java.io.IOException;

+import org.egothor.stemmer.CompiledPatchCommand;
 import org.egothor.stemmer.FrequencyTrie;
-import org.egothor.stemmer.PatchCommandEncoder;
 import org.egothor.stemmer.ReductionMode;
 import org.egothor.stemmer.StemmerPatchTrieLoader;

@@ -133,14 +140,14 @@ public final class SingleStemExample {
    }

    public static void main(final String[] arguments) throws IOException {
-        final FrequencyTrie<String> trie = StemmerPatchTrieLoader.load(
+        final FrequencyTrie<CompiledPatchCommand> trie = StemmerPatchTrieLoader.loadCompiled(
                StemmerPatchTrieLoader.Language.US_UK,
                true,
                ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS);

        final String word = "running";
-        final String patch = trie.get(word);
-        final String stem = PatchCommandEncoder.apply(word, patch);
+        final CompiledPatchCommand patch = trie.get(word);
+        final String stem = patch == null ? word : patch.apply(word);

        System.out.println(word + " -> " + stem + " (" + patch + ")");
    }
@@ -153,10 +160,10 @@ Use `getAll(...)` when the application should preserve ambiguity instead of coll

 ```java
 final String word = "axes";
-final String[] patches = trie.getAll(word);
+final CompiledPatchCommand[] patches = trie.getAll(word);

-for (final String patch : patches) {
-    final String stem = PatchCommandEncoder.apply(word, patch);
+for (final CompiledPatchCommand patch : patches) {
+    final String stem = patch.apply(word);
    System.out.println(word + " -> " + stem + " (" + patch + ")");
 }
 ```
@@ -168,11 +175,12 @@ For diagnostics or advanced ranking logic, use `getEntries(...)` to obtain value
 ```java
 import java.util.List;

+import org.egothor.stemmer.CompiledPatchCommand;
 import org.egothor.stemmer.ValueCount;

-final List<ValueCount<String>> entries = trie.getEntries("axes");
+final List<ValueCount<CompiledPatchCommand>> entries = trie.getEntries("axes");

-for (final ValueCount<String> entry : entries) {
+for (final ValueCount<CompiledPatchCommand> entry : entries) {
    System.out.println(entry.value() + " -> " + entry.count());
 }
 ```
@@ -210,7 +218,11 @@ public final class ExtendCompiledStemmerExample {
                String[]::new,
                settings);

-        builder.put("microservices", "Na");
+        final PatchCommandEncoder encoder = PatchCommandEncoder.builder()
+                .traversalDirection(compiledTrie.traversalDirection())
+                .build();
+
+        builder.put("microservices", encoder.encode("microservices", "microservice"));

        final FrequencyTrie<String> updatedTrie = builder.build();

--- a/docs/reduction-semantics.md
+++ b/docs/reduction-semantics.md
@@ -103,6 +103,24 @@ This turns reduction into a canonicalization process:

 That is how Radixor eliminates duplicated equivalent subtrees.

+## Uniform-subtree contraction
+
+Radixor performs one additional internal reduction before each public reduction mode is applied.
+When all reachable entries below a subtree have the same preferred patch command, the subtree is
+contracted into an accepting leaf for that command.
+
+This optimization is deliberately narrower than the public reduction modes:
+
+- it is based on preferred `get()` behavior,
+- it does not depend on child edge shape once the preferred command is uniform,
+- it removes lookup depth that cannot affect the selected command,
+- it preserves the standard single-result stemming path used by `StemmerPatchTrieLoader.loadCompiled(...)`.
+
+The effect is especially visible in large dictionary tries with many inflected forms that map to
+the same command class, such as no-op preservation or common suffix deletion. Runtime lookup can
+return the accepting leaf as soon as it is reached instead of traversing the remaining characters
+only to discover the same command deeper in the trie.
+
 ## Count aggregation and compiled state

 When multiple original build-time subtrees collapse into one canonical reduced node, local counts may be aggregated.
--- a/docs/reports.md
+++ b/docs/reports.md
@@ -42,15 +42,14 @@ They are useful for dependency inspection, downstream integration, compliance-or

 ## Benchmark outputs and badge metadata

-These resources expose benchmark results and generated badge metadata derived from the latest published build:
+These resources expose benchmark results and generated badge metadata derived from the latest published build. JMH benchmark reports are published as TXT and CSV files; the historical Porter comparison badge is no longer generated.

 - [JMH benchmark results (TXT)](https://leogalambos.github.io/Radixor/builds/latest/jmh/jmh-results.txt)
 - [JMH benchmark results (CSV)](https://leogalambos.github.io/Radixor/builds/latest/jmh/jmh-results.csv)
 - [Coverage badge metadata](https://leogalambos.github.io/Radixor/builds/latest/metrics/coverage-badge.json)
 - [Mutation badge metadata](https://leogalambos.github.io/Radixor/builds/latest/metrics/pitest-badge.json)
- [Benchmark badge metadata](https://leogalambos.github.io/Radixor/builds/latest/metrics/jmh-badge.json)

-The benchmark outputs provide direct access to the published JMH result files, while the badge metadata endpoints are intended for status surfaces such as the project README or other generated dashboards.
+The benchmark outputs provide direct access to the published JMH result files. Coverage and mutation badge metadata endpoints are intended for status surfaces such as the project README or other generated dashboards.

 ## Practical usage

--- a/docs/stemming-quality.md
+++ b/docs/stemming-quality.md
@@ -0,0 +1,82 @@
+# Stemming quality evaluation
+
+The explicit `stemmingQuality` analysis measures agreement between stemmer outputs and gold-standard equivalence classes represented by bundled multilingual dictionary rows. Dictionary text remains unchanged; reports and diagnostics use English.
+
+JMH adapters, registries, third-party versions, language mappings, and preparation remain in `src/jmh`. The evaluator, reports, audits, and tests reside in the standard `src/test` source set. The former `src/stemmingQualityTest` source set was removed, and neither analytical nor JMH classes enter the production JAR.
+
+## Language and adapter coverage
+
+The authoritative Radixor universe is the validated one-to-one reconciliation of `src/main/resources/*/stemmer.gz` and every `StemmerPatchTrieLoader.Language` value. All 20 current values have exactly one resource; no sentinel or alias is excluded. Radixor is evaluated for all 20 languages, independently of third-party support. Third-party combinations come only from explicit JMH adapter metadata. Unsupported combinations are documented and never fabricated as zero-valued rows.
+
+The expected matrix is constructed before evaluation from stemmer, language, dictionary mode, and supported output policy. Generation fails on missing, duplicate, unexpected, or stale keys.
+
+## Dictionary groups and modes
+
+Every usable parsed row is one gold-standard group. Exact duplicate strings are removed only within that row; identical forms in different rows remain distinct. `ALL_WORDS` preserves every valid form. `LOWERCASE_GROUPS_ONLY` excludes a complete group containing an uppercase or titlecase Unicode code point. Retained words are not lowercased or normalized by the evaluator.
+
+## Output policies
+
+`PRIMARY_OUTPUT` uses the deterministic JMH output and defines a strict partition.
+
+For multi-output adapters, `C(w)` is the immutable, sorted, exactly deduplicated candidate set. It is non-null, non-empty, contains no null, and contains the primary output. Radixor obtains alternatives through `getAll`. The repository's Morphologik lookups can return distinct lemma strings and are multi-output. Configured Hunspell filters can emit several stems at one token position. Other adapters emit only primary rows.
+
+`ANY_CANDIDATE` is an optimistic oracle-assisted pairwise upper bound. A same-group pair succeeds when its sets intersect. A cross-group pair is an error only when both sets are the same singleton; otherwise unequal candidates can be selected for that pair. Choices may vary between pairs and need not form one realizable global assignment.
+
+`ALL_CANDIDATES` activates every candidate. Two forms are related when their sets intersect, for both same-group and cross-group pairs. This relation can overlap and need not be transitive. A pair sharing several candidates is counted once.
+
+The evaluator verifies:
+
+```text
+ANY under <= PRIMARY under
+ALL under <= PRIMARY under
+ANY under = ALL under
+ANY over <= PRIMARY over
+ALL over >= PRIMARY over
+```
+
+## Pair definitions and efficient counting
+
+For `C2(n) = n(n-1)/2`:
+
+```text
+underPossible = sum_g C2(n_g)
+overPossible = C2(N) - sum_g C2(n_g)
+```
+
+Under-stemming counts unrelated same-group pairs. Over-stemming counts related cross-group pairs. Primary output uses global and per-group stem frequencies. Candidate sets are canonical signatures counted globally and per group. An inverted candidate-to-signature index discovers intersections, and signature pairs shared through several candidates are deduplicated. `ANY_CANDIDATE` over-stemming uses only equal singleton signatures. All pair arithmetic uses checked `long` operations; complete production word pairs are never enumerated.
+
+## Confusion and aggregate metrics
+
+```text
+TP = underPossible - underError
+FN = underError
+FP = overError
+TN = overPossible - overError
+```
+
+Under-stemming is `FN/(TP+FN)` and over-stemming is `FP/(TN+FP)`; their denominators differ. The CSV also publishes precision, recall, specificity, accuracy, balanced accuracy, F0.5, F1, F2, Jaccard, Fowlkes-Mallows, Matthews correlation coefficient, and pairwise error rate. F0.5 emphasizes precision and over-stemming, F1 balances precision and recall, and F2 emphasizes recall and under-stemming. Accuracy and error rate can be dominated by the large cross-group true-negative population. Metrics use raw counts, not rounded rates. Zero denominators produce `n/a` in Markdown and empty CSV fields.
+
+Only `PRIMARY_OUTPUT` receives partition metrics: Adjusted Rand Index, homogeneity, completeness, V-measure, and normalized mutual information with arithmetic-mean entropy normalization. Candidate policies remain inapplicable rather than being forced into artificial partitions.
+
+Micro summaries sum confusion counts before calculation. Macro summaries average defined language values and retain coverage counts. Common-language comparisons use the exact language intersection and never score unsupported languages as zero. Rankings are separated by policy and metric; the default F0.5 choice is navigation, not a universal scientific preference.
+
+Pearson and average-tie-rank Spearman reports use unrounded values and separate dictionary-mode and output-policy cohorts. Fewer than three observations, undefined inputs, and zero variance produce documented missing values. The reports provide reproducible data and make no automatic scientific conclusion.
+
+## Exact accuracy and pairwise under-stemming
+
+Exact textual accuracy and pairwise grouping use different denominators. One erroneous form in a 12-form group creates 11 erroneous pairs: with 88 singleton groups, exact accuracy can be 99% while pairwise under-stemming is `11/C2(12) = 16.666667%`. Singleton groups affect word accuracy but add no within-group pairs.
+
+## Running the analysis
+
+```bash
+./gradlew stemmingQuality
+./gradlew stemmingQuality -PstemmingQualityStemmer=Radixor -PstemmingQualityLanguage=DE_DE -PstemmingQualityMode=ALL_WORDS -PstemmingQualityAudit=true
+```
+
+Optional properties are `stemmingQualityLanguage`, `stemmingQualityStemmer`, `stemmingQualityMode`, `stemmingQualityOutputPolicy`, `stemmingQualityRankMetric`, `stemmingQualityAudit`, and `stemmingQualityAuditLimit`. Policies are `PRIMARY_OUTPUT`, `ANY_CANDIDATE`, and `ALL_CANDIDATES`. Filtered reports carry `-filtered` and cannot overwrite complete output.
+
+Generated files under `build/reports/stemming-quality/` include `stemming-quality.md`, `stemming-quality.csv`, `metric-correlations-pearson.csv`, `metric-correlations-spearman.csv`, and optional audit Markdown.
+
+## Limitations
+
+These measurements evaluate agreement with the available dictionary grouping. They do not capture every semantic, morphological, downstream, or dataset-specific property. `ANY_CANDIDATE` is optimistic and may not be globally realizable. `ALL_CANDIDATES` measures an overlap graph rather than a partition. Language coverage must remain visible in cross-stemmer comparisons. No single published metric establishes universal superiority; multiple metrics and their correlations are provided for transparent scientific assessment.
--- a/docs/test-taxonomy-and-filtering.md
+++ b/docs/test-taxonomy-and-filtering.md
@@ -145,6 +145,14 @@ defensive fallback in case of future tag drift.

 All examples use Gradle with JUnit Platform integration:

+- Default fast test run:
+
+```
+./gradlew test
+```
+
+The default `test` task excludes `slow` tests. Supplying `-DincludeTags` or `-PincludeTags` still excludes `slow` unless the include expression contains `slow`; supplying an explicit exclude expression replaces the default. Long-running bundled-dictionary compilation and full-language loading checks therefore run only through an explicit tag expression such as `-DincludeTags=slow` or a dedicated profile such as `ciSlow`.
+
 - Only unit tests:

 ```
@@ -154,7 +162,7 @@ All examples use Gradle with JUnit Platform integration:
 - Integration tests only:

 ```
-./gradlew test -DincludeTags=integration
+./gradlew test -DincludeTags=integration -DexcludeTags=slow
 ```

 - Only trie subsystem tests:
--- a/gradle.lockfile
+++ b/gradle.lockfile
@@ -13,7 +13,7 @@ net.jqwik:jqwik-time:1.9.3=jmhRuntimeClasspath,testCompileClasspath,testRuntimeC
 net.jqwik:jqwik-web:1.9.3=jmhRuntimeClasspath,testCompileClasspath,testRuntimeClasspath
 net.jqwik:jqwik:1.9.3=jmhRuntimeClasspath,testCompileClasspath,testRuntimeClasspath
 net.sf.jopt-simple:jopt-simple:4.9=pitest
-net.sf.jopt-simple:jopt-simple:5.0.4=jmh,jmhCompileClasspath,jmhRuntimeClasspath
+net.sf.jopt-simple:jopt-simple:5.0.4=jmh,jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
 net.sf.saxon:Saxon-HE:12.9=pmd
 net.sourceforge.pmd:pmd-ant:7.20.0=pmd
 net.sourceforge.pmd:pmd-core:7.20.0=pmd
@@ -22,9 +22,17 @@ org.antlr:antlr4-runtime:4.9.3=pmd
 org.antlr:stringtemplate:3.2.1=pitest
 org.apache.commons:commons-lang3:3.18.0=pitest
 org.apache.commons:commons-lang3:3.20.0=pmd
-org.apache.commons:commons-math3:3.6.1=jmh,jmhCompileClasspath,jmhRuntimeClasspath
+org.apache.commons:commons-math3:3.6.1=jmh,jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
 org.apache.commons:commons-text:1.14.0=pitest
+org.apache.lucene:lucene-analysis-common:10.5.0=jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
+org.apache.lucene:lucene-analysis-morfologik:10.5.0=jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
+org.apache.lucene:lucene-analysis-stempel:10.5.0=jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
+org.apache.lucene:lucene-core:10.5.0=jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
+org.apache.opennlp:opennlp-tools:2.5.4=jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
 org.apiguardian:apiguardian-api:1.1.2=jmhRuntimeClasspath,testCompileClasspath,testRuntimeClasspath
+org.carrot2:morfologik-fsa:2.1.9=jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
+org.carrot2:morfologik-polish:2.1.9=jmhRuntimeClasspath,stemmingQualityJmhRuntime
+org.carrot2:morfologik-stemming:2.1.9=jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
 org.checkerframework:checker-qual:3.52.1=pmd
 org.jacoco:org.jacoco.agent:0.8.14=jacocoAgent,jacocoAnt
 org.jacoco:org.jacoco.ant:0.8.14=jacocoAnt
@@ -41,10 +49,10 @@ org.junit:junit-bom:5.14.3=jmhRuntimeClasspath,testCompileClasspath,testRuntimeC
 org.mockito:mockito-core:5.23.0=jmhRuntimeClasspath,mockitoAgent,testCompileClasspath,testRuntimeClasspath
 org.mockito:mockito-junit-jupiter:5.23.0=jmhRuntimeClasspath,testCompileClasspath,testRuntimeClasspath
 org.objenesis:objenesis:3.3=jmhRuntimeClasspath,testRuntimeClasspath
-org.openjdk.jmh:jmh-core:1.37=jmh,jmhCompileClasspath,jmhRuntimeClasspath
-org.openjdk.jmh:jmh-generator-asm:1.37=jmh,jmhCompileClasspath,jmhRuntimeClasspath
-org.openjdk.jmh:jmh-generator-bytecode:1.37=jmh,jmhCompileClasspath,jmhRuntimeClasspath
-org.openjdk.jmh:jmh-generator-reflection:1.37=jmh,jmhCompileClasspath,jmhRuntimeClasspath
+org.openjdk.jmh:jmh-core:1.37=jmh,jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
+org.openjdk.jmh:jmh-generator-asm:1.37=jmh,jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
+org.openjdk.jmh:jmh-generator-bytecode:1.37=jmh,jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
+org.openjdk.jmh:jmh-generator-reflection:1.37=jmh,jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
 org.opentest4j:opentest4j:1.3.0=jmhRuntimeClasspath,testCompileClasspath,testRuntimeClasspath
 org.ow2.asm:asm-analysis:9.9.1=pitest
 org.ow2.asm:asm-commons:9.9=jacocoAnt
@@ -52,7 +60,7 @@ org.ow2.asm:asm-commons:9.9.1=pitest
 org.ow2.asm:asm-tree:9.9=jacocoAnt
 org.ow2.asm:asm-tree:9.9.1=pitest
 org.ow2.asm:asm-util:9.9.1=pitest
-org.ow2.asm:asm:9.0=jmh,jmhCompileClasspath,jmhRuntimeClasspath
+org.ow2.asm:asm:9.0=jmh,jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
 org.ow2.asm:asm:9.9=jacocoAnt
 org.ow2.asm:asm:9.9.1=pitest,pmd
 org.pcollections:pcollections:4.0.2=pmd
@@ -62,5 +70,7 @@ org.pitest:pitest-html-report:1.22.1=pitest
 org.pitest:pitest-junit5-plugin:1.2.3=pitest
 org.pitest:pitest:1.22.1=pitest
 org.slf4j:jul-to-slf4j:1.7.36=pmd
+org.slf4j:slf4j-api:2.0.17=jmhCompileClasspath,jmhRuntimeClasspath,stemmingQualityJmhRuntime
 org.xmlresolver:xmlresolver:5.3.3=pmd
+ua.net.nlp:morfologik-ukrainian-search:4.9.1=jmhRuntimeClasspath,stemmingQualityJmhRuntime
 empty=annotationProcessor,compileClasspath,cyclonedxBom,jmhAnnotationProcessor,mainPmdAuxClasspath,runtimeClasspath,testAnnotationProcessor
--- a/gradle/cistem-benchmarks.gradle
+++ b/gradle/cistem-benchmarks.gradle
@@ -0,0 +1,69 @@
+def cistemGoldStandardBaseUrl = 'https://raw.githubusercontent.com/LeonieWeissweiler/CISTEM/refs/heads/master/gold_standards'
+def cistemGoldStandardFiles = [
+        'goldstandard1.txt',
+        'goldstandard2.txt'
+]
+def cistemGoldStandardDownloadDirectory = layout.buildDirectory.dir('third-party/cistem-gold-standards')
+def cistemGoldStandardGeneratedResourcesDirectory = layout.buildDirectory.dir('generated/resources/cistem-gold-standards')
+
+def cistemGoldStandardDownloadedFiles = cistemGoldStandardFiles.collect { String fileName ->
+    cistemGoldStandardDownloadDirectory.map { it.file(fileName) }
+}
+
+tasks.register('downloadCistemGoldStandards') {
+    group = 'build setup'
+    description = 'Downloads benchmark-only CISTEM German gold standards.'
+
+    outputs.files(cistemGoldStandardDownloadedFiles)
+
+    doLast {
+        cistemGoldStandardFiles.each { String fileName ->
+            final File targetFile = cistemGoldStandardDownloadDirectory.get().file(fileName).asFile
+            targetFile.parentFile.mkdirs()
+
+            if (!targetFile.exists()) {
+                final URL sourceUrl = new URL("${cistemGoldStandardBaseUrl}/${fileName}")
+                try {
+                    sourceUrl.withInputStream { inputStream ->
+                        targetFile.withOutputStream { outputStream ->
+                            outputStream << inputStream
+                        }
+                    }
+                } catch (FileNotFoundException exception) {
+                    throw new GradleException(
+                            "Unable to download CISTEM gold standard ${fileName} from ${sourceUrl}.",
+                            exception)
+                }
+            }
+
+            if (targetFile.length() <= 0L) {
+                throw new GradleException("Downloaded CISTEM gold standard ${fileName} was empty.")
+            }
+        }
+    }
+}
+
+tasks.register('prepareCistemGoldStandardResources', Copy) {
+    group = 'build setup'
+    description = 'Copies benchmark-only CISTEM German gold standards into the JMH resource output.'
+
+    dependsOn(tasks.named('downloadCistemGoldStandards'))
+
+    from(cistemGoldStandardDownloadDirectory) {
+        include 'goldstandard1.txt'
+        include 'goldstandard2.txt'
+    }
+    into(cistemGoldStandardGeneratedResourcesDirectory)
+}
+
+sourceSets {
+    jmh {
+        resources {
+            srcDir(cistemGoldStandardGeneratedResourcesDirectory)
+        }
+    }
+}
+
+tasks.named('processJmhResources') {
+    dependsOn(tasks.named('prepareCistemGoldStandardResources'))
+}
--- a/gradle/hunspell-benchmarks.gradle
+++ b/gradle/hunspell-benchmarks.gradle
@@ -0,0 +1,121 @@
+import org.gradle.plugins.ide.eclipse.model.SourceFolder
+
+
+def hunspellDictionaryBaseUrl = 'https://raw.githubusercontent.com/wooorm/dictionaries/main/dictionaries'
+def hunspellDictionaryLanguages = [
+        en: 'English',
+        cs: 'Czech',
+        de: 'German',
+        es: 'Spanish',
+        fr: 'French',
+        nl: 'Dutch',
+        pl: 'Polish',
+        uk: 'Ukrainian'
+]
+def hunspellDownloadDirectory = layout.buildDirectory.dir('third-party/hunspell')
+def hunspellGeneratedResourcesDirectory = layout.buildDirectory.dir('generated/resources/hunspell')
+def hunspellGeneratedResourcesPath = provider {
+    project.relativePath(hunspellGeneratedResourcesDirectory.get().asFile)
+}
+def hunspellEclipseClasspathAttributes = [
+        gradle_scope        : 'jmh',
+        gradle_used_by_scope: 'jmh',
+        test                : 'true'
+]
+def hunspellIsAbsolutePath = { String path ->
+    path.startsWith('/') || path ==~ /^[A-Za-z]:[\\\/].*/
+}
+
+def hunspellDownloadedFiles = hunspellDictionaryLanguages.keySet().collectMany { String code ->
+    [
+            hunspellDownloadDirectory.map { it.file("${code}/index.aff") },
+            hunspellDownloadDirectory.map { it.file("${code}/index.dic") },
+            hunspellDownloadDirectory.map { it.file("${code}/license") }
+    ]
+}
+
+tasks.register('downloadHunspellBenchmarkDictionaries') {
+    group = 'build setup'
+    description = 'Downloads benchmark-only Hunspell dictionaries from wooorm/dictionaries.'
+
+    outputs.files(hunspellDownloadedFiles)
+
+    doLast {
+        hunspellDictionaryLanguages.each { String code, String displayName ->
+            ['index.aff', 'index.dic', 'license'].each { String fileName ->
+                final File targetFile = hunspellDownloadDirectory.get().file("${code}/${fileName}").asFile
+                targetFile.parentFile.mkdirs()
+
+                if (!targetFile.exists()) {
+                    final URL sourceUrl = new URL("${hunspellDictionaryBaseUrl}/${code}/${fileName}")
+                    try {
+                        sourceUrl.withInputStream { inputStream ->
+                            targetFile.withOutputStream { outputStream ->
+                                outputStream << inputStream
+                            }
+                        }
+                    } catch (FileNotFoundException exception) {
+                        throw new GradleException(
+                                "Unable to download Hunspell ${fileName} file for ${displayName} (${code}) from ${sourceUrl}.",
+                                exception)
+                    }
+                }
+
+                if (targetFile.length() <= 0L) {
+                    throw new GradleException("Downloaded Hunspell ${fileName} file for ${displayName} was empty.")
+                }
+            }
+        }
+    }
+}
+
+tasks.register('prepareHunspellBenchmarkResources', Copy) {
+    group = 'build setup'
+    description = 'Copies benchmark-only Hunspell dictionaries into the JMH resource output.'
+
+    dependsOn(tasks.named('downloadHunspellBenchmarkDictionaries'))
+
+    from(hunspellDownloadDirectory) {
+        include '**/index.aff'
+        include '**/index.dic'
+        include '**/license'
+        into 'hunspell'
+    }
+    into(hunspellGeneratedResourcesDirectory)
+}
+
+sourceSets {
+    jmh {
+        resources {
+            srcDir(hunspellGeneratedResourcesDirectory)
+        }
+    }
+}
+
+tasks.named('processJmhResources') {
+    dependsOn(tasks.named('prepareHunspellBenchmarkResources'))
+}
+
+eclipse {
+    classpath {
+        file {
+            whenMerged { classpath ->
+                String generatedPath = hunspellGeneratedResourcesPath.get()
+
+                classpath.entries.removeAll { entry ->
+                    entry.hasProperty('path') && (
+                            entry.path == generatedPath ||
+                            hunspellIsAbsolutePath(entry.path)
+                    )
+                }
+
+                SourceFolder hunspellEntry = new SourceFolder(generatedPath, null)
+                hunspellEntry.output = 'bin/jmh'
+                hunspellEclipseClasspathAttributes.each { String name, String value ->
+                    hunspellEntry.entryAttributes[name] = value
+                }
+                classpath.entries.add(hunspellEntry)
+            }
+        }
+    }
+}
--- a/gradle/lucene-benchmarks.gradle
+++ b/gradle/lucene-benchmarks.gradle
@@ -0,0 +1,223 @@
+import org.gradle.plugins.ide.eclipse.model.SourceFolder
+
+
+def luceneVersion = '10.5.0'
+def luceneRootRelativePath = 'third-party/lucene'
+def luceneSourceArtifacts = ['lucene-analysis-common', 'lucene-analyzers-common']
+def luceneSourceDirectory = layout.buildDirectory.dir("${luceneRootRelativePath}/source/analyzers-common")
+
+def luceneGeneratedSourceDirectory = layout.buildDirectory.dir('generated/sources/lucene')
+def luceneGeneratedPorterFile = luceneGeneratedSourceDirectory.map { it.file('org/egothor/stemmer/benchmark/LucenePorterStemmerCopied.java') }
+
+def luceneSourceDownloadFile = layout.buildDirectory.file("${luceneRootRelativePath}/lucene-${luceneVersion}-sources.jar")
+
+dependencies {
+    jmhImplementation "org.apache.lucene:lucene-analysis-common:${luceneVersion}"
+    jmhImplementation "org.apache.lucene:lucene-analysis-stempel:${luceneVersion}"
+    jmhImplementation "org.apache.lucene:lucene-analysis-morfologik:${luceneVersion}"
+}
+
+def buildLuceneSourcesName = { final String artifact ->
+    "${artifact}-${luceneVersion}-sources.jar"
+}
+def buildLuceneSourcesUrl = { final String artifact ->
+    "https://repo1.maven.org/maven2/org/apache/lucene/${artifact}/${luceneVersion}/${buildLuceneSourcesName(artifact)}"
+}
+
+def isLuceneSourcesDownloadable = { final String artifact ->
+    try {
+        final URL sourceUrl = new URL(buildLuceneSourcesUrl(artifact))
+        final java.net.HttpURLConnection connection = (java.net.HttpURLConnection) sourceUrl.openConnection()
+        connection.requestMethod = 'HEAD'
+        connection.instanceFollowRedirects = true
+        connection.connectTimeout = 10000
+        connection.readTimeout = 10000
+        final int responseCode = connection.responseCode
+        connection.disconnect()
+        return responseCode == 200
+    } catch (Exception ignored) {
+        return false
+    }
+}
+
+def downloadLuceneSourcesJar = { ->
+    final File targetFile = luceneSourceDownloadFile.get().asFile
+
+    for (String artifact : luceneSourceArtifacts) {
+        if (!isLuceneSourcesDownloadable(artifact)) {
+            continue
+        }
+
+        final String sourceUrl = buildLuceneSourcesUrl(artifact)
+        final File tempFile = new File(targetFile.parentFile, "${artifact}.${luceneVersion}.tmp")
+        try {
+            new URL(sourceUrl).withInputStream { inputStream ->
+                tempFile.parentFile.mkdirs()
+                tempFile.withOutputStream { outputStream ->
+                    outputStream << inputStream
+                }
+            }
+            if (!tempFile.exists() || tempFile.length() <= 0L) {
+                throw new GradleException("Downloaded Lucene source artifact for ${artifact} was empty.")
+            }
+            targetFile.delete()
+            if (!tempFile.renameTo(targetFile)) {
+                throw new GradleException("Failed to persist downloaded Lucene source artifact for ${artifact}.")
+            }
+            return
+        } catch (Exception ignored) {
+            tempFile.delete()
+        }
+    }
+
+    throw new GradleException(
+            "Failed to download Apache Lucene source artifacts ${luceneSourceArtifacts} for version ${luceneVersion}.")
+}
+
+def luceneSourceClasspathPath = provider {
+    project.relativePath(luceneGeneratedSourceDirectory.get().asFile)
+}
+def luceneEclipseClasspathAttributes = [
+        gradle_scope        : 'jmh',
+        gradle_used_by_scope: 'jmh',
+        test                : 'true'
+]
+
+def isAbsoluteClasspathPath = { String path ->
+    path.startsWith('/') || path ==~ /^[A-Za-z]:[\\\/].*/
+}
+
+def luceneGeneratedPorterNotice = '''
+/**
+ * Generated at benchmark execution time from Apache Lucene source.
+ *
+ * This source copy is compiled only for the JMH benchmark source set and is
+ * not committed as production code.
+ */
+'''
+
+def transformPorterStemmerSource = { final File sourceFile, final File targetFile ->
+    if (!sourceFile.exists()) {
+        throw new GradleException("Apache Lucene PorterStemmer source was not available at ${sourceFile}.")
+    }
+
+    final String sourceText = sourceFile.getText('UTF-8')
+    String transformedText = sourceText
+
+    if (transformedText.contains('package org.apache.lucene.analysis.en;')) {
+        transformedText = transformedText.replaceFirst(/(?m)^\s*package\s+org\.apache\.lucene\.analysis\.en\s*;/,
+                'package org.egothor.stemmer.benchmark;')
+    } else {
+        throw new GradleException(
+                'Expected Lucene package-private PorterStemmer in org.apache.lucene.analysis.en package was not found in downloaded source.')
+    }
+
+    transformedText = transformedText.replaceFirst(/(?m)^\s*class\s+PorterStemmer\s*\{/, 'public final class LucenePorterStemmerCopied {')
+    transformedText = transformedText.replaceFirst(/(?m)^\s*public\s+PorterStemmer\(\)/, 'public LucenePorterStemmerCopied()')
+
+    if (!transformedText.contains('class LucenePorterStemmerCopied')) {
+        throw new GradleException("Failed to rename PorterStemmer class when generating ${targetFile}.")
+    }
+
+    targetFile.parentFile.mkdirs()
+    targetFile.text = transformedText
+}
+
+def resolveLucenePorterStemmerSource = { ->
+    final File sourceRoot = luceneSourceDirectory.get().asFile
+    final List<String> candidates = [
+            'org/apache/lucene/analysis/en/org/apache/lucene/analysis/en/PorterStemmer.java',
+            'org/apache/lucene/analysis/en/PorterStemmer.java',
+            'org/apache/lucene/analysis/en/org/tartarus/snowball/ext/PorterStemmer.java'
+    ]
+
+    for (String candidate : candidates) {
+        final File file = new File(sourceRoot, candidate)
+        if (file.exists()) {
+            return file
+        }
+    }
+
+    final FileTree porterCandidates = fileTree(sourceRoot).matching { include '**/PorterStemmer.java' }
+    for (File file : porterCandidates.files) {
+        if (file.text.contains('class PorterStemmer') && file.text.contains('package org.apache.lucene.analysis.en;')) {
+            return file
+        }
+    }
+
+    throw new GradleException('Unable to resolve Lucene PorterStemmer source file from extracted artifact.')
+}
+
+tasks.register('downloadLuceneAnalyzersSources') {
+    group = 'build setup'
+    description = 'Downloads Apache Lucene analysis sources for benchmark-only code generation.'
+
+    outputs.file(luceneSourceDownloadFile)
+
+    doLast {
+        if (!luceneSourceDownloadFile.get().asFile.exists()) {
+            downloadLuceneSourcesJar()
+        }
+    }
+}
+
+tasks.register('extractLuceneAnalyzersSources', Copy) {
+    group = 'build setup'
+    description = 'Extracts Apache Lucene analysis source JAR for benchmark-only extraction.'
+
+    dependsOn(tasks.named('downloadLuceneAnalyzersSources'))
+
+    from(zipTree(luceneSourceDownloadFile))
+    into(luceneSourceDirectory)
+}
+
+tasks.register('generateLucenePorterStemmerCopied') {
+    group = 'build setup'
+    description = 'Generates LucenePorterStemmerCopied into the build-only benchmark source directory.'
+
+    dependsOn(tasks.named('extractLuceneAnalyzersSources'))
+
+    inputs.dir(luceneSourceDirectory)
+    outputs.file(luceneGeneratedPorterFile)
+
+    doLast {
+        final File sourceFile = resolveLucenePorterStemmerSource()
+        transformPorterStemmerSource(sourceFile, luceneGeneratedPorterFile.get().asFile)
+    }
+}
+
+sourceSets {
+    jmh {
+        java {
+            srcDir(luceneGeneratedSourceDirectory)
+        }
+    }
+}
+
+tasks.named('compileJmhJava') {
+    dependsOn(tasks.named('generateLucenePorterStemmerCopied'))
+}
+
+eclipse {
+    classpath {
+        file {
+            whenMerged { classpath ->
+                String generatedPath = luceneSourceClasspathPath.get()
+
+                classpath.entries.removeAll { entry ->
+                    entry.hasProperty('path') && (
+                            entry.path == generatedPath ||
+                            isAbsoluteClasspathPath(entry.path)
+                    )
+                }
+
+                SourceFolder luceneEntry = new SourceFolder(generatedPath, null)
+                luceneEntry.output = 'bin/jmh'
+                luceneEclipseClasspathAttributes.each { String name, String value ->
+                    luceneEntry.entryAttributes[name] = value
+                }
+                classpath.entries.add(luceneEntry)
+            }
+        }
+    }
+}
--- a/gradle/opennlp-benchmarks.gradle
+++ b/gradle/opennlp-benchmarks.gradle
@@ -0,0 +1,5 @@
+def openNlpVersion = '2.5.4'
+
+dependencies {
+    jmhImplementation "org.apache.opennlp:opennlp-tools:${openNlpVersion}"
+}
--- a/gradle/paicehusk-benchmarks.gradle
+++ b/gradle/paicehusk-benchmarks.gradle
@@ -0,0 +1,219 @@
+import org.gradle.plugins.ide.eclipse.model.SourceFolder
+
+
+def paicehuskVersion = 'master'
+def paicehuskArchiveName = "paice-husk-stemmer-${paicehuskVersion}.zip"
+def paicehuskDownloadUrl = "https://github.com/Hopper262/paice-husk-stemmer/archive/refs/heads/${paicehuskVersion}.zip"
+def paicehuskDownloadFile = layout.buildDirectory.file("third-party/paicehusk/${paicehuskArchiveName}")
+def paicehuskExtractDirectory = layout.buildDirectory.dir('third-party/paicehusk/source')
+def paicehuskArchiveDirectory = paicehuskExtractDirectory.map { it.dir('paice-husk-stemmer-master') }
+def paicehuskJavaFile = paicehuskArchiveDirectory.map { it.file('paicehusk_java.java') }
+def paicehuskRulesFile = paicehuskArchiveDirectory.map { it.file('paicehusk_rules.txt') }
+
+def paicehuskGeneratedSourceDirectory = layout.buildDirectory.dir('generated/sources/paicehusk')
+def paicehuskGeneratedStemmerFile = paicehuskGeneratedSourceDirectory.map { it.file('org/egothor/stemmer/benchmark/PaiceHuskLancasterStemmer.java') }
+
+def paicehuskGeneratedSourcePath = provider {
+    project.relativePath(paicehuskGeneratedSourceDirectory.get().asFile)
+}
+def paicehuskSourceEclipseClasspathAttributes = [
+        gradle_scope        : 'jmh',
+        gradle_used_by_scope: 'jmh',
+        test                : 'true'
+]
+
+def paicehuskIsAbsolutePath = { String path ->
+    path.startsWith('/') || path ==~ /^[A-Za-z]:[\\\/].*/
+}
+
+def paicehuskGeneratedNotice = '''
+/**
+ * Generated at benchmark execution time from upstream
+ * https://github.com/Hopper262/paice-husk-stemmer .
+ *
+ * This source copy is compiled only for the JMH benchmark source set and is
+ * not committed as production code.
+ */
+'''
+
+def escapeForJava = { final String text ->
+    return text.replace('\\\\', '\\\\\\\\')
+            .replace('\"', '\\\"')
+}
+
+def toRuleLines = { final File rulesFile ->
+    final List<String> lines = rulesFile.readLines('UTF-8')
+    final StringBuilder ruleLines = new StringBuilder()
+    for (int index = 0; index < lines.size(); index++) {
+        final String line = lines.get(index)
+        ruleLines.append('            "')
+        ruleLines.append(escapeForJava(line))
+        ruleLines.append('"')
+        if (index < lines.size() - 1) {
+            ruleLines.append(',')
+        }
+        ruleLines.append('\n')
+    }
+    return ruleLines.toString()
+}
+
+def paicehuskEngineInsertion = { final String ruleLines ->
+    return """
+
+    public static final String[] RULE_LINES = {
+${ruleLines}
+    };
+
+    private static final java.util.HashMap RULES = createRulesFromEmbeddedRules();
+
+    /**
+     * Creates benchmark stemmer instance.
+     */
+    public PaiceHuskLancasterStemmer() {
+    }
+
+    /**
+     * Applies Paice/Husk stemming to one token.
+     *
+     * @param token input token
+     * @return stemmed token
+     */
+    public String stem(final String token) {
+        if (token == null) {
+            return null;
+        }
+        return stemWord(token, RULES, null);
+    }
+
+    /**
+     * Loads bundled rule lines directly from the generated benchmark source.
+     *
+     * @return initialized rule map
+     */
+    private static java.util.HashMap createRulesFromEmbeddedRules() {
+        try {
+            final java.io.File ruleFile = java.io.File.createTempFile("paicehusk-rules", ".txt");
+            ruleFile.deleteOnExit();
+            try (java.io.PrintWriter writer = new java.io.PrintWriter(new java.io.FileWriter(ruleFile))) {
+                for (String line : RULE_LINES) {
+                    writer.println(line);
+                }
+            }
+            return loadRules(ruleFile.getAbsolutePath());
+        } catch (Exception exception) {
+            throw new IllegalStateException("Unable to initialize benchmark Paice/Husk rules.", exception);
+        }
+    }
+"""
+}
+
+def transformPaiceHuskSource = { final File sourceFile, final File rulesFile, final File targetFile ->
+    if (!sourceFile.exists()) {
+        throw new GradleException("Paice/Husk Java source was not available at ${sourceFile}.")
+    }
+    if (!rulesFile.exists()) {
+        throw new GradleException("Paice/Husk rule file was not available at ${rulesFile}.")
+    }
+
+    final String sourceText = sourceFile.getText('UTF-8')
+    String transformedText = sourceText
+
+    transformedText = 'package org.egothor.stemmer.benchmark;' + '\n\n' + transformedText
+    transformedText = transformedText.replaceFirst(/(?m)^\s*class\s+PaiceHusk\s*\{/, 'public final class PaiceHuskLancasterStemmer {')
+
+    final int packageEnd = transformedText.indexOf('\n', transformedText.indexOf('package org.egothor.stemmer.benchmark;'))
+    if (packageEnd >= 0) {
+        transformedText = transformedText.substring(0, packageEnd + 1) + '\n' + paicehuskGeneratedNotice + transformedText.substring(packageEnd + 1)
+    }
+
+    final String marker = '\n} // end class PaiceHusk'
+    final int markerIndex = transformedText.lastIndexOf(marker)
+    if (markerIndex < 0) {
+        throw new GradleException("Unexpected Paice/Husk source structure at ${sourceFile}.")
+    }
+    final String replacement = paicehuskEngineInsertion(toRuleLines(rulesFile))
+    transformedText = transformedText.substring(0, markerIndex) + '\n' + replacement + '\n}' 
+
+    targetFile.parentFile.mkdirs()
+    targetFile.text = transformedText
+}
+
+tasks.register('downloadPaiceHuskStemmer') {
+    group = 'build setup'
+    description = 'Downloads the upstream Paice/Husk benchmark source for dynamic extraction.'
+
+    outputs.file(paicehuskDownloadFile)
+
+    doLast {
+        final File targetFile = paicehuskDownloadFile.get().asFile
+        targetFile.parentFile.mkdirs()
+
+        if (!targetFile.exists()) {
+            new URL(paicehuskDownloadUrl).withInputStream { inputStream ->
+                targetFile.withOutputStream { outputStream ->
+                    outputStream << inputStream
+                }
+            }
+        }
+    }
+}
+
+tasks.register('extractPaiceHuskStemmer', Copy) {
+    group = 'build setup'
+    description = 'Extracts the upstream Paice/Husk benchmark archive.'
+
+    dependsOn(tasks.named('downloadPaiceHuskStemmer'))
+
+    from(zipTree(paicehuskDownloadFile))
+    into(paicehuskExtractDirectory)
+}
+
+tasks.register('generatePaiceHuskLancasterStemmer') {
+    group = 'build setup'
+    description = 'Generates PaiceHuskLancasterStemmer into a benchmark-only generated source directory.'
+
+    dependsOn(tasks.named('extractPaiceHuskStemmer'))
+
+    inputs.files(paicehuskJavaFile, paicehuskRulesFile)
+    outputs.file(paicehuskGeneratedStemmerFile)
+
+    doLast {
+        transformPaiceHuskSource(paicehuskJavaFile.get().asFile, paicehuskRulesFile.get().asFile, paicehuskGeneratedStemmerFile.get().asFile)
+    }
+}
+
+sourceSets {
+    jmh {
+        java {
+            srcDir(paicehuskGeneratedSourceDirectory)
+        }
+    }
+}
+
+tasks.named('compileJmhJava') {
+    dependsOn(tasks.named('generatePaiceHuskLancasterStemmer'))
+}
+
+eclipse {
+    classpath {
+        file {
+            whenMerged { classpath ->
+                String generatedPath = paicehuskGeneratedSourcePath.get()
+
+                classpath.entries.removeAll { entry ->
+                    entry.hasProperty('path') && (
+                            entry.path == generatedPath ||
+                            paicehuskIsAbsolutePath(entry.path)
+                    )
+                }
+
+                SourceFolder paicehuskEntry = new SourceFolder(generatedPath, null)
+                paicehuskEntry.output = 'bin/jmh'
+                paicehuskSourceEclipseClasspathAttributes.each { String name, String value ->
+                    paicehuskEntry.entryAttributes[name] = value
+                }
+                classpath.entries.add(paicehuskEntry)
+            }
+        }
+    }
+}
--- a/gradle/snowball-benchmarks.gradle
+++ b/gradle/snowball-benchmarks.gradle
@@ -7,13 +7,43 @@ def snowballDistributionDirectoryName = "libstemmer_java-${snowballVersion}"
 def snowballRootRelativePath = 'third-party/snowball'
 def snowballSourceRelativePath = "${snowballRootRelativePath}/source"
 def snowballJavaSourceRelativePath = "${snowballSourceRelativePath}/${snowballDistributionDirectoryName}/java"
+def snowballGeneratedSourceRelativePath = 'generated/sources/snowball'
 def snowballDownloadUrl = "https://snowballstem.org/dist/${snowballArchiveName}"
 def snowballDownloadFile = layout.buildDirectory.file("${snowballRootRelativePath}/${snowballArchiveName}")
 def snowballExtractDirectory = layout.buildDirectory.dir(snowballSourceRelativePath)
 def snowballJavaSourceDirectory = layout.buildDirectory.dir(snowballJavaSourceRelativePath)
+def snowballGeneratedSourceDirectory = layout.buildDirectory.dir(snowballGeneratedSourceRelativePath)
 def snowballJavaSourceClasspathPath = provider {
    project.relativePath(snowballJavaSourceDirectory.get().asFile)
 }
+def snowballGeneratedSourceClasspathPath = provider {
+    project.relativePath(snowballGeneratedSourceDirectory.get().asFile)
+}
+def transformSnowballSourceText = { final String sourceText ->
+    String transformedText = sourceText
+
+    transformedText = transformedText.replaceAll(/(?m)^\s*package\s+org\.tartarus\.snowball\.ext\s*;/,
+            'package org.egothor.stemmer.benchmark.snowball.ext;')
+    transformedText = transformedText.replaceAll(/(?m)^\s*package\s+org\.tartarus\.snowball\s*;/,
+            'package org.egothor.stemmer.benchmark.snowball;')
+    transformedText = transformedText.replace('org.tartarus.snowball.', 'org.egothor.stemmer.benchmark.snowball.')
+
+    return transformedText
+}
+def copySnowballSourcesWithPackageIsolation = { final File sourceDirectory, final File targetDirectory ->
+    final FileTree sourceFiles = fileTree(sourceDirectory).matching { include '**/*.java' }
+    if (targetDirectory.exists()) {
+        targetDirectory.deleteDir()
+    }
+
+    for (File sourceFile : sourceFiles.files) {
+        final String relativePath = sourceDirectory.toPath().relativize(sourceFile.toPath()).toString()
+        final File outputFile = new File(targetDirectory, relativePath)
+        outputFile.parentFile.mkdirs()
+        outputFile.text = transformSnowballSourceText(sourceFile.getText('UTF-8'))
+    }
+}
+
 def snowballEclipseClasspathAttributes = [
        gradle_scope        : 'jmh',
        gradle_used_by_scope: 'jmh',
@@ -53,16 +83,33 @@ tasks.register('extractSnowballJava', Copy) {
    into(snowballExtractDirectory)
 }

+tasks.register('generateIsolatedSnowballSources') {
+    group = 'build setup'
+    description = 'Copies Snowball source to benchmark-only package-isolated package paths.'
+
+    dependsOn(tasks.named('extractSnowballJava'))
+
+    inputs.dir(snowballJavaSourceDirectory)
+    outputs.dir(snowballGeneratedSourceDirectory)
+
+    doLast {
+        copySnowballSourcesWithPackageIsolation(
+                snowballJavaSourceDirectory.get().asFile,
+                snowballGeneratedSourceDirectory.get().asFile
+        )
+    }
+}
+
 sourceSets {
    jmh {
        java {
-            srcDir(snowballJavaSourceDirectory)
+            srcDir(snowballGeneratedSourceDirectory)
        }
    }
 }

 tasks.named('compileJmhJava') {
-    dependsOn(tasks.named('extractSnowballJava'))
+    dependsOn(tasks.named('generateIsolatedSnowballSources'))
 }

 eclipse {
@@ -70,17 +117,19 @@ eclipse {
        file {
            whenMerged { classpath ->
                String generatedSnowballPath = snowballJavaSourceClasspathPath.get()
+                String generatedIsolatedSnowballPath = snowballGeneratedSourceClasspathPath.get()
                String modelSnowballPath = snowballJavaSourceRelativePath

                classpath.entries.removeAll { entry ->
                    entry.hasProperty('path') && (
                            entry.path == generatedSnowballPath ||
+                            entry.path == generatedIsolatedSnowballPath ||
                            entry.path == modelSnowballPath ||
                            isAbsoluteClasspathPath(entry.path)
                    )
                }

-                SourceFolder snowballEntry = new SourceFolder(generatedSnowballPath, null)
+                SourceFolder snowballEntry = new SourceFolder(generatedIsolatedSnowballPath, null)
                snowballEntry.output = 'bin/jmh'
                snowballEclipseClasspathAttributes.each { String name, String value ->
                    snowballEntry.entryAttributes[name] = value
--- a/gradle/verification-metadata.xml
+++ b/gradle/verification-metadata.xml
@@ -710,6 +710,11 @@
            <sha256 value="d78bd8524c5f8380a190a6525686629a95dfe512df21111383a6d8c0923a4415" origin="Generated by Gradle"/>
         </artifact>
      </component>
+      <component group="org.apache" name="apache" version="34">
+         <artifact name="apache-34.pom">
+            <sha256 value="3671ae9d4d062ae3bb985731c76088bb2f6f7d7254e2d304ee9f690b97651328" origin="Generated by Gradle"/>
+         </artifact>
+      </component>
      <component group="org.apache" name="apache" version="35">
         <artifact name="apache-35.pom">
            <sha256 value="ea297dcd114136e8b8e8b630230d52a76c2fc69f6c5db25d672b1857000728b8" origin="Generated by Gradle"/>
@@ -905,6 +910,14 @@
            <sha256 value="d8ef04000565affac019b7a55de5bb7cc82ab0403295285ef49f6c8c2745afeb" origin="Generated by Gradle"/>
         </artifact>
      </component>
+      <component group="org.apache.lucene" name="lucene-analysis-common" version="10.5.0">
+         <artifact name="lucene-analysis-common-10.5.0.jar">
+            <sha256 value="922e217fe5cc88305b5a8e057cd80a30b7de96be2ff20dc4327fb524d26e6a25" origin="Generated by Gradle"/>
+         </artifact>
+         <artifact name="lucene-analysis-common-10.5.0.pom">
+            <sha256 value="8ad3288be355a6dca42678ae6abaa4726fe4a320b1657e7a7a455ae684edad27" origin="Generated by Gradle"/>
+         </artifact>
+      </component>
      <component group="org.apache.lucene" name="lucene-analysis-common" version="9.12.3">
         <artifact name="lucene-analysis-common-9.12.3.jar">
            <sha256 value="fa571bd7caf0f0b4faf46a72ca004a7836f348c31d92bd522dddcc3d128d287e" origin="Generated by Gradle"/>
@@ -913,6 +926,30 @@
            <sha256 value="3242d6696252c6ce33744087bbb821c384373c7819c3a054539b59876e3df39a" origin="Generated by Gradle"/>
         </artifact>
      </component>
+      <component group="org.apache.lucene" name="lucene-analysis-morfologik" version="10.5.0">
+         <artifact name="lucene-analysis-morfologik-10.5.0.jar">
+            <sha256 value="7438fa8afd11dc9b606e911ef5a922cf642a3d45e32c5bbd456e54d731810dbe" origin="Generated by Gradle"/>
+         </artifact>
+         <artifact name="lucene-analysis-morfologik-10.5.0.pom">
+            <sha256 value="df9863e0db7416ba1d0d1fa657b2728e31a2ac8150bdd87fa272b100aa0838fb" origin="Generated by Gradle"/>
+         </artifact>
+      </component>
+      <component group="org.apache.lucene" name="lucene-analysis-stempel" version="10.5.0">
+         <artifact name="lucene-analysis-stempel-10.5.0.jar">
+            <sha256 value="f46699a4457e1cec1035737c37be1f061eaf32d96b6679968bbda19d1b12ed2b" origin="Generated by Gradle"/>
+         </artifact>
+         <artifact name="lucene-analysis-stempel-10.5.0.pom">
+            <sha256 value="da627cd84d6e29eeae63fd1881420b1b9f1b3619c106016a5ba0338351a292ac" origin="Generated by Gradle"/>
+         </artifact>
+      </component>
+      <component group="org.apache.lucene" name="lucene-core" version="10.5.0">
+         <artifact name="lucene-core-10.5.0.jar">
+            <sha256 value="ec05ee432860dc6116765fc6bd9ffccf65311cb0430a42a3a90f09fffcd310b1" origin="Generated by Gradle"/>
+         </artifact>
+         <artifact name="lucene-core-10.5.0.pom">
+            <sha256 value="356efef0e44ed7e1979af9d39c4403fea268edc5ede1c6fe4c006e58f218a0dc" origin="Generated by Gradle"/>
+         </artifact>
+      </component>
      <component group="org.apache.lucene" name="lucene-core" version="9.12.3">
         <artifact name="lucene-core-9.12.3.jar">
            <sha256 value="b64a3f8098a7572034fb30085cdee01b34ec81fb0e5a31b471536af58dc6c01b" origin="Generated by Gradle"/>
@@ -1106,6 +1143,19 @@
            <sha256 value="a941745d7faeb8dc9a75edc2c330c994b7440b9a44d21142716b6053967a41c1" origin="Generated by Gradle"/>
         </artifact>
      </component>
+      <component group="org.apache.opennlp" name="opennlp" version="2.5.4">
+         <artifact name="opennlp-2.5.4.pom">
+            <sha256 value="433d0873ec27cfe1b1a20ea643773a2f29a40a1fdea1bb5d1828b2f35350c29b" origin="Generated by Gradle"/>
+         </artifact>
+      </component>
+      <component group="org.apache.opennlp" name="opennlp-tools" version="2.5.4">
+         <artifact name="opennlp-tools-2.5.4.jar">
+            <sha256 value="5efedb26d0e97e53707a8d2f0e2e28f05fcf654fce2a33f8bfe11c2d4c2a4fe9" origin="Generated by Gradle"/>
+         </artifact>
+         <artifact name="opennlp-tools-2.5.4.pom">
+            <sha256 value="fd4557916e65775956c37e8cdda7d9ad2bca778a29c98d6a5c6e31b61f83d560" origin="Generated by Gradle"/>
+         </artifact>
+      </component>
      <component group="org.apache.velocity" name="velocity-engine-core" version="2.4.1">
         <artifact name="velocity-engine-core-2.4.1.jar">
            <sha256 value="1c19157d1171d560088e485be97c93a7a2f7e9f56e517f0a30273c5c39df6231" origin="Generated by Gradle"/>
@@ -1137,6 +1187,35 @@
            <sha256 value="22b87dda9aab83fa1d0f3ea409b524e7a44921cf8f5f87999cf59046f0fe0bc3" origin="Generated by Gradle"/>
         </artifact>
      </component>
+      <component group="org.carrot2" name="morfologik-fsa" version="2.1.9">
+         <artifact name="morfologik-fsa-2.1.9.jar">
+            <sha256 value="1bfefce937df14cc94d32a98ce59c33f4d5b6c0eddbb436b6bfe27ff2120a23d" origin="Generated by Gradle"/>
+         </artifact>
+         <artifact name="morfologik-fsa-2.1.9.pom">
+            <sha256 value="1097b12e6ede04b5a4e09b77233ac0943d8a6020edced7fd65ec97f2b02e103c" origin="Generated by Gradle"/>
+         </artifact>
+      </component>
+      <component group="org.carrot2" name="morfologik-parent" version="2.1.9">
+         <artifact name="morfologik-parent-2.1.9.pom">
+            <sha256 value="59c72168787ba151785125e34472e7841c5ff18bde176c2db7349c06917a0627" origin="Generated by Gradle"/>
+         </artifact>
+      </component>
+      <component group="org.carrot2" name="morfologik-polish" version="2.1.9">
+         <artifact name="morfologik-polish-2.1.9.jar">
+            <sha256 value="e503682b3f4e8bb7a5d05820b0e2a4a19d4bad43dae20f64741786658a9cf478" origin="Generated by Gradle"/>
+         </artifact>
+         <artifact name="morfologik-polish-2.1.9.pom">
+            <sha256 value="85595c01c592576b91f59600c040bf2ceabaf61afc25ffa8507e4da981a54fb5" origin="Generated by Gradle"/>
+         </artifact>
+      </component>
+      <component group="org.carrot2" name="morfologik-stemming" version="2.1.9">
+         <artifact name="morfologik-stemming-2.1.9.jar">
+            <sha256 value="6170895b2315b697f4da5630caf57c6c441f1cb419d89d1cb5326b0673293e8a" origin="Generated by Gradle"/>
+         </artifact>
+         <artifact name="morfologik-stemming-2.1.9.pom">
+            <sha256 value="0b1495ad4d54b8dd4d309e1b445625ff2c788cc68818b5ded5cfe4f0d2f891a2" origin="Generated by Gradle"/>
+         </artifact>
+      </component>
      <component group="org.checkerframework" name="checker-qual" version="3.52.1">
         <artifact name="checker-qual-3.52.1.jar">
            <sha256 value="934641a18c8461bf66d7e939b2b054bf2a518ed4188fd7d6836a65b038f5364a" origin="Generated by Gradle"/>
@@ -1955,6 +2034,14 @@
            <sha256 value="ba01ae7a744cb52fe8ecf3b023cbc32e0ccc8c6beef9f26de77a47808239447d" origin="Generated by Gradle"/>
         </artifact>
      </component>
+      <component group="ua.net.nlp" name="morfologik-ukrainian-search" version="4.9.1">
+         <artifact name="morfologik-ukrainian-search-4.9.1.jar">
+            <sha256 value="463d9054b8d4cfacb9cd69566395826a6fe32fb6e7da91dbf79d37ddf7d56ba0" origin="Generated by Gradle"/>
+         </artifact>
+         <artifact name="morfologik-ukrainian-search-4.9.1.pom">
+            <sha256 value="6a66d8efe6a5c774932401b04411a8feca7d6628631e4ef80819cfafa133dc38" origin="Generated by Gradle"/>
+         </artifact>
+      </component>
      <component group="us.springett" name="cpe-parser" version="3.0.1">
         <artifact name="cpe-parser-3.0.1.jar">
            <sha256 value="f98a50dce0a381e08f0f0ac067801c7f4c51dc7f8f1fe7a3c9960c684a809705" origin="Generated by Gradle"/>
--- a/mkdocs.yml
+++ b/mkdocs.yml
@@ -38,17 +38,23 @@ markdown_extensions:
 nav:
  - Home: index.md

-  - Getting Started:
+  - Start:
+     - Fast Track: fast-track.md
     - Quick Start: quick-start.md
-     - Built-in Languages: built-in-languages.md
-     - Dictionary Format: dictionary-format.md
-     - CLI Compilation: cli-compilation.md
+     - Integration Deep Dive: integration-deep-dive.md

-  - Programmatic Usage:
+  - Integration:
     - Overview: programmatic-usage.md
     - Loading and Building Stemmers: programmatic-loading-and-building.md
     - Querying and Ambiguity Handling: programmatic-querying-and-ambiguity.md
     - Extending and Persisting Compiled Tries: programmatic-extending-and-persistence.md
+     - Migration and Backward Compatibility: migration-and-backward-compatibility.md
+     - CLI Compilation: cli-compilation.md
+
+  - Dictionaries and Languages:
+     - Built-in Languages: built-in-languages.md
+     - Dictionary Format: dictionary-format.md
+     - Contributing Dictionaries: contributing-dictionaries.md

  - Architecture and Semantics:
     - Overview: architecture-and-reduction.md
@@ -57,11 +63,42 @@ nav:
     - Lookup Edge Optimization: lookup-edge-optimization.md
     - Compatibility and Guarantees: compatibility-and-guarantees.md

-  - Dictionaries:
-     - Contributing Dictionaries: contributing-dictionaries.md
+  - Benchmarks:
+     - How to Read Benchmarks: benchmarking.md
+     - Benchmark Results: benchmarks/index.md
+     - Reference:
+        - Methodology: benchmarks/reference/methodology.md
+        - Linguistic Quality Methodology: benchmarks/reference/linguistic-quality.md
+        - Tested Stemmers: benchmarks/reference/tested-stemmers.md
+        - Reproducibility and Raw Data: benchmarks/reference/reproducibility.md
+        - Corpora: benchmarks/reference/corpora.md
+        - Environment and Reports: benchmarks/reference/environment.md
+        - English Dictionary Coverage: benchmarks/reference/english-coverage.md
+        - Candidate Evaluation: benchmarks/reference/candidates.md
+     - Language Results:
+        - Overview: benchmarks/languages/index.md
+        - Czech: benchmarks/languages/czech.md
+        - Danish: benchmarks/languages/danish.md
+        - Dutch: benchmarks/languages/dutch.md
+        - English: benchmarks/languages/english.md
+        - Finnish: benchmarks/languages/finnish.md
+        - French: benchmarks/languages/french.md
+        - German: benchmarks/languages/german.md
+        - Hungarian: benchmarks/languages/hungarian.md
+        - Italian: benchmarks/languages/italian.md
+        - Norwegian Bokmal: benchmarks/languages/norwegian-bokmal.md
+        - Norwegian Nynorsk: benchmarks/languages/norwegian-nynorsk.md
+        - Persian: benchmarks/languages/persian.md
+        - Polish: benchmarks/languages/polish.md
+        - Portuguese: benchmarks/languages/portuguese.md
+        - Russian: benchmarks/languages/russian.md
+        - Spanish: benchmarks/languages/spanish.md
+        - Swedish: benchmarks/languages/swedish.md
+        - Ukrainian: benchmarks/languages/ukrainian.md
+        - Yiddish: benchmarks/languages/yiddish.md

  - Quality and Operations:
     - Quality and Operations: quality-and-operations.md
-     - Benchmarking: benchmarking.md
+     - Stemming Quality: stemming-quality.md
     - Reports: reports.md
     - Test taxonomy and execution filtering: test-taxonomy-and-filtering.md
--- a/src/jmh/java/org/egothor/stemmer/benchmark/BenchmarkCorpusSupport.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/BenchmarkCorpusSupport.java
@@ -113,12 +113,12 @@ final class BenchmarkCorpusSupport {
            dictionaryBuilder.append(stem);
            lookupKeys.add(stem);
            for (String variant : variants) {
-                dictionaryBuilder.append(' ').append(variant);
+                dictionaryBuilder.append('\t').append(variant);
                lookupKeys.add(variant);
            }

            final String homograph = createHomograph(index);
-            dictionaryBuilder.append(' ').append(homograph);
+            dictionaryBuilder.append('\t').append(homograph);
            lookupKeys.add(homograph);
            ambiguousLookupKeys.add(homograph);

--- a/src/jmh/java/org/egothor/stemmer/benchmark/BenchmarkTokenSequence.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/BenchmarkTokenSequence.java
@@ -0,0 +1,185 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ * 
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ * 
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ * 
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ * 
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ * 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.util.Objects;
+
+/**
+ * Reusable deterministic token sequence for benchmark-only token streams.
+ *
+ * <p>
+ * The sequence keeps stable token ordering and offset progression while avoiding
+ * per-token object creation during iteration.
+ * </p>
+ */
+final class BenchmarkTokenSequence {
+
+    /**
+     * Shared backing corpus as character arrays.
+     */
+    private char[][] tokenCharacters;
+
+    /**
+     * Number of active tokens in the sequence.
+     */
+    private int tokenCount;
+
+    /**
+     * Cursor index for the currently emitted token.
+     */
+    private int cursor;
+
+    /**
+     * Current token character array.
+     */
+    private char[] currentToken;
+
+    /**
+     * Start offset of the current token.
+     */
+    private int currentStartOffset;
+
+    /**
+     * End offset of the current token.
+     */
+    private int currentEndOffset;
+
+    /**
+     * Offset of the next token start.
+     */
+    private int nextOffset;
+
+    /**
+     * Creates a reusable token sequence.
+ *
+     * @param tokens token corpus source
+     */
+    BenchmarkTokenSequence(final String[] tokens) {
+        setTokens(tokens);
+    }
+
+    /**
+     * Sets a new token corpus for this sequence.
+     *
+     * <p>
+     * The sequence stores copied character arrays so token reads can be reused
+     * without creating per-token objects during benchmark iteration.
+     * </p>
+     *
+     * @param tokens new token corpus
+     */
+    void setTokens(final String[] tokens) {
+        Objects.requireNonNull(tokens, "tokens");
+        this.tokenCharacters = new char[tokens.length][];
+        for (int index = 0; index < tokens.length; index++) {
+            final String token = Objects.requireNonNull(tokens[index], "tokens[" + index + "]");
+            this.tokenCharacters[index] = token.toCharArray();
+        }
+
+        this.tokenCount = this.tokenCharacters.length;
+        reset();
+    }
+
+    /**
+     * Resets stream position for reuse.
+     */
+    void reset() {
+        this.cursor = 0;
+        this.nextOffset = 0;
+        this.currentStartOffset = 0;
+        this.currentEndOffset = 0;
+        this.currentToken = null;
+    }
+
+    /**
+     * Returns whether at least one token remains in the sequence.
+     *
+     * @return true if a token can be emitted
+     */
+    boolean hasNext() {
+        return this.cursor < this.tokenCount;
+    }
+
+    /**
+     * Advances to the next token.
+     *
+     * @return true if a token was emitted
+     */
+    boolean advance() {
+        if (!hasNext()) {
+            return false;
+        }
+
+        final char[] token = this.tokenCharacters[this.cursor];
+        this.currentToken = token;
+        this.currentStartOffset = this.nextOffset;
+        this.currentEndOffset = this.currentStartOffset + token.length;
+        this.nextOffset = this.currentEndOffset + 1;
+        this.cursor++;
+        return true;
+    }
+
+    /**
+     * Returns the current token in the sequence.
+     *
+     * @return current token character array
+     */
+    char[] currentToken() {
+        return this.currentToken;
+    }
+
+    /**
+     * Returns current token start offset for token stream attributes.
+     *
+     * @return start offset
+     */
+    int currentStartOffset() {
+        return this.currentStartOffset;
+    }
+
+    /**
+     * Returns current token end offset for token stream attributes.
+ *
+     * @return end offset
+     */
+    int currentEndOffset() {
+        return this.currentEndOffset;
+    }
+
+    /**
+     * Returns final stream offset value used by {@code end()}.
+     *
+     * @return final offset
+     */
+    int endOffset() {
+        return this.nextOffset > 0 ? this.nextOffset - 1 : 0;
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/BenchmarkTokenStream.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/BenchmarkTokenStream.java
@@ -0,0 +1,146 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.io.IOException;
+
+import org.apache.lucene.analysis.TokenStream;
+import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;
+import org.apache.lucene.analysis.tokenattributes.OffsetAttribute;
+import org.apache.lucene.analysis.tokenattributes.PositionIncrementAttribute;
+
+/**
+ * Reusable Lucene {@link TokenStream} backed by a deterministic token array.
+ *
+ * <p>
+ * Instances are mutable and intended for one JMH worker thread. The stream
+ * copies configured token text into reusable character storage so benchmark
+ * iteration can replay the same token sequence without mutating the shared
+ * source array.
+ * </p>
+ */
+final class BenchmarkTokenStream extends TokenStream {
+
+    /**
+     * Current token text attribute.
+     */
+    private final CharTermAttribute charTermAttribute;
+
+    /**
+     * Offset attribute used by Lucene filters that inspect offsets.
+     */
+    private final OffsetAttribute offsetAttribute;
+
+    /**
+     * Position increment attribute for one-token-at-a-time streams.
+     */
+    private final PositionIncrementAttribute positionIncrementAttribute;
+
+    /**
+     * Reusable token sequence.
+     */
+    private final BenchmarkTokenSequence tokenSequence;
+
+    /**
+     * Creates a stream over the supplied tokens.
+     *
+     * @param tokens initial token corpus
+     */
+    BenchmarkTokenStream(final String[] tokens) {
+        this.tokenSequence = new BenchmarkTokenSequence(tokens);
+        this.charTermAttribute = addAttribute(CharTermAttribute.class);
+        this.offsetAttribute = addAttribute(OffsetAttribute.class);
+        this.positionIncrementAttribute = addAttribute(PositionIncrementAttribute.class);
+    }
+
+    /**
+     * Replaces the configured token corpus.
+     *
+     * @param tokens new token corpus
+     */
+    void setTokens(final String[] tokens) {
+        this.tokenSequence.setTokens(tokens);
+    }
+
+    /**
+     * Returns whether all configured tokens have been emitted.
+     *
+     * @return {@code true} after the current pass is exhausted
+     */
+    boolean isDrained() {
+        return !this.tokenSequence.hasNext();
+    }
+
+    /**
+     * {@inheritDoc}
+     */
+    @Override
+    public boolean incrementToken() throws IOException {
+        if (!this.tokenSequence.advance()) {
+            return false;
+        }
+
+        clearAttributes();
+        final char[] token = this.tokenSequence.currentToken();
+        this.charTermAttribute.copyBuffer(token, 0, token.length);
+        this.positionIncrementAttribute.setPositionIncrement(1);
+        this.offsetAttribute.setOffset(this.tokenSequence.currentStartOffset(), this.tokenSequence.currentEndOffset());
+        return true;
+    }
+
+    /**
+     * {@inheritDoc}
+     */
+    @Override
+    public void reset() throws IOException {
+        super.reset();
+        this.tokenSequence.reset();
+    }
+
+    /**
+     * {@inheritDoc}
+     */
+    @Override
+    public void end() throws IOException {
+        super.end();
+        final int endOffset = this.tokenSequence.endOffset();
+        this.offsetAttribute.setOffset(endOffset, endOffset);
+    }
+
+    /**
+     * {@inheritDoc}
+     */
+    @Override
+    public void close() throws IOException {
+        super.close();
+        this.charTermAttribute.setEmpty();
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/Cistem.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/Cistem.java
@@ -0,0 +1,178 @@
+/*******************************************************************************
+ * MIT License
+ *
+ * Copyright (c) 2017 Leonie Weißweiler
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * Source: CISTEM German stemmer
+ *   Authors: Leonie Weissweiler, Alexander Fraser
+ *   https://github.com/LeonieWeissweiler/CISTEM
+ *   https://www.cis.lmu.de/~weissweiler/cistem/
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.util.regex.Pattern;
+
+public final class Cistem {
+
+    private static final Pattern GE_PATTERN = Pattern.compile("^ge(.{4,})");
+    private static final Pattern DOLLAR1_PATTERN = Pattern.compile("(.)\\1");
+    private static final Pattern ND_PATTERN = Pattern.compile("nd$");
+    private static final Pattern EMR_PATTERN = Pattern.compile("e[mr]$");
+    private static final Pattern T_PATTERN = Pattern.compile("t$");
+    private static final Pattern ESN_PATTERN = Pattern.compile("[esn]$");
+    private static final Pattern STAR_PATTERN = Pattern.compile("(.)\\*");
+
+    private Cistem() {
+    }
+
+    public static String stem(final String word) {
+        return stem(word, false);
+    }
+
+    public static String stem(final String word, final boolean caseInsensitive) {
+        if (word.isEmpty()) {
+            return word;
+        }
+
+        String normalized = word;
+        normalized = normalized.replace("Ü", "U");
+        normalized = normalized.replace("Ö", "O");
+        normalized = normalized.replace("Ä", "A");
+        normalized = normalized.replace("ü", "u");
+        normalized = normalized.replace("ö", "o");
+        normalized = normalized.replace("ä", "a");
+
+        final boolean uppercase = Character.isUpperCase(normalized.charAt(0));
+
+        normalized = normalized.toLowerCase();
+        normalized = normalized.replace("ß", "ss");
+        normalized = GE_PATTERN.matcher(normalized).replaceAll("$1");
+        normalized = normalized.replace("sch", "$");
+        normalized = normalized.replace("ei", "%");
+        normalized = normalized.replace("ie", "&");
+        normalized = DOLLAR1_PATTERN.matcher(normalized).replaceAll("$1*");
+
+        while (normalized.length() > 3) {
+            if (normalized.length() > 5) {
+                String newWord = EMR_PATTERN.matcher(normalized).replaceAll("");
+                if (!normalized.equals(newWord)) {
+                    normalized = newWord;
+                    continue;
+                }
+
+                newWord = ND_PATTERN.matcher(normalized).replaceAll("");
+                if (!normalized.equals(newWord)) {
+                    normalized = newWord;
+                    continue;
+                }
+            }
+
+            if (!uppercase || caseInsensitive) {
+                final String newWord = T_PATTERN.matcher(normalized).replaceAll("");
+                if (!normalized.equals(newWord)) {
+                    normalized = newWord;
+                    continue;
+                }
+            }
+
+            final String newWord = ESN_PATTERN.matcher(normalized).replaceAll("");
+            if (!normalized.equals(newWord)) {
+                normalized = newWord;
+            } else {
+                break;
+            }
+        }
+
+        normalized = STAR_PATTERN.matcher(normalized).replaceAll("$1$1");
+        normalized = normalized.replace("&", "ie");
+        normalized = normalized.replace("%", "ei");
+        normalized = normalized.replace("$", "sch");
+
+        return normalized;
+    }
+
+    public static String[] segment(final String word) {
+        return segment(word, false);
+    }
+
+    public static String[] segment(final String word, final boolean caseInsensitive) {
+        if (word.isEmpty()) {
+            return new String[] {"", ""};
+        }
+
+        int restLength = 0;
+        final boolean uppercase = Character.isUpperCase(word.charAt(0));
+        String normalized = word.toLowerCase();
+        final String original = new String(normalized);
+
+        normalized = normalized.replace("sch", "$");
+        normalized = normalized.replace("ei", "%");
+        normalized = normalized.replace("ie", "&");
+        normalized = DOLLAR1_PATTERN.matcher(normalized).replaceAll("$1*");
+
+        while (normalized.length() > 3) {
+            if (normalized.length() > 5) {
+                String newWord = normalized.replaceAll("e[mr]$", "");
+                if (!normalized.equals(newWord)) {
+                    restLength += 2;
+                    normalized = newWord;
+                    continue;
+                }
+
+                newWord = normalized.replaceAll("nd$", "");
+                if (!normalized.equals(newWord)) {
+                    restLength += 2;
+                    normalized = newWord;
+                    continue;
+                }
+            }
+
+            if (!uppercase || caseInsensitive) {
+                final String newWord = normalized.replaceAll("t$", "");
+                if (!normalized.equals(newWord)) {
+                    restLength += 1;
+                    normalized = newWord;
+                    continue;
+                }
+            }
+
+            final String newWord = normalized.replaceAll("[esn]$", "");
+            if (!normalized.equals(newWord)) {
+                restLength += 1;
+                normalized = newWord;
+            } else {
+                break;
+            }
+        }
+
+        normalized = normalized.replaceAll("(.)\\*", "$1$1");
+        normalized = normalized.replace("&", "ie");
+        normalized = normalized.replace("%", "ei");
+        normalized = normalized.replace("$", "sch");
+
+        String rest = "";
+        if (restLength != 0) {
+            rest = original.substring(original.length() - restLength);
+        }
+
+        return new String[] {normalized, rest};
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/EnglishComparisonCorpus.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/EnglishComparisonCorpus.java
@@ -30,43 +30,24 @@
 ******************************************************************************/
 package org.egothor.stemmer.benchmark;

-import java.util.ArrayList;
-import java.util.List;
-import java.util.Locale;
+import java.io.IOException;
+
+import org.egothor.stemmer.StemmerPatchTrieLoader;

 /**
 * Builds a deterministic English token corpus for side-by-side stemming
- * benchmarks.
+ * benchmarks from the bundled Radixor English dictionary resource.
 *
 * <p>
- * The generated corpus mixes:
- * </p>
- * <ul>
- * <li>simple inflections</li>
- * <li>common derivational forms</li>
- * <li>US/UK spelling families</li>
- * <li>forms that are suitable for comparison against the bundled
- * {@code US_UK_PROFI} Radixor dictionary</li>
- * </ul>
- *
- * <p>
- * The goal is not to simulate natural language frequency distribution exactly,
- * but to provide a stable and reproducible comparison workload for benchmark
- * runs and regression tracking.
+ * The dictionary resource stores the expected stem as the first tab-separated
+ * field on each line and its surface variants on the same line. This helper
+ * uses only token/root pairs where the token differs from the expected root for
+ * timing. Resources smaller than the shared timing minimum are repeated
+ * deterministically by {@link LanguageBenchmarkCorpus}.
 * </p>
 */
 final class EnglishComparisonCorpus {

-    /**
-     * Canonical lexical bases used to generate the token workload.
-     */
-    private static final String[] BASES = { "analyze", "analyse", "color", "colour", "center", "centre", "organize",
-            "organise", "optimize", "optimise", "characterize", "characterise", "connect", "construct", "compute",
-            "design", "develop", "engineer", "govern", "improve", "index", "inform", "manage", "model", "observe",
-            "operate", "perform", "predict", "prepare", "process", "project", "protect", "publish", "query", "reduce",
-            "refresh", "render", "resolve", "return", "search", "select", "signal", "store", "structure", "support",
-            "transform", "update", "validate", "value" };
-
    /**
     * Utility class.
     */
@@ -77,64 +58,21 @@ final class EnglishComparisonCorpus {
    /**
     * Creates a deterministic token corpus for English stemming comparison.
     *
-     * @param familyCount number of generated lexical families
     * @return token array in stable order
+     * @throws IOException if the bundled English resource cannot be read
     */
-    static String[] createTokens(final int familyCount) {
-        if (familyCount < 1) {
-            throw new IllegalArgumentException("familyCount must be at least 1.");
-        }
-
-        final List<String> tokens = new ArrayList<>(familyCount * 14);
-
-        for (int index = 0; index < familyCount; index++) {
-            final String base = createBase(index);
-
-            tokens.add(base);
-            tokens.add(base + "s");
-            tokens.add(base + "ed");
-            tokens.add(base + "ing");
-            tokens.add(base + "er");
-            tokens.add(base + "ers");
-            tokens.add(base + "ly");
-            tokens.add(base + "ness");
-            tokens.add(base + "ment");
-            tokens.add(base + "ments");
-            tokens.add(base + "able");
-            tokens.add(base + "ability");
-
-            if (base.endsWith("ize")) {
-                tokens.add(base.substring(0, base.length() - 3) + "isation");
-                tokens.add(base.substring(0, base.length() - 3) + "ised");
-            }
-
-            if (base.endsWith("ise")) {
-                tokens.add(base.substring(0, base.length() - 3) + "ization");
-                tokens.add(base.substring(0, base.length() - 3) + "ized");
-            }
-        }
-
-        return tokens.toArray(String[]::new);
+    static String[] createTokens() throws IOException {
+        return createCorpus().tokens();
    }

    /**
-     * Creates one deterministic base token.
+     * Creates a deterministic changed-token corpus and expected root array for
+     * English stemming comparison.
     *
-     * @param index base ordinal
-     * @return generated lexical base
+     * @return token corpus with expected roots
+     * @throws IOException if the bundled English resource cannot be read
     */
-    private static String createBase(final int index) {
-        return (BASES[index % BASES.length] + suffix(index)).toLowerCase(Locale.ROOT);
-    }
-
-    /**
-     * Creates a compact discriminator suffix so that large corpora remain unique
-     * while retaining stable lexical families.
-     *
-     * @param value ordinal value
-     * @return compact discriminator
-     */
-    private static String suffix(final int value) {
-        return Integer.toString(value, Character.MAX_RADIX);
+    static LanguageBenchmarkCorpus.Corpus createCorpus() throws IOException {
+        return LanguageBenchmarkCorpus.createChangedCorpus(StemmerPatchTrieLoader.Language.US_UK);
    }
 }
--- a/src/jmh/java/org/egothor/stemmer/benchmark/EnglishHunspellStemmerComparisonBenchmarkQuality.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/EnglishHunspellStemmerComparisonBenchmarkQuality.java
@@ -0,0 +1,280 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.io.IOException;
+import java.io.InputStream;
+import java.text.ParseException;
+import java.util.List;
+import java.util.Objects;
+import java.util.concurrent.TimeUnit;
+
+import org.apache.lucene.analysis.LowerCaseFilter;
+import org.apache.lucene.analysis.TokenStream;
+import org.apache.lucene.analysis.hunspell.Dictionary;
+import org.apache.lucene.analysis.hunspell.HunspellStemFilter;
+import org.apache.lucene.analysis.hunspell.SortingStrategy;
+import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;
+import org.apache.lucene.analysis.tokenattributes.PositionIncrementAttribute;
+import org.egothor.stemmer.StemmerPatchTrieLoader;
+import org.openjdk.jmh.annotations.AuxCounters;
+import org.openjdk.jmh.annotations.Benchmark;
+import org.openjdk.jmh.annotations.BenchmarkMode;
+import org.openjdk.jmh.annotations.Fork;
+import org.openjdk.jmh.annotations.Level;
+import org.openjdk.jmh.annotations.Measurement;
+import org.openjdk.jmh.annotations.Mode;
+import org.openjdk.jmh.annotations.OutputTimeUnit;
+import org.openjdk.jmh.annotations.Scope;
+import org.openjdk.jmh.annotations.Setup;
+import org.openjdk.jmh.annotations.State;
+import org.openjdk.jmh.annotations.Warmup;
+import org.openjdk.jmh.infra.Blackhole;
+
+/**
+ * Emits exact-root agreement metrics for the benchmark-only English Hunspell
+ * comparison.
+ */
+@BenchmarkMode(Mode.AverageTime)
+@OutputTimeUnit(TimeUnit.NANOSECONDS)
+@Warmup(iterations = 0)
+@Measurement(iterations = 1, time = 1, timeUnit = TimeUnit.MILLISECONDS)
+@Fork(0)
+public class EnglishHunspellStemmerComparisonBenchmarkQuality {
+
+    /**
+     * Shared English quality corpus and Hunspell dictionary.
+     */
+    @State(Scope.Benchmark)
+    public static class SharedState {
+
+        /**
+         * Complete English resource-derived corpus.
+         */
+        private LanguageBenchmarkCorpus.Corpus corpus;
+
+        /**
+         * Parsed benchmark-only Hunspell dictionary.
+         */
+        private Dictionary dictionary;
+
+        /**
+         * Initializes quality resources.
+         *
+         * @throws IOException    if corpus or dictionary loading fails
+         * @throws ParseException if the Hunspell dictionary cannot be parsed
+         */
+        @Setup(Level.Trial)
+        public void setUp() throws IOException, ParseException {
+            this.corpus = LanguageBenchmarkCorpus.createFullCorpus(StemmerPatchTrieLoader.Language.US_UK);
+            this.dictionary = loadEnglishDictionary();
+        }
+    }
+
+    /**
+     * JMH auxiliary counters for exact-root agreement.
+     */
+    @State(Scope.Thread)
+    @AuxCounters(AuxCounters.Type.EVENTS)
+    public static class AccuracyCounters {
+
+        /**
+         * Number of exact-root matches.
+         */
+        public long correctMatches;
+
+        /**
+         * Number of evaluated tokens.
+         */
+        public long evaluatedTokens;
+
+        /**
+         * Number of exact-root matches where the input token differs from the
+         * expected root.
+         */
+        public long changedCorrectMatches;
+
+        /**
+         * Number of evaluated tokens where the input token differs from the expected
+         * root.
+         */
+        public long changedEvaluatedTokens;
+
+        /**
+         * Number of exact-root matches where the input token is already the expected
+         * root.
+         */
+        public long rootPreservedMatches;
+
+        /**
+         * Number of evaluated tokens where the input token is already the expected
+         * root.
+         */
+        public long rootEvaluatedTokens;
+
+        /**
+         * Resets counters before the measured iteration.
+         */
+        @Setup(Level.Iteration)
+        public void reset() {
+            this.correctMatches = 0L;
+            this.evaluatedTokens = 0L;
+            this.changedCorrectMatches = 0L;
+            this.changedEvaluatedTokens = 0L;
+            this.rootPreservedMatches = 0L;
+            this.rootEvaluatedTokens = 0L;
+        }
+    }
+
+    /**
+     * Evaluates exact-root agreement for English Hunspell.
+     *
+     * @param sharedState shared English quality state
+     * @param counters    JMH auxiliary counters
+     * @param blackhole   result sink
+     * @return exact-root match count
+     * @throws IOException if Lucene token streaming fails
+     */
+    @Benchmark
+    public int luceneHunspellStemFilterAccuracy(final SharedState sharedState, final AccuracyCounters counters,
+            final Blackhole blackhole) throws IOException {
+        final String[] actualStems = firstHunspellOutputs(sharedState.corpus.tokens(), sharedState.dictionary,
+                blackhole);
+        final String[] tokens = sharedState.corpus.tokens();
+        final String[] expectedRoots = sharedState.corpus.expectedRoots();
+
+        int correct = 0;
+        int changedCorrect = 0;
+        int changedEvaluated = 0;
+        int rootPreserved = 0;
+        int rootEvaluated = 0;
+        for (int index = 0; index < actualStems.length; index++) {
+            final String token = tokens[index];
+            final String expectedRoot = expectedRoots[index];
+            final boolean exact = Objects.equals(expectedRoot, actualStems[index]);
+            if (exact) {
+                correct++;
+            }
+            if (Objects.equals(token, expectedRoot)) {
+                rootEvaluated++;
+                if (exact) {
+                    rootPreserved++;
+                }
+            } else {
+                changedEvaluated++;
+                if (exact) {
+                    changedCorrect++;
+                }
+            }
+        }
+
+        counters.correctMatches += correct;
+        counters.evaluatedTokens += actualStems.length;
+        counters.changedCorrectMatches += changedCorrect;
+        counters.changedEvaluatedTokens += changedEvaluated;
+        counters.rootPreservedMatches += rootPreserved;
+        counters.rootEvaluatedTokens += rootEvaluated;
+        return correct;
+    }
+
+    /**
+     * Extracts the first emitted Hunspell stem for each input token.
+     *
+     * @param tokens     token corpus
+     * @param dictionary Hunspell dictionary
+     * @param blackhole  result sink
+     * @return first emitted term per input token
+     * @throws IOException if Lucene streaming fails
+     */
+    private static String[] firstHunspellOutputs(final String[] tokens, final Dictionary dictionary,
+            final Blackhole blackhole) throws IOException {
+        final String[] outputs = new String[tokens.length];
+        final BenchmarkTokenStream input = new BenchmarkTokenStream(tokens);
+        final TokenStream output = new HunspellStemFilter(new LowerCaseFilter(input), dictionary, true);
+        final CharTermAttribute termAttribute = output.addAttribute(CharTermAttribute.class);
+        final PositionIncrementAttribute positionAttribute = output.addAttribute(PositionIncrementAttribute.class);
+        int inputIndex = -1;
+        boolean recordedForPosition = false;
+
+        output.reset();
+        while (output.incrementToken()) {
+            final int positionIncrement = positionAttribute.getPositionIncrement();
+            if (positionIncrement > 0) {
+                inputIndex += positionIncrement;
+                recordedForPosition = false;
+            }
+            if (inputIndex >= 0 && inputIndex < outputs.length && !recordedForPosition) {
+                outputs[inputIndex] = termAttribute.toString();
+                recordedForPosition = true;
+            }
+            blackhole.consume(termAttribute);
+        }
+        output.end();
+        output.close();
+
+        for (int index = 0; index < outputs.length; index++) {
+            if (outputs[index] == null) {
+                outputs[index] = tokens[index];
+            }
+        }
+        return outputs;
+    }
+
+    /**
+     * Loads the benchmark-only English Hunspell dictionary.
+     *
+     * @return parsed Hunspell dictionary
+     * @throws IOException    if dictionary resources cannot be read
+     * @throws ParseException if the Hunspell dictionary cannot be parsed
+     */
+    private static Dictionary loadEnglishDictionary() throws IOException, ParseException {
+        final ClassLoader classLoader = EnglishHunspellStemmerComparisonBenchmarkQuality.class.getClassLoader();
+        try (InputStream affixStream = openRequiredResource(classLoader, "hunspell/en/index.aff");
+                InputStream dictionaryStream = openRequiredResource(classLoader, "hunspell/en/index.dic")) {
+            return new Dictionary(affixStream, List.of(dictionaryStream), true, SortingStrategy.inMemory());
+        }
+    }
+
+    /**
+     * Opens a required classpath resource.
+     *
+     * @param classLoader class loader
+     * @param path        resource path
+     * @return resource stream
+     */
+    private static InputStream openRequiredResource(final ClassLoader classLoader, final String path) {
+        final InputStream stream = classLoader.getResourceAsStream(path);
+        if (stream == null) {
+            throw new IllegalStateException("Missing benchmark-only Hunspell resource: " + path);
+        }
+        return stream;
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/EnglishRadixorDictionaryCoverageBenchmark.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/EnglishRadixorDictionaryCoverageBenchmark.java
@@ -0,0 +1,431 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.io.BufferedReader;
+import java.io.IOException;
+import java.io.InputStream;
+import java.io.InputStreamReader;
+import java.nio.charset.StandardCharsets;
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.HashMap;
+import java.util.HashSet;
+import java.util.List;
+import java.util.Map;
+import java.util.Objects;
+import java.util.Set;
+import java.util.concurrent.TimeUnit;
+import java.util.zip.GZIPInputStream;
+
+import org.egothor.stemmer.CompiledPatchCommand;
+import org.egothor.stemmer.FrequencyTrie;
+import org.egothor.stemmer.FrequencyTrieBuilders;
+import org.egothor.stemmer.PatchCommandEncoder;
+import org.egothor.stemmer.ReductionMode;
+import org.egothor.stemmer.ReductionSettings;
+import org.egothor.stemmer.StemmerDictionaryParser;
+import org.egothor.stemmer.StemmerPatchTrieLoader;
+import org.egothor.stemmer.WordTraversalDirection;
+import org.openjdk.jmh.annotations.AuxCounters;
+import org.openjdk.jmh.annotations.Benchmark;
+import org.openjdk.jmh.annotations.BenchmarkMode;
+import org.openjdk.jmh.annotations.Level;
+import org.openjdk.jmh.annotations.Measurement;
+import org.openjdk.jmh.annotations.Mode;
+import org.openjdk.jmh.annotations.OutputTimeUnit;
+import org.openjdk.jmh.annotations.Param;
+import org.openjdk.jmh.annotations.Scope;
+import org.openjdk.jmh.annotations.Setup;
+import org.openjdk.jmh.annotations.State;
+import org.openjdk.jmh.annotations.Warmup;
+import org.openjdk.jmh.infra.Blackhole;
+
+/**
+ * Measures Radixor English stemming quality and changed-token speed when the
+ * runtime trie is trained from a deterministic percentage of dictionary rows.
+ *
+ * <p>
+ * The measured stemmer always uses {@link CompiledPatchCommand} values. Quality
+ * is evaluated against the complete English dictionary corpus, while speed is
+ * measured over the complete changed-token English corpus used by the comparison
+ * benchmarks.
+ * </p>
+ */
+@BenchmarkMode(Mode.AverageTime)
+@OutputTimeUnit(TimeUnit.NANOSECONDS)
+@Warmup(iterations = 3, time = 1, timeUnit = TimeUnit.SECONDS)
+@Measurement(iterations = 5, time = 1, timeUnit = TimeUnit.SECONDS)
+public class EnglishRadixorDictionaryCoverageBenchmark {
+
+    /**
+     * Shared benchmark state for one dictionary-row coverage percentage.
+     */
+    @State(Scope.Benchmark)
+    public static class CoverageState {
+
+        /**
+         * Percentage of parsed English dictionary rows used to build the Radixor trie.
+         */
+        @Param({ "100", "90", "80", "70", "60", "50", "40", "30", "20", "10" })
+        public int coveragePercent;
+
+        /**
+         * Full English corpus used for exact-root accounting.
+         */
+        private LanguageBenchmarkCorpus.Corpus fullCorpus;
+
+        /**
+         * Complete changed-token English corpus used for speed measurement.
+         */
+        private LanguageBenchmarkCorpus.Corpus changedCorpus;
+
+        /**
+         * Radixor stemmer backed by a trie built from selected dictionary rows.
+         */
+        private RadixorBenchmarkStemmer stemmer;
+
+        /**
+         * Parsed dictionary row count before deterministic coverage selection.
+         */
+        private int totalRowCount;
+
+        /**
+         * Selected dictionary row count for the configured coverage percentage.
+         */
+        private int selectedRowCount;
+
+        /**
+         * Builds the reduced dictionary trie and shared corpora before measurement.
+         *
+         * @throws IOException if the English dictionary resource cannot be read
+         */
+        @Setup(Level.Trial)
+        public void setUp() throws IOException {
+            final List<DictionaryRow> rows = readEnglishRows();
+            this.totalRowCount = rows.size();
+            final List<DictionaryRow> selectedRows = selectRows(rows, this.coveragePercent);
+            this.selectedRowCount = selectedRows.size();
+            this.fullCorpus = LanguageBenchmarkCorpus.createFullCorpus(StemmerPatchTrieLoader.Language.US_UK);
+            this.changedCorpus = LanguageBenchmarkCorpus.createChangedCorpus(StemmerPatchTrieLoader.Language.US_UK);
+            this.stemmer = new RadixorBenchmarkStemmer(buildCompiledTrie(selectedRows));
+        }
+    }
+
+    /**
+     * JMH auxiliary counters for dictionary-row coverage and exact-root agreement.
+     */
+    @State(Scope.Thread)
+    @AuxCounters(AuxCounters.Type.EVENTS)
+    public static class CoverageCounters {
+
+        /**
+         * Number of exact output/root matches over the full dictionary corpus.
+         */
+        public long correctMatches;
+
+        /**
+         * Number of evaluated tokens over the full dictionary corpus.
+         */
+        public long evaluatedTokens;
+
+        /**
+         * Number of exact output/root matches where token and root differ.
+         */
+        public long changedCorrectMatches;
+
+        /**
+         * Number of evaluated tokens where token and root differ.
+         */
+        public long changedEvaluatedTokens;
+
+        /**
+         * Number of exact output/root matches where token already equals root.
+         */
+        public long rootPreservedMatches;
+
+        /**
+         * Number of evaluated tokens where token already equals root.
+         */
+        public long rootEvaluatedTokens;
+
+        /**
+         * Number of parsed dictionary rows used for trie construction.
+         */
+        public long selectedRows;
+
+        /**
+         * Total number of parsed dictionary rows available.
+         */
+        public long totalRows;
+
+        /**
+         * Resets counters before each measured iteration.
+         */
+        @Setup(Level.Iteration)
+        public void reset() {
+            this.correctMatches = 0L;
+            this.evaluatedTokens = 0L;
+            this.changedCorrectMatches = 0L;
+            this.changedEvaluatedTokens = 0L;
+            this.rootPreservedMatches = 0L;
+            this.rootEvaluatedTokens = 0L;
+            this.selectedRows = 0L;
+            this.totalRows = 0L;
+        }
+    }
+
+    /**
+     * Measures direct Radixor stemming over the complete English changed-token
+     * corpus.
+     *
+     * @param state     shared coverage state
+     * @param blackhole result sink
+     */
+    @Benchmark
+    public void changedTokenStemmingSpeed(final CoverageState state, final Blackhole blackhole) {
+        final String[] tokens = state.changedCorpus.tokens();
+        final RadixorBenchmarkStemmer stemmer = state.stemmer;
+        for (String token : tokens) {
+            blackhole.consume(stemmer.stem(token));
+        }
+    }
+
+    /**
+     * Measures exact-root agreement over the complete English dictionary corpus.
+     *
+     * @param state     shared coverage state
+     * @param counters  auxiliary exact-root counters
+     * @param blackhole result sink
+     * @return exact-root match count for one benchmark operation
+     */
+    @Benchmark
+    public int exactRootAgreement(final CoverageState state, final CoverageCounters counters,
+            final Blackhole blackhole) {
+        final QualityCounts counts = evaluate(state.fullCorpus, state.stemmer, blackhole);
+        counters.correctMatches += counts.correctMatches();
+        counters.evaluatedTokens += counts.evaluatedTokens();
+        counters.changedCorrectMatches += counts.changedCorrectMatches();
+        counters.changedEvaluatedTokens += counts.changedEvaluatedTokens();
+        counters.rootPreservedMatches += counts.rootPreservedMatches();
+        counters.rootEvaluatedTokens += counts.rootEvaluatedTokens();
+        counters.selectedRows += state.selectedRowCount;
+        counters.totalRows += state.totalRowCount;
+        return counts.correctMatches();
+    }
+
+    private static QualityCounts evaluate(final LanguageBenchmarkCorpus.Corpus corpus,
+            final RadixorBenchmarkStemmer stemmer, final Blackhole blackhole) {
+        final String[] tokens = corpus.tokens();
+        final String[] roots = corpus.expectedRoots();
+        int correct = 0;
+        int changedCorrect = 0;
+        int changedEvaluated = 0;
+        int rootPreserved = 0;
+        int rootEvaluated = 0;
+        for (int index = 0; index < tokens.length; index++) {
+            final String token = tokens[index];
+            final String root = roots[index];
+            final String actual = stemmer.stem(token);
+            blackhole.consume(actual);
+            final boolean exact = Objects.equals(root, actual);
+            if (exact) {
+                correct++;
+            }
+            if (Objects.equals(token, root)) {
+                rootEvaluated++;
+                if (exact) {
+                    rootPreserved++;
+                }
+            } else {
+                changedEvaluated++;
+                if (exact) {
+                    changedCorrect++;
+                }
+            }
+        }
+        return new QualityCounts(correct, tokens.length, changedCorrect, changedEvaluated, rootPreserved,
+                rootEvaluated);
+    }
+
+    private static FrequencyTrie<CompiledPatchCommand> buildCompiledTrie(final List<DictionaryRow> rows) {
+        final ReductionSettings settings = new ReductionSettings(
+                ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS,
+                ReductionSettings.DEFAULT_DOMINANT_WINNER_MIN_PERCENT,
+                ReductionSettings.DEFAULT_DOMINANT_WINNER_OVER_SECOND_RATIO,
+                true);
+        final FrequencyTrie.Builder<String> builder = new FrequencyTrie.Builder<>(String[]::new, settings,
+                WordTraversalDirection.BACKWARD);
+        final PatchCommandEncoder encoder = PatchCommandEncoder.builder()
+                .traversalDirection(WordTraversalDirection.BACKWARD)
+                .build();
+
+        for (DictionaryRow row : rows) {
+            builder.put(row.stem(), encoder.encode(row.stem(), row.stem()));
+            for (String variant : row.variants()) {
+                if (!variant.equals(row.stem())) {
+                    builder.put(variant, encoder.encode(variant, row.stem()));
+                }
+            }
+        }
+
+        final FrequencyTrie<String> trie = builder.build();
+        final Map<String, CompiledPatchCommand> compiledCommands = new HashMap<String, CompiledPatchCommand>(4096);
+        return FrequencyTrieBuilders.mapValues(trie, CompiledPatchCommand[]::new, trie.metadata().reductionSettings(),
+                patch -> compiledCommands.computeIfAbsent(patch,
+                        value -> CompiledPatchCommand.compile(value, trie.traversalDirection())));
+    }
+
+    private static List<DictionaryRow> selectRows(final List<DictionaryRow> rows, final int coveragePercent) {
+        if (coveragePercent < 1 || coveragePercent > 100) {
+            throw new IllegalArgumentException("coveragePercent must be between 1 and 100.");
+        }
+        if (coveragePercent == 100) {
+            return List.copyOf(rows);
+        }
+
+        final int selectedCount = Math.max(1, Math.round(rows.size() * coveragePercent / 100.0F));
+        final List<DictionaryRow> rankedRows = new ArrayList<DictionaryRow>(rows);
+        rankedRows.sort(Comparator.comparingLong(DictionaryRow::rank).thenComparingInt(DictionaryRow::lineNumber));
+
+        final Set<Integer> selectedLineNumbers = new HashSet<Integer>(selectedCount);
+        for (int index = 0; index < selectedCount; index++) {
+            selectedLineNumbers.add(rankedRows.get(index).lineNumber());
+        }
+
+        final List<DictionaryRow> selectedRows = new ArrayList<DictionaryRow>(selectedCount);
+        for (DictionaryRow row : rows) {
+            if (selectedLineNumbers.contains(row.lineNumber())) {
+                selectedRows.add(row);
+            }
+        }
+        return selectedRows;
+    }
+
+    private static List<DictionaryRow> readEnglishRows() throws IOException {
+        final String resourcePath = StemmerPatchTrieLoader.Language.US_UK.resourcePath();
+        final InputStream resource = StemmerPatchTrieLoader.class.getClassLoader().getResourceAsStream(resourcePath);
+        if (resource == null) {
+            throw new IllegalStateException("Missing bundled English dictionary resource " + resourcePath + ".");
+        }
+
+        final List<DictionaryRow> rows = new ArrayList<DictionaryRow>(400_000);
+        try (InputStream inputStream = resource;
+                GZIPInputStream gzipInputStream = new GZIPInputStream(inputStream);
+                InputStreamReader inputStreamReader = new InputStreamReader(gzipInputStream, StandardCharsets.UTF_8);
+                BufferedReader reader = new BufferedReader(inputStreamReader)) {
+            StemmerDictionaryParser.parse(reader, resourcePath, (stem, variants, lineNumber) -> {
+                rows.add(new DictionaryRow(lineNumber, stem, variants, rank(lineNumber, stem, variants)));
+            });
+        }
+        return rows;
+    }
+
+    private static long rank(final int lineNumber, final String stem, final String[] variants) {
+        long hash = 0xcbf29ce484222325L;
+        hash = mix(hash, lineNumber);
+        hash = mix(hash, stem);
+        for (String variant : variants) {
+            hash = mix(hash, variant);
+        }
+        return hash;
+    }
+
+    private static long mix(final long hash, final int value) {
+        long result = hash;
+        result ^= value & 0xFFL;
+        result *= 0x100000001b3L;
+        result ^= value >>> 8 & 0xFFL;
+        result *= 0x100000001b3L;
+        result ^= value >>> 16 & 0xFFL;
+        result *= 0x100000001b3L;
+        result ^= value >>> 24 & 0xFFL;
+        result *= 0x100000001b3L;
+        return result;
+    }
+
+    private static long mix(final long hash, final String value) {
+        long result = hash;
+        for (int index = 0; index < value.length(); index++) {
+            final char character = value.charAt(index);
+            result ^= character & 0xFFL;
+            result *= 0x100000001b3L;
+            result ^= character >>> 8;
+            result *= 0x100000001b3L;
+        }
+        result ^= 0xFFL;
+        result *= 0x100000001b3L;
+        return result;
+    }
+
+    /**
+     * One parsed dictionary row with deterministic selection rank.
+     *
+     * @param lineNumber source dictionary line number
+     * @param stem       canonical stem from the first column
+     * @param variants   normalized variants from following columns
+     * @param rank       deterministic selection rank
+     */
+    private record DictionaryRow(int lineNumber, String stem, String[] variants, long rank) {
+
+        /**
+         * Creates one immutable dictionary row snapshot.
+         *
+         * @param lineNumber source dictionary line number
+         * @param stem       canonical stem from the first column
+         * @param variants   normalized variants from following columns
+         * @param rank       deterministic selection rank
+         */
+        DictionaryRow {
+            Objects.requireNonNull(stem, "stem");
+            variants = variants.clone();
+        }
+
+        @Override
+        public String[] variants() {
+            return this.variants.clone();
+        }
+    }
+
+    /**
+     * Exact-root accounting result for one quality operation.
+     *
+     * @param correctMatches         exact-root matches for all tokens
+     * @param evaluatedTokens        evaluated token count
+     * @param changedCorrectMatches  exact-root matches for changed tokens
+     * @param changedEvaluatedTokens evaluated changed-token count
+     * @param rootPreservedMatches   exact-root matches for root-equal tokens
+     * @param rootEvaluatedTokens    evaluated root-equal token count
+     */
+    private record QualityCounts(int correctMatches, int evaluatedTokens, int changedCorrectMatches,
+            int changedEvaluatedTokens, int rootPreservedMatches, int rootEvaluatedTokens) {
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/EnglishStemmerComparisonBenchmark.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/EnglishStemmerComparisonBenchmark.java
@@ -32,140 +32,314 @@ package org.egothor.stemmer.benchmark;

 import java.io.IOException;
 import java.util.concurrent.TimeUnit;
+import java.util.logging.Logger;
+
+import org.apache.lucene.analysis.TokenStream;
+import org.apache.lucene.analysis.en.EnglishMinimalStemFilter;
+import org.apache.lucene.analysis.en.EnglishPossessiveFilter;
+import org.apache.lucene.analysis.en.KStemFilter;
+import org.apache.lucene.analysis.en.PorterStemFilter;
+import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;
+import org.egothor.stemmer.benchmark.snowball.ext.englishStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.porterStemmer;

-import org.egothor.stemmer.FrequencyTrie;
-import org.egothor.stemmer.PatchCommandEncoder;
-import org.egothor.stemmer.ReductionMode;
-import org.egothor.stemmer.StemmerPatchTrieLoader;
 import org.openjdk.jmh.annotations.Benchmark;
 import org.openjdk.jmh.annotations.BenchmarkMode;
 import org.openjdk.jmh.annotations.Level;
 import org.openjdk.jmh.annotations.Measurement;
 import org.openjdk.jmh.annotations.Mode;
 import org.openjdk.jmh.annotations.OutputTimeUnit;
-import org.openjdk.jmh.annotations.Param;
 import org.openjdk.jmh.annotations.Scope;
 import org.openjdk.jmh.annotations.Setup;
 import org.openjdk.jmh.annotations.State;
 import org.openjdk.jmh.annotations.Warmup;
 import org.openjdk.jmh.infra.Blackhole;
-import org.tartarus.snowball.ext.englishStemmer;
-import org.tartarus.snowball.ext.porterStemmer;
+
+import org.egothor.stemmer.FrequencyTrie;
+import org.egothor.stemmer.StemmerDictionaryParser;
+import org.egothor.stemmer.ReductionMode;
+import org.egothor.stemmer.StemmerPatchTrieLoader;

 /**
- * Compares English stemming throughput across Radixor and Snowball stemmers.
+ * Compares English stemming throughput across Radixor and selected Java
+ * algorithm paths with a shared deterministic corpus.
 *
 * <p>
- * The benchmark processes the same deterministic token array with:
+ * The comparison uses one shared changed-token dictionary array for all methods:
 * </p>
 * <ul>
- * <li>Radixor using bundled {@link StemmerPatchTrieLoader.Language#US_UK}</li>
- * <li>Snowball original Porter stemmer</li>
- * <li>Snowball English stemmer, commonly referred to as Porter2</li>
+ * <li>Radixor direct dictionary lookup</li>
+ * <li>Snowball Porter</li>
+ * <li>Snowball English (Porter2)</li>
+ * <li>Lucene direct Porter API (generated copy)</li>
+ * <li>Lucene Porter, KStem, and EnglishMinimal token-filter paths</li>
+ * <li>Benchmark-only Paice/Husk Lancaster baseline</li>
 * </ul>
- *
- * <p>
- * This benchmark compares throughput on a shared workload. It does not imply
- * that the algorithms are linguistically equivalent.
- * </p>
 */
@BenchmarkMode(Mode.AverageTime)
@OutputTimeUnit(TimeUnit.NANOSECONDS)
-@Warmup(iterations = 3, time = 1)
-@Measurement(iterations = 5, time = 1)
+@Warmup(iterations = 3, time = 1, timeUnit = TimeUnit.SECONDS)
+@Measurement(iterations = 5, time = 1, timeUnit = TimeUnit.SECONDS)
 public class EnglishStemmerComparisonBenchmark {

    /**
-     * Shared benchmark data.
+     * Shared, parameterized benchmark corpus state.
     */
    @State(Scope.Benchmark)
    public static class SharedState {

        /**
-         * Number of generated lexical families.
-         */
-        @Param({ "1000", "5000" })
-        public int familyCount;
-
-        /**
-         * Token workload processed by all compared stemmers.
+         * Shared deterministic token corpus.
         */
        private String[] tokens;

        /**
-         * Radixor trie loaded from the bundled professional English dictionary.
+         * Radixor benchmark adapter for the US/UK benchmark corpus.
         */
-        private FrequencyTrie<String> radixorTrie;
+        private RadixorBenchmarkStemmer radixorStemmer;

        /**
-         * Initializes the shared benchmark state.
-         *
-         * @throws IOException if the bundled Radixor dictionary cannot be loaded
+         * Initializes shared corpus and trie state once per trial.
         */
        @Setup(Level.Trial)
-        public void setUp() throws IOException {
-            this.tokens = EnglishComparisonCorpus.createTokens(this.familyCount);
-            this.radixorTrie = StemmerPatchTrieLoader.load(StemmerPatchTrieLoader.Language.US_UK, true,
-                    ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS);
+        public void setUp() throws java.io.IOException {
+            Logger.getLogger(StemmerDictionaryParser.class.getName())
+                    .setLevel(java.util.logging.Level.OFF);
+            Logger.getLogger(StemmerDictionaryParser.class.getName()).setUseParentHandlers(false);
+            this.tokens = EnglishComparisonCorpus.createTokens();
+            this.radixorStemmer = new RadixorBenchmarkStemmer(StemmerPatchTrieLoader.loadCompiled(
+                    StemmerPatchTrieLoader.Language.US_UK, true,
+                    ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS));
        }
    }

    /**
-     * Per-thread reusable Snowball stemmers.
+     * Reusable direct stemmer instances.
     */
    @State(Scope.Thread)
-    public static class SnowballState {
+    public static class DirectStemmerState {

        /**
-         * Adapter for the original Porter stemmer.
+         * Snowball classic Porter.
         */
        private SnowballStemmerAdapter porterStemmer;

        /**
-         * Adapter for the Snowball English stemmer.
+         * Snowball English (Porter2) for legacy and dictionary comparison.
         */
-        private SnowballStemmerAdapter englishStemmer;
+        private SnowballStemmerAdapter englishPorterStemmer;

        /**
-         * Initializes reusable Snowball stemmers for the executing thread.
+         * Generated Lucene direct Porter implementation copy.
+         */
+        private LucenePorterStemmerCopied lucenePorter;
+
+        /**
+         * Benchmark-only Paice/Husk Lancaster implementation.
+         */
+        private PaiceHuskLancasterStemmer paiceHuskLancaster;
+
+        /**
+         * Apache OpenNLP Porter stemmer.
+         */
+        private opennlp.tools.stemmer.PorterStemmer openNlpPorterStemmer;
+
+        /**
+         * Initializes mutable stemmer instances reused by all benchmark calls.
         */
        @Setup(Level.Trial)
        public void setUp() {
            this.porterStemmer = new SnowballStemmerAdapter(porterStemmer::new);
-            this.englishStemmer = new SnowballStemmerAdapter(englishStemmer::new);
+            this.englishPorterStemmer = new SnowballStemmerAdapter(englishStemmer::new);
+            this.lucenePorter = new LucenePorterStemmerCopied();
+            this.paiceHuskLancaster = new PaiceHuskLancasterStemmer();
+            this.openNlpPorterStemmer = new opennlp.tools.stemmer.PorterStemmer();
+        }
+    }
+
+    /**
+     * Reusable Lucene token streams and filters.
+     */
+    @State(Scope.Thread)
+    public static class LuceneFilterState {
+
+        /**
+         * Reused Porter filter stream input.
+         */
+        private final EnglishStemmerComparisonTokenStream porterStemFilterInput;
+
+        /**
+         * Porter token filter for public API integration-path comparison.
+         */
+        private final PorterStemFilter porterStemFilter;
+
+        /**
+         * Porter filter attributes.
+         */
+        private final CharTermAttribute porterStemFilterTerm;
+
+        /**
+         * Reused KStem filter stream input.
+         */
+        private final EnglishStemmerComparisonTokenStream kStemFilterInput;
+
+        /**
+         * KStem token filter for a second Lucene English baseline.
+         */
+        private final KStemFilter kStemFilter;
+
+        /**
+         * KStem filter attributes.
+         */
+        private final CharTermAttribute kStemTerm;
+
+        /**
+         * Reused minimal stem filter stream input.
+         */
+        private final EnglishStemmerComparisonTokenStream englishMinimalStemFilterInput;
+
+        /**
+         * EnglishMinimal token filter.
+         */
+        private final EnglishMinimalStemFilter englishMinimalStemFilter;
+
+        /**
+         * EnglishMinimal filter attributes.
+         */
+        private final CharTermAttribute englishMinimalTerm;
+
+        /**
+         * Reused English possessive filter stream input.
+         */
+        private final EnglishStemmerComparisonTokenStream englishPossessiveFilterInput;
+
+        /**
+         * English possessive filter.
+         */
+        private final EnglishPossessiveFilter englishPossessiveFilter;
+
+        /**
+         * English possessive filter attributes.
+         */
+        private final CharTermAttribute englishPossessiveTerm;
+
+        /**
+         * Creates benchmark stream/filter state and attaches token attributes.
+         */
+        public LuceneFilterState() {
+            this.porterStemFilterInput = new EnglishStemmerComparisonTokenStream(new String[0]);
+            this.porterStemFilter = new PorterStemFilter(this.porterStemFilterInput);
+            this.porterStemFilterTerm = this.porterStemFilter.getAttribute(CharTermAttribute.class);
+
+            this.kStemFilterInput = new EnglishStemmerComparisonTokenStream(new String[0]);
+            this.kStemFilter = new KStemFilter(this.kStemFilterInput);
+            this.kStemTerm = this.kStemFilter.getAttribute(CharTermAttribute.class);
+
+            this.englishMinimalStemFilterInput = new EnglishStemmerComparisonTokenStream(new String[0]);
+            this.englishMinimalStemFilter = new EnglishMinimalStemFilter(this.englishMinimalStemFilterInput);
+            this.englishMinimalTerm = this.englishMinimalStemFilter.getAttribute(CharTermAttribute.class);
+
+            this.englishPossessiveFilterInput = new EnglishStemmerComparisonTokenStream(new String[0]);
+            this.englishPossessiveFilter = new EnglishPossessiveFilter(this.englishPossessiveFilterInput);
+            this.englishPossessiveTerm = this.englishPossessiveFilter.getAttribute(CharTermAttribute.class);
+        }
+
+        /**
+         * Rebinds the shared corpus and resets all streams for another measured
+         * operation.
+         *
+         * <p>
+         * The {@code String[]} to Lucene character-buffer conversion is deliberately
+         * performed every time so TokenFilter benchmarks include the cost of adapting
+         * the benchmark's canonical string corpus to Lucene's mutable token
+         * attributes.
+         * </p>
+         *
+         * @param tokens benchmark token corpus
+         */
+        void configure(final String[] tokens) throws IOException {
+            this.porterStemFilterInput.setTokens(tokens);
+            this.kStemFilterInput.setTokens(tokens);
+            this.englishMinimalStemFilterInput.setTokens(tokens);
+            this.englishPossessiveFilterInput.setTokens(tokens);
+
+            this.porterStemFilter.reset();
+            this.kStemFilter.reset();
+            this.englishMinimalStemFilter.reset();
+            this.englishPossessiveFilter.reset();
+        }
+
+        /**
+         * Reuses one mutable filter stream and returns all emitted tokens to blackhole.
+         *
+         * @param stream     benchmark token stream with configured filter
+         * @param term       token text attribute
+         * @param blackhole  sink
+         * @throws IOException on token stream failure
+         */
+        private static void consume(final TokenStream stream, final CharTermAttribute term, final Blackhole blackhole)
+                throws IOException {
+            while (stream.incrementToken()) {
+                blackhole.consume(term.toString());
+            }
+            stream.end();
+        }
+
+        /**
+         * Executes Porter filter over the shared corpus.
+         *
+         * @param blackhole sink
+         * @throws IOException if tokenization fails
+         */
+        void runPorterStemFilter(final Blackhole blackhole) throws IOException {
+            consume(this.porterStemFilter, this.porterStemFilterTerm, blackhole);
+        }
+
+        /**
+         * Executes KStem filter over the shared corpus.
+         *
+         * @param blackhole sink
+         * @throws IOException if tokenization fails
+         */
+        void runKStemFilter(final Blackhole blackhole) throws IOException {
+            consume(this.kStemFilter, this.kStemTerm, blackhole);
+        }
+
+        /**
+         * Executes English minimal filter over the shared corpus.
+         *
+         * @param blackhole sink
+         * @throws IOException if tokenization fails
+         */
+        void runEnglishMinimalStemFilter(final Blackhole blackhole) throws IOException {
+            consume(this.englishMinimalStemFilter, this.englishMinimalTerm, blackhole);
+        }
+
+        /**
+         * Executes English possessive filter over the shared corpus.
+         *
+         * @param blackhole sink
+         * @throws IOException if tokenization fails
+         */
+        void runEnglishPossessiveFilter(final Blackhole blackhole) throws IOException {
+            consume(this.englishPossessiveFilter, this.englishPossessiveTerm, blackhole);
        }
    }

    /**
     * Measures Radixor preferred-result stemming throughput.
     *
-     * @param sharedState shared benchmark data
-     * @param blackhole   sink preventing dead-code elimination
+     * <p>
+     * This path uses a single shared dictionary lookup and patch application.
+     * </p>
+     *
+     * @param sharedState shared corpus and trie
+     * @param blackhole   result sink
     */
    @Benchmark
    public void radixorUsUkProfiPreferredStem(final SharedState sharedState, final Blackhole blackhole) {
        final String[] tokens = sharedState.tokens;
-        final FrequencyTrie<String> trie = sharedState.radixorTrie;
-
-        for (String token : tokens) {
-            final String patch = trie.get(token);
-            final String stem = patch == null ? token : PatchCommandEncoder.apply(token, patch);
-            blackhole.consume(stem);
-        }
-    }
-
-    /**
-     * Measures Snowball original Porter stemming throughput.
-     *
-     * @param sharedState   shared benchmark data
-     * @param snowballState reusable Snowball stemmers
-     * @param blackhole     sink preventing dead-code elimination
-     */
-    @Benchmark
-    public void snowballOriginalPorter(final SharedState sharedState, final SnowballState snowballState,
-            final Blackhole blackhole) {
-        final String[] tokens = sharedState.tokens;
-        final SnowballStemmerAdapter stemmer = snowballState.porterStemmer;
+        final RadixorBenchmarkStemmer stemmer = sharedState.radixorStemmer;

        for (String token : tokens) {
            blackhole.consume(stemmer.stem(token));
@@ -173,25 +347,169 @@ public class EnglishStemmerComparisonBenchmark {
    }

    /**
-     * Measures Snowball English stemming throughput.
+     * Measures the canonical Snowball Porter stemming throughput used by the
+     * performance badge.
     *
     * <p>
-     * Snowball English is the newer English stemmer commonly referred to as
-     * Porter2.
+     * This uses Snowball classic Porter as a direct stemmer API call and includes
+     * no Lucene token stream integration overhead.
     * </p>
     *
-     * @param sharedState   shared benchmark data
-     * @param snowballState reusable Snowball stemmers
-     * @param blackhole     sink preventing dead-code elimination
+     * @param sharedState  shared corpus
+     * @param stemmerState reusable Snowball adapter state
+     * @param blackhole    result sink
     */
    @Benchmark
-    public void snowballEnglishPorter2(final SharedState sharedState, final SnowballState snowballState,
+    public void snowballOriginalPorter(final SharedState sharedState, final DirectStemmerState stemmerState,
            final Blackhole blackhole) {
        final String[] tokens = sharedState.tokens;
-        final SnowballStemmerAdapter stemmer = snowballState.englishStemmer;
+        final SnowballStemmerAdapter stemmer = stemmerState.porterStemmer;

        for (String token : tokens) {
            blackhole.consume(stemmer.stem(token));
        }
    }
+
+    /**
+     * Measures Snowball English (Porter2) direct API throughput.
+     *
+     * @param sharedState  shared corpus
+     * @param stemmerState reusable Snowball adapter state
+     * @param blackhole    result sink
+     */
+    @Benchmark
+    public void snowballEnglishPorter2(final SharedState sharedState, final DirectStemmerState stemmerState,
+            final Blackhole blackhole) {
+        final String[] tokens = sharedState.tokens;
+        final SnowballStemmerAdapter stemmer = stemmerState.englishPorterStemmer;
+
+        for (String token : tokens) {
+            blackhole.consume(stemmer.stem(token));
+        }
+    }
+
+    /**
+     * Measures Lucene generated Porter stemmer API throughput.
+     *
+     * <p>
+     * This path is a generated copy of Lucene&apos;s package-private PorterStemmer
+     * class, compiled into the JMH source set only.
+     * </p>
+     *
+     * @param sharedState  shared corpus
+     * @param stemmerState reusable Lucene copied API state
+     * @param blackhole    result sink
+     */
+    @Benchmark
+    public void lucenePorterStemmerCopied(final SharedState sharedState, final DirectStemmerState stemmerState,
+            final Blackhole blackhole) {
+        final String[] tokens = sharedState.tokens;
+        final LucenePorterStemmerCopied stemmer = stemmerState.lucenePorter;
+
+        for (String token : tokens) {
+            blackhole.consume(stemmer.stem(token));
+        }
+    }
+
+    /**
+     * Measures Lucene Porter token-filter integration throughput.
+     *
+     * <p>
+     * This includes stream, reusable token attributes, and filter overhead and is
+     * not equivalent to a direct API stemmer call.
+     * </p>
+     *
+     * @param sharedState shared corpus
+     * @param filterState reusable filter state
+     * @param blackhole    sink
+     * @throws IOException if token stream fails
+     */
+    @Benchmark
+    public void lucenePorterStemFilter(final SharedState sharedState, final LuceneFilterState filterState,
+            final Blackhole blackhole) throws IOException {
+        filterState.configure(sharedState.tokens);
+        filterState.runPorterStemFilter(blackhole);
+    }
+
+    /**
+     * Measures Lucene KStem integration-path throughput.
+     *
+     * @param sharedState shared corpus
+     * @param filterState reusable filter state
+     * @param blackhole    sink
+     * @throws IOException if token stream fails
+     */
+    @Benchmark
+    public void luceneKStemFilter(final SharedState sharedState, final LuceneFilterState filterState,
+            final Blackhole blackhole) throws IOException {
+        filterState.configure(sharedState.tokens);
+        filterState.runKStemFilter(blackhole);
+    }
+
+    /**
+     * Measures Lucene EnglishMinimal integration-path throughput.
+     *
+     * @param sharedState shared corpus
+     * @param filterState reusable filter state
+     * @param blackhole    sink
+     * @throws IOException if token stream fails
+     */
+    @Benchmark
+    public void luceneEnglishMinimalStemFilter(final SharedState sharedState, final LuceneFilterState filterState,
+            final Blackhole blackhole) throws IOException {
+        filterState.configure(sharedState.tokens);
+        filterState.runEnglishMinimalStemFilter(blackhole);
+    }
+
+    /**
+     * Measures benchmark-only Paice/Husk Lancaster throughput.
+     *
+     * @param sharedState  shared corpus
+     * @param stemmerState reusable Paice/Husk instance
+     * @param blackhole    sink
+     */
+    @Benchmark
+    public void paiceHuskLancaster(final SharedState sharedState, final DirectStemmerState stemmerState,
+            final Blackhole blackhole) {
+        final String[] tokens = sharedState.tokens;
+        final PaiceHuskLancasterStemmer stemmer = stemmerState.paiceHuskLancaster;
+
+        for (String token : tokens) {
+            blackhole.consume(stemmer.stem(token));
+        }
+    }
+
+    /**
+     * Measures Apache OpenNLP Porter stemming throughput.
+     *
+     * @param sharedState  shared corpus
+     * @param stemmerState reusable OpenNLP Porter instance
+     * @param blackhole    sink
+     */
+    @Benchmark
+    public void opennlpPorterStemmer(final SharedState sharedState, final DirectStemmerState stemmerState,
+            final Blackhole blackhole) {
+        final String[] tokens = sharedState.tokens;
+        final opennlp.tools.stemmer.PorterStemmer stemmer = stemmerState.openNlpPorterStemmer;
+
+        for (String token : tokens) {
+            blackhole.consume(stemmer.stem(token).toString());
+        }
+    }
+
+    /**
+     * Measures Lucene EnglishPossessiveFilter as a narrow possessive-removal
+     * baseline.
+     *
+     * @param sharedState shared corpus
+     * @param filterState reusable filter state
+     * @param blackhole   sink
+     * @throws IOException if token stream fails
+     */
+    @Benchmark
+    public void luceneEnglishPossessiveFilter(final SharedState sharedState, final LuceneFilterState filterState,
+            final Blackhole blackhole) throws IOException {
+        filterState.configure(sharedState.tokens);
+        filterState.runEnglishPossessiveFilter(blackhole);
+    }
 }
--- a/src/jmh/java/org/egothor/stemmer/benchmark/EnglishStemmerComparisonBenchmarkQuality.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/EnglishStemmerComparisonBenchmarkQuality.java
@@ -0,0 +1,258 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.io.IOException;
+import java.util.Objects;
+import java.util.concurrent.TimeUnit;
+
+import org.egothor.stemmer.FrequencyTrie;
+import org.egothor.stemmer.ReductionMode;
+import org.egothor.stemmer.StemmerPatchTrieLoader;
+import org.egothor.stemmer.benchmark.snowball.ext.porterStemmer;
+import org.openjdk.jmh.annotations.AuxCounters;
+import org.openjdk.jmh.annotations.Benchmark;
+import org.openjdk.jmh.annotations.BenchmarkMode;
+import org.openjdk.jmh.annotations.Fork;
+import org.openjdk.jmh.annotations.Level;
+import org.openjdk.jmh.annotations.Measurement;
+import org.openjdk.jmh.annotations.Mode;
+import org.openjdk.jmh.annotations.OutputTimeUnit;
+import org.openjdk.jmh.annotations.Scope;
+import org.openjdk.jmh.annotations.Setup;
+import org.openjdk.jmh.annotations.State;
+import org.openjdk.jmh.annotations.Warmup;
+import org.openjdk.jmh.infra.Blackhole;
+
+/**
+ * Emits exact-root agreement metrics for the canonical English badge pair.
+ *
+ * <p>
+ * This class is deliberately named so the existing focused include pattern for
+ * English stemmer comparison benchmarks includes it. The benchmark methods are
+ * separate from throughput methods so equality checks do not contaminate timing
+ * scores.
+ * </p>
+ */
+@BenchmarkMode(Mode.AverageTime)
+@OutputTimeUnit(TimeUnit.NANOSECONDS)
+@Warmup(iterations = 0)
+@Measurement(iterations = 1, time = 1, timeUnit = TimeUnit.MILLISECONDS)
+@Fork(0)
+public class EnglishStemmerComparisonBenchmarkQuality {
+
+    /**
+     * Shared English quality corpus and stemmer state.
+     */
+    @State(Scope.Benchmark)
+    public static class SharedState {
+
+        /**
+         * Complete English resource-derived corpus.
+         */
+        private LanguageBenchmarkCorpus.Corpus corpus;
+
+        /**
+         * Compiled Radixor English trie.
+         */
+        private RadixorBenchmarkStemmer radixorStemmer;
+
+        /**
+         * Reusable Snowball Porter adapter.
+         */
+        private SnowballStemmerAdapter porterStemmer;
+
+        /**
+         * Initializes quality resources.
+         *
+         * @throws IOException if corpus or trie loading fails
+         */
+        @Setup(Level.Trial)
+        public void setUp() throws IOException {
+            this.corpus = LanguageBenchmarkCorpus.createFullCorpus(StemmerPatchTrieLoader.Language.US_UK);
+            this.radixorStemmer = new RadixorBenchmarkStemmer(StemmerPatchTrieLoader.loadCompiled(
+                    StemmerPatchTrieLoader.Language.US_UK, true,
+                    ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS));
+            this.porterStemmer = new SnowballStemmerAdapter(porterStemmer::new);
+        }
+    }
+
+    /**
+     * JMH auxiliary counters for exact-root agreement.
+     */
+    @State(Scope.Thread)
+    @AuxCounters(AuxCounters.Type.EVENTS)
+    public static class AccuracyCounters {
+
+        /**
+         * Number of exact-root matches.
+         */
+        public long correctMatches;
+
+        /**
+         * Number of evaluated tokens.
+         */
+        public long evaluatedTokens;
+
+        /**
+         * Number of exact-root matches where the input token differs from the
+         * expected root.
+         */
+        public long changedCorrectMatches;
+
+        /**
+         * Number of evaluated tokens where the input token differs from the expected
+         * root.
+         */
+        public long changedEvaluatedTokens;
+
+        /**
+         * Number of exact-root matches where the input token is already the expected
+         * root.
+         */
+        public long rootPreservedMatches;
+
+        /**
+         * Number of evaluated tokens where the input token is already the expected
+         * root.
+         */
+        public long rootEvaluatedTokens;
+
+        /**
+         * Resets counters before each measured iteration.
+         */
+        @Setup(Level.Iteration)
+        public void reset() {
+            this.correctMatches = 0L;
+            this.evaluatedTokens = 0L;
+            this.changedCorrectMatches = 0L;
+            this.changedEvaluatedTokens = 0L;
+            this.rootPreservedMatches = 0L;
+            this.rootEvaluatedTokens = 0L;
+        }
+    }
+
+    /**
+     * Evaluates exact-root agreement for the canonical Radixor badge method.
+     *
+     * @param sharedState shared English quality state
+     * @param counters JMH auxiliary counters
+     * @param blackhole result sink
+     * @return exact-root match count
+     */
+    @Benchmark
+    public int radixorUsUkProfiPreferredStemAccuracy(final SharedState sharedState,
+            final AccuracyCounters counters, final Blackhole blackhole) {
+        return evaluate(sharedState.corpus, sharedState.radixorStemmer::stem, counters, blackhole);
+    }
+
+    /**
+     * Evaluates exact-root agreement for the canonical Snowball Porter badge
+     * method.
+     *
+     * @param sharedState shared English quality state
+     * @param counters JMH auxiliary counters
+     * @param blackhole result sink
+     * @return exact-root match count
+     */
+    @Benchmark
+    public int snowballOriginalPorterAccuracy(final SharedState sharedState,
+            final AccuracyCounters counters, final Blackhole blackhole) {
+        return evaluate(sharedState.corpus, sharedState.porterStemmer::stem, counters, blackhole);
+    }
+
+    /**
+     * Evaluates one stemmer against the expected roots.
+     *
+     * @param corpus token/root corpus
+     * @param stemmer stemmer under evaluation
+     * @param counters JMH auxiliary counters
+     * @param blackhole result sink
+     * @return exact-root match count
+     */
+    private static int evaluate(final LanguageBenchmarkCorpus.Corpus corpus, final Stemmer stemmer,
+            final AccuracyCounters counters, final Blackhole blackhole) {
+        Objects.requireNonNull(corpus, "corpus");
+        Objects.requireNonNull(stemmer, "stemmer");
+
+        int correct = 0;
+        int changedCorrect = 0;
+        int changedEvaluated = 0;
+        int rootPreserved = 0;
+        int rootEvaluated = 0;
+        final String[] tokens = corpus.tokens();
+        final String[] expectedRoots = corpus.expectedRoots();
+        for (int index = 0; index < tokens.length; index++) {
+            final String token = tokens[index];
+            final String expectedRoot = expectedRoots[index];
+            final String actual = stemmer.stem(token);
+            blackhole.consume(actual);
+            final boolean exact = Objects.equals(expectedRoot, actual);
+            if (exact) {
+                correct++;
+            }
+            if (Objects.equals(token, expectedRoot)) {
+                rootEvaluated++;
+                if (exact) {
+                    rootPreserved++;
+                }
+            } else {
+                changedEvaluated++;
+                if (exact) {
+                    changedCorrect++;
+                }
+            }
+        }
+
+        counters.correctMatches += correct;
+        counters.evaluatedTokens += tokens.length;
+        counters.changedCorrectMatches += changedCorrect;
+        counters.changedEvaluatedTokens += changedEvaluated;
+        counters.rootPreservedMatches += rootPreserved;
+        counters.rootEvaluatedTokens += rootEvaluated;
+        return correct;
+    }
+
+    /**
+     * Direct stemmer function.
+     */
+    @FunctionalInterface
+    private interface Stemmer {
+
+        /**
+         * Produces one stem.
+         *
+         * @param token input token
+         * @return produced stem
+         */
+        String stem(String token);
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/EnglishStemmerComparisonTokenStream.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/EnglishStemmerComparisonTokenStream.java
@@ -0,0 +1,145 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ * 
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ * 
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ * 
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ * 
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ * 
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.io.IOException;
+
+import org.apache.lucene.analysis.TokenStream;
+import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;
+import org.apache.lucene.analysis.tokenattributes.OffsetAttribute;
+import org.apache.lucene.analysis.tokenattributes.PositionIncrementAttribute;
+
+/**
+ * Reusable token stream driven by a deterministic token corpus.
+ *
+ * <p>
+ * The stream emits each token from a shared array and supports repeated
+ * {@link #reset()} + {@link #incrementToken()} cycles without per-token
+ * object allocation.
+ * </p>
+ */
+final class EnglishStemmerComparisonTokenStream extends TokenStream {
+
+    /**
+     * Current token text.
+     */
+    private final CharTermAttribute charTermAttribute;
+
+    /**
+     * Token offsets for benchmark stream compliance.
+     */
+    private final OffsetAttribute offsetAttribute;
+
+    /**
+     * Position increment attribute for benchmark stream compliance.
+     */
+    private final PositionIncrementAttribute positionIncrementAttribute;
+
+    /**
+     * Reusable token source.
+     */
+    private final BenchmarkTokenSequence tokenSequence;
+
+    /**
+     * Creates a deterministic token stream for benchmark reuse.
+     *
+     * @param tokens tokens emitted by the stream
+     */
+    EnglishStemmerComparisonTokenStream(final String[] tokens) {
+        this.tokenSequence = new BenchmarkTokenSequence(tokens);
+        this.charTermAttribute = addAttribute(CharTermAttribute.class);
+        this.offsetAttribute = addAttribute(OffsetAttribute.class);
+        this.positionIncrementAttribute = addAttribute(PositionIncrementAttribute.class);
+    }
+
+    /**
+     * Replaces the token corpus for this stream.
+     *
+     * @param tokens new token corpus
+     */
+    void setTokens(final String[] tokens) {
+        this.tokenSequence.setTokens(tokens);
+    }
+
+    /**
+     * Returns whether the stream is drained and ready to be exhausted.
+     *
+     * @return true if all configured tokens were consumed
+     */
+    boolean isDrained() {
+        return !this.tokenSequence.hasNext();
+    }
+
+    /**
+     * {@inheritDoc}
+     */
+    @Override
+    public boolean incrementToken() throws IOException {
+        if (!this.tokenSequence.advance()) {
+            return false;
+        }
+
+        clearAttributes();
+        final char[] token = this.tokenSequence.currentToken();
+        this.charTermAttribute.copyBuffer(token, 0, token.length);
+        this.positionIncrementAttribute.setPositionIncrement(1);
+        this.offsetAttribute.setOffset(this.tokenSequence.currentStartOffset(), this.tokenSequence.currentEndOffset());
+        return true;
+    }
+
+    /**
+     * {@inheritDoc}
+     */
+    @Override
+    public void reset() throws IOException {
+        super.reset();
+        this.tokenSequence.reset();
+    }
+
+    /**
+     * {@inheritDoc}
+     */
+    @Override
+    public void end() throws IOException {
+        super.end();
+        final int endOffset = this.tokenSequence.endOffset();
+        this.offsetAttribute.setOffset(endOffset, endOffset);
+    }
+
+    /**
+     * {@inheritDoc}
+     */
+    @Override
+    public void close() throws IOException {
+        super.close();
+        this.charTermAttribute.setEmpty();
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/FrequencyTrieLookupBenchmark.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/FrequencyTrieLookupBenchmark.java
@@ -65,6 +65,7 @@ import org.openjdk.jmh.infra.Blackhole;
@OutputTimeUnit(TimeUnit.NANOSECONDS)
@Warmup(iterations = 3, time = 1)
@Measurement(iterations = 5, time = 1)
+@SuppressWarnings("deprecation")
 public class FrequencyTrieLookupBenchmark {

    /**
--- a/src/jmh/java/org/egothor/stemmer/benchmark/GermanGoldstandardStemmerComparisonBenchmark.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/GermanGoldstandardStemmerComparisonBenchmark.java
@@ -0,0 +1,579 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.io.BufferedReader;
+import java.io.IOException;
+import java.io.InputStream;
+import java.io.InputStreamReader;
+import java.nio.charset.StandardCharsets;
+import java.util.ArrayList;
+import java.util.LinkedHashMap;
+import java.util.List;
+import java.util.Map;
+import java.util.Objects;
+import java.util.function.Function;
+
+import org.apache.lucene.analysis.LowerCaseFilter;
+import org.apache.lucene.analysis.TokenStream;
+import org.apache.lucene.analysis.de.GermanLightStemFilter;
+import org.apache.lucene.analysis.de.GermanMinimalStemFilter;
+import org.apache.lucene.analysis.de.GermanNormalizationFilter;
+import org.apache.lucene.analysis.de.GermanStemFilter;
+import org.apache.lucene.analysis.snowball.SnowballFilter;
+import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;
+import org.apache.lucene.analysis.tokenattributes.PositionIncrementAttribute;
+import org.openjdk.jmh.annotations.AuxCounters;
+import org.openjdk.jmh.annotations.Benchmark;
+import org.openjdk.jmh.annotations.BenchmarkMode;
+import org.openjdk.jmh.annotations.Fork;
+import org.openjdk.jmh.annotations.Level;
+import org.openjdk.jmh.annotations.Measurement;
+import org.openjdk.jmh.annotations.Mode;
+import org.openjdk.jmh.annotations.OutputTimeUnit;
+import org.openjdk.jmh.annotations.Param;
+import org.openjdk.jmh.annotations.Scope;
+import org.openjdk.jmh.annotations.Setup;
+import org.openjdk.jmh.annotations.State;
+import org.openjdk.jmh.annotations.Warmup;
+import org.openjdk.jmh.infra.Blackhole;
+
+import org.egothor.stemmer.FrequencyTrie;
+import org.egothor.stemmer.ReductionMode;
+import org.egothor.stemmer.StemmerPatchTrieLoader;
+
+/**
+ * German-only stemmer comparison on CISTEM gold standards.
+ *
+ * <p>
+ * Each benchmark operation is fed by one cluster file. The same candidate set is
+ * evaluated twice, once per file, to produce one precision/recall/f-measure
+ * table for each gold standard.
+ * </p>
+ */
+@BenchmarkMode(Mode.AverageTime)
+@OutputTimeUnit(java.util.concurrent.TimeUnit.NANOSECONDS)
+@Warmup(iterations = 3, time = 1, timeUnit = java.util.concurrent.TimeUnit.SECONDS)
+@Measurement(iterations = 5, time = 1, timeUnit = java.util.concurrent.TimeUnit.SECONDS)
+@Fork(1)
+public class GermanGoldstandardStemmerComparisonBenchmark {
+
+    /**
+     * Shared German benchmark state for one dataset and one candidate.
+     */
+    @State(Scope.Benchmark)
+    public static class SharedState {
+
+        /**
+         * Gold standard dataset.
+         */
+        @Param({"goldstandard1.txt", "goldstandard2.txt"})
+        public String goldStandardFileName;
+
+        /**
+         * Candidate stemmer.
+         */
+        @Param({
+                "GERMAN_RADIXOR",
+                "GERMAN_LUCENE_GERMAN_STEM_FILTER",
+                "GERMAN_LUCENE_GERMAN_LIGHT_STEM_FILTER",
+                "GERMAN_LUCENE_GERMAN_MINIMAL_STEM_FILTER",
+                "GERMAN_CISTEM",
+                "SNOWBALL_GERMAN_DIRECT",
+                "SNOWBALL_GERMAN_LUCENE_FILTER"
+        })
+        public String candidateName;
+
+        /**
+         * Parsed gold standard corpus.
+         */
+        private GermanGoldstandardCorpus corpus;
+
+        /**
+         * Gold standard words flattened by cluster order.
+         */
+        private String[] allTokens;
+
+        /**
+         * Candidate evaluator.
+         */
+        private GoldstandardStemmer stemmer;
+
+        /**
+         * Initializes one candidate on one gold standard corpus.
+         *
+         * @throws IOException when the corpus cannot be loaded
+         */
+        @Setup(Level.Trial)
+        public void setUp() throws IOException {
+            this.corpus = loadCorpus(this.goldStandardFileName);
+            this.allTokens = flattenCorpusTokens(this.corpus);
+            this.stemmer = GermanCandidate.valueOf(this.candidateName).createEvaluator();
+        }
+    }
+
+    /**
+     * JMH auxiliary counters for CISTEM-style cluster accounting.
+     */
+    @State(Scope.Thread)
+    @AuxCounters(AuxCounters.Type.EVENTS)
+    public static class GoldstandardQualityCounters {
+
+        /**
+         * True positives across clusters.
+         */
+        public long truePositives;
+
+        /**
+         * False positives across clusters.
+         */
+        public long falsePositives;
+
+        /**
+         * False negatives across clusters.
+         */
+        public long falseNegatives;
+
+        /**
+         * Evaluated clusters.
+         */
+        public long evaluatedClusters;
+
+        /**
+         * Evaluated tokens.
+         */
+        public long evaluatedTokens;
+
+        /**
+         * Resets counters before each measured iteration.
+         */
+        @Setup(Level.Iteration)
+        public void reset() {
+            this.truePositives = 0L;
+            this.falsePositives = 0L;
+            this.falseNegatives = 0L;
+            this.evaluatedClusters = 0L;
+            this.evaluatedTokens = 0L;
+        }
+    }
+
+    /**
+     * Evaluates CISTEM-style precision, recall, and F1-relevant counts.
+     *
+     * @param state shared benchmark state
+     * @param counters quality counters
+     * @param blackhole result sink
+     * @return evaluated token count for this operation
+     * @throws IOException if token filtering cannot run
+     */
+    @Benchmark
+    @Warmup(iterations = 0)
+    @Measurement(iterations = 1, time = 1, timeUnit = java.util.concurrent.TimeUnit.MILLISECONDS)
+    @Fork(0)
+    public long cistemStyleQuality(final SharedState state, final GoldstandardQualityCounters counters,
+            final Blackhole blackhole) throws IOException {
+        final GoldstandardResult result = evaluateCistemStyle(state.corpus, state.allTokens, state.stemmer, blackhole);
+        counters.truePositives += result.truePositives();
+        counters.falsePositives += result.falsePositives();
+        counters.falseNegatives += result.falseNegatives();
+        counters.evaluatedClusters += result.evaluatedClusters();
+        counters.evaluatedTokens += result.evaluatedTokens();
+        return result.evaluatedTokens();
+    }
+
+    /**
+     * Benchmarks candidate throughput over the selected gold standard.
+     *
+     * @param state shared benchmark state
+     * @param blackhole result sink
+     * @throws IOException if token filtering cannot run
+     */
+    @Benchmark
+    public void cistemStyleSpeed(final SharedState state, final Blackhole blackhole) throws IOException {
+        state.stemmer.stem(state.allTokens, blackhole);
+    }
+
+    /**
+     * Named German candidates used for the CISTEM gold-standard comparison.
+     */
+    private enum GermanCandidate {
+        GERMAN_RADIXOR,
+        GERMAN_LUCENE_GERMAN_STEM_FILTER,
+        GERMAN_LUCENE_GERMAN_LIGHT_STEM_FILTER,
+        GERMAN_LUCENE_GERMAN_MINIMAL_STEM_FILTER,
+        GERMAN_CISTEM,
+        SNOWBALL_GERMAN_DIRECT,
+        SNOWBALL_GERMAN_LUCENE_FILTER;
+
+        /**
+         * Creates a candidate evaluator.
+         *
+         * @return stemmer evaluator
+         * @throws IOException if trie resources cannot be loaded
+         */
+        GoldstandardStemmer createEvaluator() throws IOException {
+            return switch (this) {
+                case GERMAN_RADIXOR -> direct(createGermanRadixorStemmer());
+                case GERMAN_LUCENE_GERMAN_STEM_FILTER ->
+                    tokenFilter(input -> new GermanStemFilter(lowercase(input)));
+                case GERMAN_LUCENE_GERMAN_LIGHT_STEM_FILTER ->
+                    tokenFilter(input -> new GermanLightStemFilter(germanNormalize(input)));
+                case GERMAN_LUCENE_GERMAN_MINIMAL_STEM_FILTER ->
+                    tokenFilter(input -> new GermanMinimalStemFilter(germanNormalize(input)));
+                case GERMAN_CISTEM -> direct(Cistem::stem);
+                case SNOWBALL_GERMAN_DIRECT -> direct(SnowballLanguageCase.GERMAN.createDirectStemmer()::stem);
+                case SNOWBALL_GERMAN_LUCENE_FILTER ->
+                    tokenFilter(input -> new SnowballFilter(new LowerCaseFilter(input),
+                            SnowballLanguageCase.GERMAN.luceneSnowballName()));
+            };
+        }
+    }
+
+    /**
+     * Evaluates one full corpus through CISTEM-style cluster scoring.
+     *
+     * <p>
+     * For each cluster, the most frequent predicted stem is considered the
+     * cluster main stem. TP are cluster words mapped to this stem, FN are
+     * words mapped elsewhere inside the same cluster, and FP are words from
+     * other clusters mapped to the same main stem.
+     * </p>
+     *
+     * @param corpus parsed gold standard corpus
+     * @param allTokens flattened token sequence
+     * @param stemmer candidate stemmer
+     * @param blackhole result sink
+     * @return aggregated TP/FP/FN counters and token metrics
+     * @throws IOException when token filtering cannot run
+     */
+    private static GoldstandardResult evaluateCistemStyle(final GermanGoldstandardCorpus corpus,
+            final String[] allTokens, final GoldstandardStemmer stemmer, final Blackhole blackhole) throws IOException {
+        final String[] predicted = stemmer.stem(allTokens, blackhole);
+        final Map<String, Integer> globalPredictions = new LinkedHashMap<>();
+        for (int index = 0; index < allTokens.length; index++) {
+            final String prediction = normalizePrediction(predicted[index], allTokens[index]);
+            globalPredictions.put(prediction, globalPredictions.getOrDefault(prediction, 0) + 1);
+        }
+
+        long truePositives = 0L;
+        long falsePositives = 0L;
+        long falseNegatives = 0L;
+        int tokenOffset = 0;
+        for (final String[] cluster : corpus.clusters()) {
+            if (cluster.length == 0) {
+                continue;
+            }
+
+            final Map<String, Integer> localPredictions = new LinkedHashMap<>();
+            for (int index = 0; index < cluster.length; index++) {
+                final int tokenIndex = tokenOffset + index;
+                final String word = allTokens[tokenIndex];
+                final String prediction = normalizePrediction(predicted[tokenIndex], word);
+                localPredictions.put(prediction, localPredictions.getOrDefault(prediction, 0) + 1);
+            }
+
+            final String mainStem = mostFrequent(localPredictions);
+            final int predictedAsMain = localPredictions.get(mainStem);
+            final int clusterSize = cluster.length;
+            final int falseNegative = clusterSize - predictedAsMain;
+            final int falsePositive = globalPredictions.get(mainStem) - predictedAsMain;
+
+            truePositives += predictedAsMain;
+            falseNegatives += falseNegative;
+            falsePositives += falsePositive;
+            tokenOffset += clusterSize;
+        }
+
+        return new GoldstandardResult(truePositives, falsePositives, falseNegatives, corpus.clusters().length,
+                allTokens.length);
+    }
+
+    /**
+     * Creates a direct evaluator.
+     *
+     * @param stemmer direct word stemmer
+     * @return evaluator
+     */
+    private static GoldstandardStemmer direct(final Stemmer stemmer) {
+        Objects.requireNonNull(stemmer, "stemmer");
+        return (tokens, blackhole) -> {
+            final String[] outputs = new String[tokens.length];
+            for (int index = 0; index < tokens.length; index++) {
+                final String output = stemmer.stem(tokens[index]);
+                outputs[index] = output;
+                blackhole.consume(output);
+            }
+            return outputs;
+        };
+    }
+
+    /**
+     * Creates a TokenFilter evaluator.
+     *
+     * @param factory filter stream factory
+     * @return evaluator
+     */
+    private static GoldstandardStemmer tokenFilter(final Function<TokenStream, TokenStream> factory) {
+        Objects.requireNonNull(factory, "factory");
+        return (tokens, blackhole) -> firstTokenFilterOutputs(tokens, factory, blackhole);
+    }
+
+    /**
+     * Loads and parses one gold standard file from generated JMH resources.
+     *
+     * @param resourceName gold standard file name
+     * @return parsed corpus
+     * @throws IOException if reading fails
+     */
+    private static GermanGoldstandardCorpus loadCorpus(final String resourceName) throws IOException {
+        final ClassLoader classLoader = GermanGoldstandardStemmerComparisonBenchmark.class.getClassLoader();
+        final InputStream resourceStream = classLoader.getResourceAsStream(resourceName);
+        if (resourceStream == null) {
+            throw new IllegalStateException("Missing generated CISTEM gold standard resource: " + resourceName
+                    + ". Run the Gradle JMH resource preparation task to download benchmark-only inputs.");
+        }
+        try (InputStream input = resourceStream) {
+            return parseCorpus(new BufferedReader(new InputStreamReader(input, StandardCharsets.UTF_8)));
+        }
+    }
+
+    /**
+     * Parses CISTEM gold standard format into clustered candidates.
+     *
+     * @param reader UTF-8 reader
+     * @return parsed corpus
+     * @throws IOException if input cannot be read
+     */
+    private static GermanGoldstandardCorpus parseCorpus(final BufferedReader reader) throws IOException {
+        final List<String[]> clusters = new ArrayList<>();
+        String line = reader.readLine();
+        while (line != null) {
+            final String trimmed = line.trim();
+            if (!trimmed.isEmpty()) {
+                final String[] words = trimmed.split("\\s+");
+                if (words.length > 0) {
+                    clusters.add(words);
+                }
+            }
+            line = reader.readLine();
+        }
+        return new GermanGoldstandardCorpus(clusters.toArray(String[][]::new));
+    }
+
+    /**
+     * Flattens the corpus in deterministic cluster order.
+     *
+     * @param corpus corpus to flatten
+     * @return flattened token array
+     */
+    private static String[] flattenCorpusTokens(final GermanGoldstandardCorpus corpus) {
+        int total = 0;
+        for (final String[] cluster : corpus.clusters()) {
+            total += cluster.length;
+        }
+        final String[] tokens = new String[total];
+        int index = 0;
+        for (final String[] cluster : corpus.clusters()) {
+            System.arraycopy(cluster, 0, tokens, index, cluster.length);
+            index += cluster.length;
+        }
+        return tokens;
+    }
+
+    /**
+     * Returns the most frequent key; insertion order is preserved on ties.
+     *
+     * @param frequencies predicted stem frequencies
+     * @return most frequent stem
+     */
+    private static String mostFrequent(final Map<String, Integer> frequencies) {
+        String best = null;
+        int bestCount = -1;
+        for (final Map.Entry<String, Integer> entry : frequencies.entrySet()) {
+            if (entry.getValue() > bestCount) {
+                best = entry.getKey();
+                bestCount = entry.getValue();
+            }
+        }
+        return best;
+    }
+
+    /**
+     * Normalizes a null/empty prediction using the input token as fallback.
+     *
+     * @param prediction stemmed token
+     * @param fallback fallback token
+     * @return safe prediction
+     */
+    private static String normalizePrediction(final String prediction, final String fallback) {
+        if (prediction == null || prediction.isEmpty()) {
+            return fallback;
+        }
+        return prediction;
+    }
+
+    /**
+     * Applies one TokenFilter to all input tokens and returns the first emitted term
+     * for each input token.
+     *
+     * @param tokens input token corpus
+     * @param factory TokenFilter factory
+     * @param blackhole result sink
+     * @return first emitted term per input token
+     * @throws IOException if token streaming fails
+     */
+    private static String[] firstTokenFilterOutputs(final String[] tokens, final Function<TokenStream, TokenStream> factory,
+            final Blackhole blackhole) throws IOException {
+        final String[] outputs = new String[tokens.length];
+        final BenchmarkTokenStream input = new BenchmarkTokenStream(tokens);
+        final TokenStream output = factory.apply(input);
+        final CharTermAttribute termAttribute = output.addAttribute(CharTermAttribute.class);
+        final PositionIncrementAttribute positionAttribute = output.addAttribute(PositionIncrementAttribute.class);
+
+        int inputIndex = -1;
+        boolean recordedForPosition = false;
+        output.reset();
+        while (output.incrementToken()) {
+            final int positionIncrement = positionAttribute.getPositionIncrement();
+            if (positionIncrement > 0) {
+                inputIndex += positionIncrement;
+                recordedForPosition = false;
+            }
+            if (inputIndex >= 0 && inputIndex < outputs.length && !recordedForPosition) {
+                outputs[inputIndex] = termAttribute.toString();
+                blackhole.consume(termAttribute);
+                recordedForPosition = true;
+            }
+        }
+        output.end();
+        output.close();
+
+        for (int index = 0; index < outputs.length; index++) {
+            if (outputs[index] == null) {
+                outputs[index] = tokens[index];
+            }
+        }
+        return outputs;
+    }
+
+    /**
+     * Creates a direct Radixor evaluator using the contracted dictionary trie.
+     *
+     * @return direct Radixor stemmer
+     * @throws IOException if the trie cannot be loaded
+     */
+    private static Stemmer createGermanRadixorStemmer() throws IOException {
+        return new RadixorBenchmarkStemmer(StemmerPatchTrieLoader.loadCompiled(
+                StemmerPatchTrieLoader.Language.DE_DE, true,
+                ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS))::stem;
+    }
+
+    /**
+     * Adds Lucene lower-case normalization.
+     *
+     * @param input token stream
+     * @return normalized token stream
+     */
+    private static TokenStream lowercase(final TokenStream input) {
+        return new LowerCaseFilter(input);
+    }
+
+    /**
+     * Adds Lucene German normalization for light and minimal filters.
+     *
+     * @param input token stream
+     * @return normalized token stream
+     */
+    private static TokenStream germanNormalize(final TokenStream input) {
+        return new GermanNormalizationFilter(lowercase(input));
+    }
+
+    /**
+     * Direct or filter stemmer adapter used by this benchmark.
+     */
+    @FunctionalInterface
+    private interface GoldstandardStemmer {
+
+        /**
+         * Runs one complete token list.
+         *
+         * @param tokens input tokens
+         * @param blackhole result sink
+         * @return per-token outputs
+         * @throws IOException if filter processing fails
+         */
+        String[] stem(String[] tokens, Blackhole blackhole) throws IOException;
+    }
+
+    /**
+     * Deterministic direct word stem function.
+     */
+    @FunctionalInterface
+    private interface Stemmer {
+
+        /**
+         * Stems one token.
+         *
+         * @param token input token
+         * @return stemmed token
+         */
+        String stem(String token);
+    }
+
+    /**
+     * Immutable parsed CISTEM gold standard corpus.
+     */
+    private static final class GermanGoldstandardCorpus {
+
+        private final String[][] clusters;
+
+        GermanGoldstandardCorpus(final String[][] clusters) {
+            this.clusters = clusters;
+        }
+
+        String[][] clusters() {
+            return this.clusters;
+        }
+    }
+
+    /**
+     * Aggregated quality result for one benchmark operation.
+     *
+     * @param truePositives true positives
+     * @param falsePositives false positives
+     * @param falseNegatives false negatives
+     * @param evaluatedClusters evaluated clusters
+     * @param evaluatedTokens evaluated tokens
+     */
+    private record GoldstandardResult(long truePositives, long falsePositives, long falseNegatives,
+            long evaluatedClusters, long evaluatedTokens) {
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/HunspellStemmerComparisonBenchmark.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/HunspellStemmerComparisonBenchmark.java
@@ -0,0 +1,315 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.io.IOException;
+import java.io.InputStream;
+import java.text.ParseException;
+import java.util.List;
+import java.util.Locale;
+import java.util.concurrent.TimeUnit;
+
+import org.apache.lucene.analysis.LowerCaseFilter;
+import org.apache.lucene.analysis.TokenStream;
+import org.apache.lucene.analysis.hunspell.Dictionary;
+import org.apache.lucene.analysis.hunspell.HunspellStemFilter;
+import org.apache.lucene.analysis.hunspell.SortingStrategy;
+import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;
+import org.egothor.stemmer.CompiledPatchCommand;
+import org.egothor.stemmer.FrequencyTrie;
+import org.egothor.stemmer.ReductionMode;
+import org.egothor.stemmer.StemmerPatchTrieLoader;
+import org.openjdk.jmh.annotations.Benchmark;
+import org.openjdk.jmh.annotations.BenchmarkMode;
+import org.openjdk.jmh.annotations.Level;
+import org.openjdk.jmh.annotations.Measurement;
+import org.openjdk.jmh.annotations.Mode;
+import org.openjdk.jmh.annotations.OutputTimeUnit;
+import org.openjdk.jmh.annotations.Param;
+import org.openjdk.jmh.annotations.Scope;
+import org.openjdk.jmh.annotations.Setup;
+import org.openjdk.jmh.annotations.State;
+import org.openjdk.jmh.annotations.Warmup;
+import org.openjdk.jmh.infra.Blackhole;
+
+/**
+ * Compares Radixor with Lucene's Hunspell integration over selected
+ * benchmark-only Hunspell dictionaries.
+ */
+@BenchmarkMode(Mode.AverageTime)
+@OutputTimeUnit(TimeUnit.NANOSECONDS)
+@Warmup(iterations = 3, time = 1, timeUnit = TimeUnit.SECONDS)
+@Measurement(iterations = 5, time = 1, timeUnit = TimeUnit.SECONDS)
+public class HunspellStemmerComparisonBenchmark {
+
+    /**
+     * Parameterized benchmark case.
+     */
+    @State(Scope.Benchmark)
+    public static class SharedState {
+
+        /**
+         * Selected language case.
+         */
+        @Param({ "ENGLISH", "CZECH", "GERMAN", "SPANISH", "FRENCH", "DUTCH", "POLISH", "UKRAINIAN" })
+        public String languageCaseName;
+
+        /**
+         * Selected language case descriptor.
+         */
+        private HunspellLanguageCase languageCase;
+
+        /**
+         * Shared deterministic changed-token corpus.
+         */
+        private String[] tokens;
+
+        /**
+         * Radixor benchmark adapter.
+         */
+        private RadixorBenchmarkStemmer radixorStemmer;
+
+        /**
+         * Initializes the selected language corpus and Radixor stemmer.
+         *
+         * @throws IOException if the Radixor corpus or trie cannot be loaded
+         */
+        @Setup(Level.Trial)
+        public void setUp() throws IOException {
+            this.languageCase = HunspellLanguageCase.valueOf(this.languageCaseName);
+            this.tokens = LanguageBenchmarkCorpus.createTokens(this.languageCase.radixorLanguage());
+            final FrequencyTrie<CompiledPatchCommand> trie = StemmerPatchTrieLoader.loadCompiled(
+                    this.languageCase.radixorLanguage(), true,
+                    ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS);
+            this.radixorStemmer = new RadixorBenchmarkStemmer(trie);
+        }
+    }
+
+    /**
+     * Reusable Hunspell filter state.
+     */
+    @State(Scope.Thread)
+    public static class HunspellState {
+
+        /**
+         * Reusable benchmark input stream.
+         */
+        private BenchmarkTokenStream input;
+
+        /**
+         * Reusable Hunspell filter output stream.
+         */
+        private TokenStream output;
+
+        /**
+         * Output term attribute.
+         */
+        private CharTermAttribute termAttribute;
+
+        /**
+         * Initializes the Hunspell dictionary and filter for the selected language.
+         *
+         * @param sharedState selected language state
+         * @throws IOException    if dictionary resources cannot be read
+         * @throws ParseException if the Hunspell dictionary cannot be parsed
+         */
+        @Setup(Level.Trial)
+        public void setUp(final SharedState sharedState) throws IOException, ParseException {
+            this.input = new BenchmarkTokenStream(new String[0]);
+            final Dictionary dictionary = loadDictionary(sharedState.languageCase);
+            this.output = new HunspellStemFilter(new LowerCaseFilter(this.input), dictionary, true);
+            this.termAttribute = this.output.addAttribute(CharTermAttribute.class);
+        }
+
+        /**
+         * Runs Hunspell over one token corpus.
+         *
+         * @param tokens    token corpus
+         * @param blackhole result sink
+         * @throws IOException if Lucene token streaming fails
+         */
+        private void run(final String[] tokens, final Blackhole blackhole) throws IOException {
+            this.input.setTokens(tokens);
+            this.output.reset();
+            while (this.output.incrementToken()) {
+                blackhole.consume(this.termAttribute.toString());
+            }
+            this.output.end();
+        }
+    }
+
+    /**
+     * Runs Radixor direct lookup and patch application.
+     *
+     * @param sharedState selected language state
+     * @param blackhole   result sink
+     */
+    @Benchmark
+    public void radixor(final SharedState sharedState, final Blackhole blackhole) {
+        final String[] tokens = sharedState.tokens;
+        final RadixorBenchmarkStemmer stemmer = sharedState.radixorStemmer;
+        for (String token : tokens) {
+            blackhole.consume(stemmer.stem(token));
+        }
+    }
+
+    /**
+     * Runs Lucene HunspellStemFilter over the selected language corpus.
+     *
+     * @param sharedState selected language state
+     * @param hunspellState reusable Hunspell state
+     * @param blackhole result sink
+     * @throws IOException if Lucene token streaming fails
+     */
+    @Benchmark
+    public void luceneHunspellStemFilter(final SharedState sharedState, final HunspellState hunspellState,
+            final Blackhole blackhole) throws IOException {
+        hunspellState.run(sharedState.tokens, blackhole);
+    }
+
+    /**
+     * Loads a benchmark-only Hunspell dictionary from generated JMH resources.
+     *
+     * @param languageCase selected language case
+     * @return parsed Hunspell dictionary
+     * @throws IOException    if dictionary resources cannot be read
+     * @throws ParseException if the Hunspell dictionary cannot be parsed
+     */
+    private static Dictionary loadDictionary(final HunspellLanguageCase languageCase) throws IOException,
+            ParseException {
+        final ClassLoader classLoader = HunspellStemmerComparisonBenchmark.class.getClassLoader();
+        final String basePath = "hunspell/" + languageCase.hunspellResourceCode() + "/index.";
+        try (InputStream affixStream = openRequiredResource(classLoader, basePath + "aff");
+                InputStream dictionaryStream = openRequiredResource(classLoader, basePath + "dic")) {
+            return new Dictionary(affixStream, List.of(dictionaryStream), true, SortingStrategy.inMemory());
+        }
+    }
+
+    /**
+     * Opens a classpath resource or fails with a descriptive exception.
+     *
+     * @param classLoader class loader
+     * @param path resource path
+     * @return resource stream
+     */
+    private static InputStream openRequiredResource(final ClassLoader classLoader, final String path) {
+        final InputStream stream = classLoader.getResourceAsStream(path);
+        if (stream == null) {
+            throw new IllegalStateException("Missing benchmark-only Hunspell resource: " + path);
+        }
+        return stream;
+    }
+
+    /**
+     * Benchmark language mapping.
+     */
+    private enum HunspellLanguageCase {
+
+        /**
+         * English Hunspell dictionary over the Radixor English corpus.
+         */
+        ENGLISH("en", StemmerPatchTrieLoader.Language.US_UK),
+
+        /**
+         * Czech Hunspell dictionary over the Radixor Czech corpus.
+         */
+        CZECH("cs", StemmerPatchTrieLoader.Language.CS_CZ),
+
+        /**
+         * German Hunspell dictionary over the Radixor German corpus.
+         */
+        GERMAN("de", StemmerPatchTrieLoader.Language.DE_DE),
+
+        /**
+         * Spanish Hunspell dictionary over the Radixor Spanish corpus.
+         */
+        SPANISH("es", StemmerPatchTrieLoader.Language.ES_ES),
+
+        /**
+         * French Hunspell dictionary over the Radixor French corpus.
+         */
+        FRENCH("fr", StemmerPatchTrieLoader.Language.FR_FR),
+
+        /**
+         * Dutch Hunspell dictionary over the Radixor Dutch corpus.
+         */
+        DUTCH("nl", StemmerPatchTrieLoader.Language.NL_NL),
+
+        /**
+         * Polish Hunspell dictionary over the Radixor Polish corpus.
+         */
+        POLISH("pl", StemmerPatchTrieLoader.Language.PL_PL),
+
+        /**
+         * Ukrainian Hunspell dictionary over the Radixor Ukrainian corpus.
+         */
+        UKRAINIAN("uk", StemmerPatchTrieLoader.Language.UK_UA);
+
+        /**
+         * wooorm/dictionaries resource code.
+         */
+        private final String hunspellResourceCode;
+
+        /**
+         * Matching Radixor language.
+         */
+        private final StemmerPatchTrieLoader.Language radixorLanguage;
+
+        /**
+         * Creates a language mapping.
+         *
+         * @param hunspellResourceCode Hunspell resource code
+         * @param radixorLanguage      Radixor language
+         */
+        HunspellLanguageCase(final String hunspellResourceCode, final StemmerPatchTrieLoader.Language radixorLanguage) {
+            this.hunspellResourceCode = hunspellResourceCode.toLowerCase(Locale.ROOT);
+            this.radixorLanguage = radixorLanguage;
+        }
+
+        /**
+         * Returns the Hunspell resource code.
+         *
+         * @return resource code
+         */
+        String hunspellResourceCode() {
+            return this.hunspellResourceCode;
+        }
+
+        /**
+         * Returns the matching Radixor language.
+         *
+         * @return Radixor language
+         */
+        StemmerPatchTrieLoader.Language radixorLanguage() {
+            return this.radixorLanguage;
+        }
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/HunspellStemmerComparisonBenchmarkQuality.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/HunspellStemmerComparisonBenchmarkQuality.java
@@ -0,0 +1,436 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.io.IOException;
+import java.io.InputStream;
+import java.text.ParseException;
+import java.util.List;
+import java.util.Locale;
+import java.util.Objects;
+import java.util.concurrent.TimeUnit;
+
+import org.apache.lucene.analysis.LowerCaseFilter;
+import org.apache.lucene.analysis.TokenStream;
+import org.apache.lucene.analysis.hunspell.Dictionary;
+import org.apache.lucene.analysis.hunspell.HunspellStemFilter;
+import org.apache.lucene.analysis.hunspell.SortingStrategy;
+import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;
+import org.apache.lucene.analysis.tokenattributes.PositionIncrementAttribute;
+import org.egothor.stemmer.StemmerPatchTrieLoader;
+import org.openjdk.jmh.annotations.AuxCounters;
+import org.openjdk.jmh.annotations.Benchmark;
+import org.openjdk.jmh.annotations.BenchmarkMode;
+import org.openjdk.jmh.annotations.Fork;
+import org.openjdk.jmh.annotations.Level;
+import org.openjdk.jmh.annotations.Measurement;
+import org.openjdk.jmh.annotations.Mode;
+import org.openjdk.jmh.annotations.OutputTimeUnit;
+import org.openjdk.jmh.annotations.Param;
+import org.openjdk.jmh.annotations.Scope;
+import org.openjdk.jmh.annotations.Setup;
+import org.openjdk.jmh.annotations.State;
+import org.openjdk.jmh.annotations.Warmup;
+import org.openjdk.jmh.infra.Blackhole;
+
+/**
+ * Emits exact-root agreement metrics for the benchmark-only Hunspell comparisons.
+ *
+ * <p>
+ * This class mirrors the existing Hunspell throughput setup but adds
+ * quality-style accuracy counters for every Hunspell language dictionary used
+ * in benchmark-only throughput comparisons.
+ * </p>
+ */
+@BenchmarkMode(Mode.AverageTime)
+@OutputTimeUnit(TimeUnit.NANOSECONDS)
+@Warmup(iterations = 0)
+@Measurement(iterations = 1, time = 1, timeUnit = TimeUnit.MILLISECONDS)
+@Fork(0)
+public class HunspellStemmerComparisonBenchmarkQuality {
+
+    /**
+     * Shared quality corpus and Hunspell dictionary for a selected language.
+     */
+    @State(Scope.Benchmark)
+    public static class SharedState {
+
+        /**
+         * Selected language case.
+         */
+        @Param({ "ENGLISH", "CZECH", "GERMAN", "SPANISH", "FRENCH", "DUTCH", "POLISH", "UKRAINIAN" })
+        public String languageCaseName;
+
+        /**
+         * Selected language descriptor.
+         */
+        private HunspellLanguageCase languageCase;
+
+        /**
+         * Complete language dictionary corpus and expected roots.
+         */
+        private LanguageBenchmarkCorpus.Corpus corpus;
+
+        /**
+         * Parsed benchmark-only Hunspell dictionary.
+         */
+        private Dictionary dictionary;
+
+        /**
+         * Initializes quality resources.
+         *
+         * @throws IOException    if corpus or dictionary loading fails
+         * @throws ParseException if the Hunspell dictionary cannot be parsed
+         */
+        @Setup(Level.Trial)
+        public void setUp() throws IOException, ParseException {
+            this.languageCase = HunspellLanguageCase.valueOf(this.languageCaseName);
+            this.corpus = LanguageBenchmarkCorpus.createFullCorpus(this.languageCase.radixorLanguage());
+            this.dictionary = loadDictionary(this.languageCase);
+        }
+    }
+
+    /**
+     * JMH auxiliary counters for exact-root agreement.
+     */
+    @State(Scope.Thread)
+    @AuxCounters(AuxCounters.Type.EVENTS)
+    public static class AccuracyCounters {
+
+        /**
+         * Number of exact-root matches.
+         */
+        public long correctMatches;
+
+        /**
+         * Number of evaluated tokens.
+         */
+        public long evaluatedTokens;
+
+        /**
+         * Number of exact-root matches where the input token differs from the
+         * expected root.
+         */
+        public long changedCorrectMatches;
+
+        /**
+         * Number of evaluated tokens where the input token differs from the expected
+         * root.
+         */
+        public long changedEvaluatedTokens;
+
+        /**
+         * Number of exact-root matches where the input token is already the expected
+         * root.
+         */
+        public long rootPreservedMatches;
+
+        /**
+         * Number of evaluated tokens where the input token is already the expected
+         * root.
+         */
+        public long rootEvaluatedTokens;
+
+        /**
+         * Resets counters before each measured iteration.
+         */
+        @Setup(Level.Iteration)
+        public void reset() {
+            this.correctMatches = 0L;
+            this.evaluatedTokens = 0L;
+            this.changedCorrectMatches = 0L;
+            this.changedEvaluatedTokens = 0L;
+            this.rootPreservedMatches = 0L;
+            this.rootEvaluatedTokens = 0L;
+        }
+    }
+
+    /**
+     * Evaluates exact-root agreement for the selected Hunspell dictionary.
+     *
+     * @param sharedState shared quality state
+     * @param counters    JMH auxiliary counters
+     * @param blackhole   result sink
+     * @return exact-root match count
+     * @throws IOException if Lucene token streaming fails
+     */
+    @Benchmark
+    public int luceneHunspellStemFilterAccuracy(final SharedState sharedState, final AccuracyCounters counters,
+            final Blackhole blackhole) throws IOException {
+        final String[] actualStems = firstHunspellOutputs(sharedState.corpus.tokens(), sharedState.dictionary,
+                blackhole);
+        final String[] tokens = sharedState.corpus.tokens();
+        final String[] expectedRoots = sharedState.corpus.expectedRoots();
+
+        int correct = 0;
+        int changedCorrect = 0;
+        int changedEvaluated = 0;
+        int rootPreserved = 0;
+        int rootEvaluated = 0;
+        for (int index = 0; index < actualStems.length; index++) {
+            final String token = tokens[index];
+            final String expectedRoot = expectedRoots[index];
+            final boolean exact = Objects.equals(expectedRoot, actualStems[index]);
+            if (exact) {
+                correct++;
+            }
+            if (Objects.equals(token, expectedRoot)) {
+                rootEvaluated++;
+                if (exact) {
+                    rootPreserved++;
+                }
+            } else {
+                changedEvaluated++;
+                if (exact) {
+                    changedCorrect++;
+                }
+            }
+        }
+
+        counters.correctMatches += correct;
+        counters.evaluatedTokens += actualStems.length;
+        counters.changedCorrectMatches += changedCorrect;
+        counters.changedEvaluatedTokens += changedEvaluated;
+        counters.rootPreservedMatches += rootPreserved;
+        counters.rootEvaluatedTokens += rootEvaluated;
+        return correct;
+    }
+
+    /**
+     * Extracts the first emitted Hunspell stem for each input token.
+     *
+     * @param tokens     token corpus
+     * @param dictionary Hunspell dictionary
+     * @param blackhole  result sink
+     * @return first emitted term per input token
+     * @throws IOException if Lucene streaming fails
+     */
+    private static String[] firstHunspellOutputs(final String[] tokens, final Dictionary dictionary,
+            final Blackhole blackhole) throws IOException {
+        final String[] outputs = new String[tokens.length];
+        final BenchmarkTokenStream input = new BenchmarkTokenStream(tokens);
+        final TokenStream output = new HunspellStemFilter(new LowerCaseFilter(input), dictionary, true);
+        final CharTermAttribute termAttribute = output.addAttribute(CharTermAttribute.class);
+        final PositionIncrementAttribute positionAttribute = output.addAttribute(PositionIncrementAttribute.class);
+        int inputIndex = -1;
+        boolean recordedForPosition = false;
+
+        output.reset();
+        while (output.incrementToken()) {
+            final int positionIncrement = positionAttribute.getPositionIncrement();
+            if (positionIncrement > 0) {
+                inputIndex += positionIncrement;
+                recordedForPosition = false;
+            }
+            if (inputIndex >= 0 && inputIndex < outputs.length && !recordedForPosition) {
+                outputs[inputIndex] = termAttribute.toString();
+                recordedForPosition = true;
+            }
+            if (blackhole != null) {
+                blackhole.consume(termAttribute);
+            }
+        }
+        output.end();
+        output.close();
+
+        for (int index = 0; index < outputs.length; index++) {
+            if (outputs[index] == null) {
+                outputs[index] = tokens[index];
+            }
+        }
+        return outputs;
+    }
+
+    /**
+     * Loads a benchmark-only Hunspell dictionary from generated resources.
+     *
+     * @param languageCase selected language case
+     * @return parsed dictionary
+     * @throws IOException    if dictionary resources cannot be read
+     * @throws ParseException if dictionary parsing fails
+     */
+    private static Dictionary loadDictionary(final HunspellLanguageCase languageCase) throws IOException,
+            ParseException {
+        final ClassLoader classLoader = HunspellStemmerComparisonBenchmarkQuality.class.getClassLoader();
+        final String basePath = "hunspell/" + languageCase.hunspellResourceCode() + "/index.";
+        try (InputStream affixStream = openRequiredResource(classLoader, basePath + "aff");
+                InputStream dictionaryStream = openRequiredResource(classLoader, basePath + "dic")) {
+            return new Dictionary(affixStream, List.of(dictionaryStream), true, SortingStrategy.inMemory());
+        }
+    }
+
+    /**
+     * Stems one analytical batch through the exact Hunspell quality-benchmark path.
+     *
+     * @param languageCase declared Hunspell language case
+     * @param tokens original dictionary forms
+     * @return first Hunspell output per input form
+     * @throws IOException if dictionary parsing or token streaming fails
+     */
+    static String[] stemForQuality(final HunspellLanguageCase languageCase, final String[] tokens) throws IOException {
+        try {
+            return firstHunspellOutputs(tokens, loadDictionary(languageCase), null);
+        } catch (ParseException exception) {
+            throw new IOException("Unable to parse the JMH Hunspell dictionary for " + languageCase + ".", exception);
+        }
+    }
+
+    /** Returns all distinct Hunspell stems per token through the quality-benchmark dictionary. */
+    static List<List<String>> stemCandidatesForQuality(final HunspellLanguageCase languageCase,
+            final String[] tokens) throws IOException {
+        try {
+            final Dictionary dictionary = loadDictionary(languageCase);
+            final List<java.util.LinkedHashSet<String>> candidates = new java.util.ArrayList<>(tokens.length);
+            for (int index = 0; index < tokens.length; index++) { candidates.add(new java.util.LinkedHashSet<>()); }
+            final BenchmarkTokenStream input = new BenchmarkTokenStream(tokens);
+            final TokenStream output = new HunspellStemFilter(new LowerCaseFilter(input), dictionary, true);
+            final CharTermAttribute term = output.addAttribute(CharTermAttribute.class);
+            final PositionIncrementAttribute position = output.addAttribute(PositionIncrementAttribute.class);
+            int inputIndex = -1;
+            output.reset();
+            while (output.incrementToken()) {
+                if (position.getPositionIncrement() > 0) { inputIndex += position.getPositionIncrement(); }
+                if (inputIndex >= 0 && inputIndex < candidates.size()) { candidates.get(inputIndex).add(term.toString()); }
+            }
+            output.end();
+            output.close();
+            final String[] primary = firstHunspellOutputs(tokens, dictionary, null);
+            final List<List<String>> result = new java.util.ArrayList<>(tokens.length);
+            for (int index = 0; index < tokens.length; index++) {
+                candidates.get(index).add(primary[index]);
+                result.add(List.copyOf(candidates.get(index)));
+            }
+            return List.copyOf(result);
+        } catch (ParseException exception) {
+            throw new IOException("Unable to parse the JMH Hunspell dictionary for " + languageCase + ".", exception);
+        }
+    }
+
+    /**
+     * Opens a required classpath resource.
+     *
+     * @param classLoader class loader
+     * @param path        resource path
+     * @return resource stream
+     */
+    private static InputStream openRequiredResource(final ClassLoader classLoader, final String path) {
+        final InputStream stream = classLoader.getResourceAsStream(path);
+        if (stream == null) {
+            throw new IllegalStateException("Missing benchmark-only Hunspell resource: " + path);
+        }
+        return stream;
+    }
+
+    /**
+     * Benchmark language mapping.
+     */
+    enum HunspellLanguageCase {
+
+        /**
+         * English Hunspell dictionary over the Radixor English corpus.
+         */
+        ENGLISH("en", StemmerPatchTrieLoader.Language.US_UK),
+
+        /**
+         * Czech Hunspell dictionary over the Radixor Czech corpus.
+         */
+        CZECH("cs", StemmerPatchTrieLoader.Language.CS_CZ),
+
+        /**
+         * German Hunspell dictionary over the Radixor German corpus.
+         */
+        GERMAN("de", StemmerPatchTrieLoader.Language.DE_DE),
+
+        /**
+         * Spanish Hunspell dictionary over the Radixor Spanish corpus.
+         */
+        SPANISH("es", StemmerPatchTrieLoader.Language.ES_ES),
+
+        /**
+         * French Hunspell dictionary over the Radixor French corpus.
+         */
+        FRENCH("fr", StemmerPatchTrieLoader.Language.FR_FR),
+
+        /**
+         * Dutch Hunspell dictionary over the Radixor Dutch corpus.
+         */
+        DUTCH("nl", StemmerPatchTrieLoader.Language.NL_NL),
+
+        /**
+         * Polish Hunspell dictionary over the Radixor Polish corpus.
+         */
+        POLISH("pl", StemmerPatchTrieLoader.Language.PL_PL),
+
+        /**
+         * Ukrainian Hunspell dictionary over the Radixor Ukrainian corpus.
+         */
+        UKRAINIAN("uk", StemmerPatchTrieLoader.Language.UK_UA);
+
+        /**
+         * Wooorm/dictionaries resource code.
+         */
+        private final String hunspellResourceCode;
+
+        /**
+         * Matching Radixor language.
+         */
+        private final StemmerPatchTrieLoader.Language radixorLanguage;
+
+        /**
+         * Creates a language mapping.
+         *
+         * @param hunspellResourceCode Hunspell resource code
+         * @param radixorLanguage      Radixor language
+         */
+        HunspellLanguageCase(final String hunspellResourceCode, final StemmerPatchTrieLoader.Language radixorLanguage) {
+            this.hunspellResourceCode = hunspellResourceCode.toLowerCase(Locale.ROOT);
+            this.radixorLanguage = radixorLanguage;
+        }
+
+        /**
+         * Returns the Hunspell resource code.
+         *
+         * @return resource code
+         */
+        String hunspellResourceCode() {
+            return this.hunspellResourceCode;
+        }
+
+        /**
+         * Returns the matching Radixor language.
+         *
+         * @return Radixor language
+         */
+        StemmerPatchTrieLoader.Language radixorLanguage() {
+            return this.radixorLanguage;
+        }
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/LanguageBenchmarkCorpus.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/LanguageBenchmarkCorpus.java
@@ -0,0 +1,445 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.io.BufferedReader;
+import java.io.IOException;
+import java.io.InputStream;
+import java.io.InputStreamReader;
+import java.nio.charset.StandardCharsets;
+import java.util.ArrayList;
+import java.util.EnumMap;
+import java.util.List;
+import java.util.Map;
+import java.util.Locale;
+import java.util.Objects;
+import java.util.zip.GZIPInputStream;
+
+import org.egothor.stemmer.StemmerPatchTrieLoader;
+
+/**
+ * Builds deterministic language-specific benchmark corpora from bundled
+ * Radixor dictionary resources.
+ *
+ * <p>
+ * Corpus construction is setup work only. It is intentionally based on the same
+ * resource that backs the Radixor benchmark path so every competitor for a
+ * language consumes the same changed-token timing workload, while quality
+ * benchmarks can still use the complete dictionary workload.
+ * </p>
+ */
+final class LanguageBenchmarkCorpus {
+
+    /**
+     * Minimum token count for timing benchmark operations.
+     */
+    static final int MINIMUM_TIMING_TOKEN_COUNT = 5_000;
+
+    /**
+     * Shared timing corpora keyed by bundled Radixor language.
+     */
+    private static final Map<StemmerPatchTrieLoader.Language, Corpus> TIMING_CORPORA =
+            new EnumMap<>(StemmerPatchTrieLoader.Language.class);
+
+    /**
+     * Shared changed-token timing corpora keyed by bundled Radixor language.
+     */
+    private static final Map<StemmerPatchTrieLoader.Language, Corpus> CHANGED_TIMING_CORPORA =
+            new EnumMap<>(StemmerPatchTrieLoader.Language.class);
+
+    /**
+     * Shared complete corpora keyed by bundled Radixor language.
+     */
+    private static final Map<StemmerPatchTrieLoader.Language, Corpus> FULL_CORPORA =
+            new EnumMap<>(StemmerPatchTrieLoader.Language.class);
+
+    /**
+     * Utility class.
+     */
+    private LanguageBenchmarkCorpus() {
+        throw new AssertionError("No instances.");
+    }
+
+    /**
+     * Creates a deterministic changed-token timing corpus from a bundled language
+     * dictionary.
+     *
+     * <p>
+     * Only token/root pairs where the token differs from the expected root are
+     * included. Smaller changed-token resources are repeated in stable order until
+     * the timing corpus reaches 5,000 tokens.
+     * </p>
+     *
+     * @param language bundled Radixor language
+     * @return token array containing changed-token dictionary entries, repeated
+     *         only when the changed-token resource is smaller than 5,000 tokens
+     * @throws IOException if the resource cannot be read
+     */
+    static String[] createTokens(final StemmerPatchTrieLoader.Language language) throws IOException {
+        return createChangedCorpus(language).tokens();
+    }
+
+    /**
+     * Creates a deterministic changed-token timing corpus from a bundled language
+     * dictionary.
+     *
+     * @param language bundled Radixor language
+     * @return changed-token corpus with expected roots
+     * @throws IOException if the resource cannot be read
+     */
+    static Corpus createChangedCorpus(final StemmerPatchTrieLoader.Language language) throws IOException {
+        return cachedChangedCorpus(language);
+    }
+
+    /**
+     * Creates a deterministic full-dictionary timing corpus and expected root
+     * array from a bundled language dictionary.
+     *
+     * @param language bundled Radixor language
+     * @return token corpus with expected roots
+     * @throws IOException if the resource cannot be read
+     */
+    static Corpus createCorpus(final StemmerPatchTrieLoader.Language language) throws IOException {
+        return cachedCorpus(TIMING_CORPORA, language, true);
+    }
+
+    /**
+     * Creates a deterministic full-dictionary timing corpus and expected root
+     * array from a bundled language dictionary.
+     *
+     * <p>
+     * The complete dictionary token sequence is used when it contains at least
+     * {@code minimumTokenCount} tokens. Smaller resources are repeated in stable
+     * order until the minimum is reached.
+     * </p>
+     *
+     * @param language          bundled Radixor language
+     * @param minimumTokenCount minimum token count for timing
+     * @return token corpus with expected roots
+     * @throws IOException if the resource cannot be read
+     */
+    static Corpus createCorpus(final StemmerPatchTrieLoader.Language language, final int minimumTokenCount)
+            throws IOException {
+        Objects.requireNonNull(language, "language");
+        if (minimumTokenCount < 1) {
+            throw new IllegalArgumentException("minimumTokenCount must be at least 1.");
+        }
+        if (minimumTokenCount == MINIMUM_TIMING_TOKEN_COUNT) {
+            return createCorpus(language);
+        }
+
+        return buildTimingCorpus(language, minimumTokenCount);
+    }
+
+    /**
+     * Creates or returns the shared complete corpus for a bundled language.
+     *
+     * @param language bundled Radixor language
+     * @return complete token corpus with expected roots
+     * @throws IOException if the resource cannot be read
+     */
+    static Corpus createFullCorpus(final StemmerPatchTrieLoader.Language language) throws IOException {
+        return cachedCorpus(FULL_CORPORA, language, false);
+    }
+
+    /**
+     * Returns a cached corpus, creating it once per JVM when necessary.
+     *
+     * @param cache corpus cache
+     * @param language bundled Radixor language
+     * @param timing whether the timing-minimum corpus should be built
+     * @return cached corpus instance
+     * @throws IOException if the resource cannot be read
+     */
+    private static Corpus cachedCorpus(final Map<StemmerPatchTrieLoader.Language, Corpus> cache,
+            final StemmerPatchTrieLoader.Language language, final boolean timing) throws IOException {
+        Objects.requireNonNull(cache, "cache");
+        Objects.requireNonNull(language, "language");
+
+        synchronized (LanguageBenchmarkCorpus.class) {
+            final Corpus existing = cache.get(language);
+            if (existing != null) {
+                return existing;
+            }
+
+            final Corpus created = timing ? buildTimingCorpus(language, MINIMUM_TIMING_TOKEN_COUNT)
+                    : buildFullCorpus(language);
+            cache.put(language, created);
+            return created;
+        }
+    }
+
+    /**
+     * Returns a cached changed-token timing corpus, creating it once per JVM when
+     * necessary.
+     *
+     * @param language bundled Radixor language
+     * @return changed-token timing corpus
+     * @throws IOException if the resource cannot be read
+     */
+    private static Corpus cachedChangedCorpus(final StemmerPatchTrieLoader.Language language) throws IOException {
+        Objects.requireNonNull(language, "language");
+
+        synchronized (LanguageBenchmarkCorpus.class) {
+            final Corpus existing = CHANGED_TIMING_CORPORA.get(language);
+            if (existing != null) {
+                return existing;
+            }
+
+            final Corpus created = buildChangedTimingCorpus(language, MINIMUM_TIMING_TOKEN_COUNT);
+            CHANGED_TIMING_CORPORA.put(language, created);
+            return created;
+        }
+    }
+
+    /**
+     * Builds a deterministic timing corpus from a bundled language dictionary.
+     *
+     * @param language bundled Radixor language
+     * @param minimumTokenCount minimum token count for timing
+     * @return token corpus with expected roots
+     * @throws IOException if the resource cannot be read
+     */
+    private static Corpus buildTimingCorpus(final StemmerPatchTrieLoader.Language language, final int minimumTokenCount)
+            throws IOException {
+        Objects.requireNonNull(language, "language");
+        if (minimumTokenCount < 1) {
+            throw new IllegalArgumentException("minimumTokenCount must be at least 1.");
+        }
+
+        final List<Entry> candidates = readCandidates(language, Integer.MAX_VALUE);
+        if (candidates.isEmpty()) {
+            throw new IllegalStateException("No benchmark corpus tokens were available for " + language + ".");
+        }
+
+        final int timingTokenCount = Math.max(candidates.size(), minimumTokenCount);
+        final String[] tokens = new String[timingTokenCount];
+        final String[] expectedRoots = new String[timingTokenCount];
+        for (int index = 0; index < tokens.length; index++) {
+            final Entry entry = candidates.get(index % candidates.size());
+            tokens[index] = entry.token();
+            expectedRoots[index] = entry.root();
+        }
+        return new Corpus(tokens, expectedRoots);
+    }
+
+    /**
+     * Builds a deterministic changed-token timing corpus from a bundled language
+     * dictionary.
+     *
+     * @param language bundled Radixor language
+     * @param minimumTokenCount minimum token count for timing
+     * @return changed-token corpus with expected roots
+     * @throws IOException if the resource cannot be read
+     */
+    private static Corpus buildChangedTimingCorpus(final StemmerPatchTrieLoader.Language language,
+            final int minimumTokenCount) throws IOException {
+        Objects.requireNonNull(language, "language");
+        if (minimumTokenCount < 1) {
+            throw new IllegalArgumentException("minimumTokenCount must be at least 1.");
+        }
+
+        final List<Entry> allCandidates = readCandidates(language, Integer.MAX_VALUE);
+        final List<Entry> changedCandidates = new ArrayList<>(allCandidates.size());
+        for (Entry entry : allCandidates) {
+            if (!Objects.equals(entry.token(), entry.root())) {
+                changedCandidates.add(entry);
+            }
+        }
+        if (changedCandidates.isEmpty()) {
+            throw new IllegalStateException("No changed-token benchmark corpus tokens were available for "
+                    + language + ".");
+        }
+
+        final int timingTokenCount = Math.max(changedCandidates.size(), minimumTokenCount);
+        final String[] tokens = new String[timingTokenCount];
+        final String[] expectedRoots = new String[timingTokenCount];
+        for (int index = 0; index < tokens.length; index++) {
+            final Entry entry = changedCandidates.get(index % changedCandidates.size());
+            tokens[index] = entry.token();
+            expectedRoots[index] = entry.root();
+        }
+        return new Corpus(tokens, expectedRoots);
+    }
+
+    /**
+     * Creates a complete deterministic token corpus and expected root array from a
+     * bundled language dictionary.
+     *
+     * <p>
+     * This method is intended for exact-root quality accounting. It includes all
+     * single-token fields available in the dictionary resource and does not repeat
+     * small dictionaries to the timing minimum.
+     * </p>
+     *
+     * @param language bundled Radixor language
+     * @return complete token corpus with expected roots
+     * @throws IOException if the resource cannot be read
+     */
+    private static Corpus buildFullCorpus(final StemmerPatchTrieLoader.Language language) throws IOException {
+        Objects.requireNonNull(language, "language");
+
+        final List<Entry> candidates = readCandidates(language, Integer.MAX_VALUE);
+        if (candidates.isEmpty()) {
+            throw new IllegalStateException("No benchmark corpus tokens were available for " + language + ".");
+        }
+
+        final String[] tokens = new String[candidates.size()];
+        final String[] expectedRoots = new String[candidates.size()];
+        for (int index = 0; index < tokens.length; index++) {
+            final Entry entry = candidates.get(index);
+            tokens[index] = entry.token();
+            expectedRoots[index] = entry.root();
+        }
+        return new Corpus(tokens, expectedRoots);
+    }
+
+    /**
+     * Reads token candidates from a bundled compressed dictionary.
+     *
+     * @param language bundled Radixor language
+     * @param maximumTokenCount maximum token count to read
+     * @return deterministic candidate list
+     * @throws IOException if the resource cannot be read
+     */
+    private static List<Entry> readCandidates(final StemmerPatchTrieLoader.Language language, final int maximumTokenCount)
+            throws IOException {
+        final String resourcePath = language.resourcePath();
+        final InputStream resource = StemmerPatchTrieLoader.class.getClassLoader().getResourceAsStream(resourcePath);
+        if (resource == null) {
+            throw new IllegalStateException("Missing bundled benchmark resource " + resourcePath + ".");
+        }
+
+        final List<Entry> candidates = new ArrayList<>(MINIMUM_TIMING_TOKEN_COUNT);
+        try (InputStream inputStream = resource;
+                GZIPInputStream gzipInputStream = new GZIPInputStream(inputStream);
+                InputStreamReader inputStreamReader = new InputStreamReader(gzipInputStream, StandardCharsets.UTF_8);
+                BufferedReader reader = new BufferedReader(inputStreamReader)) {
+            String line = reader.readLine();
+            while (line != null && candidates.size() < maximumTokenCount) {
+                collectLineCandidates(line, candidates, maximumTokenCount);
+                line = reader.readLine();
+            }
+        }
+        return candidates;
+    }
+
+    /**
+     * Collects lower-case token candidates from one dictionary line.
+     *
+     * @param line       dictionary line
+     * @param candidates mutable candidate list
+     * @param maximumTokenCount maximum token count to read
+     */
+    private static void collectLineCandidates(final String line, final List<Entry> candidates,
+            final int maximumTokenCount) {
+        if (line == null || line.isBlank() || line.startsWith("#") || line.startsWith("//")) {
+            return;
+        }
+
+        final String[] fields = line.split("\t");
+        if (fields.length == 0) {
+            return;
+        }
+
+        final String root = normalizeToken(fields[0]);
+        if (root.isEmpty() || containsWhitespace(root)) {
+            return;
+        }
+
+        for (String field : fields) {
+            if (candidates.size() >= maximumTokenCount) {
+                return;
+            }
+            final String token = normalizeToken(field);
+            if (!token.isEmpty() && !containsWhitespace(token)) {
+                candidates.add(new Entry(token, root));
+            }
+        }
+    }
+
+    /**
+     * Normalizes dictionary token text for deterministic benchmark lookup.
+     *
+     * @param token dictionary token field
+     * @return normalized token
+     */
+    private static String normalizeToken(final String token) {
+        return token.trim().toLowerCase(Locale.ROOT);
+    }
+
+    /**
+     * Returns whether a token contains Unicode whitespace.
+     *
+     * @param token token candidate
+     * @return {@code true} when whitespace is present
+     */
+    private static boolean containsWhitespace(final String token) {
+        for (int index = 0; index < token.length(); index++) {
+            if (Character.isWhitespace(token.charAt(index))) {
+                return true;
+            }
+        }
+        return false;
+    }
+
+    /**
+     * Immutable token corpus with expected roots.
+     *
+     * @param tokens        benchmark token corpus
+     * @param expectedRoots expected root for each token
+     */
+    record Corpus(String[] tokens, String[] expectedRoots) {
+
+        /**
+         * Creates corpus data.
+         *
+         * @param tokens        benchmark token corpus
+         * @param expectedRoots expected root for each token
+         */
+        Corpus {
+            Objects.requireNonNull(tokens, "tokens");
+            Objects.requireNonNull(expectedRoots, "expectedRoots");
+            if (tokens.length != expectedRoots.length) {
+                throw new IllegalArgumentException("tokens and expectedRoots must have the same length.");
+            }
+        }
+    }
+
+    /**
+     * Immutable dictionary-derived token/root entry.
+     *
+     * @param token token form
+     * @param root  expected root
+     */
+    private record Entry(String token, String root) {
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/MultiLanguageStemmerComparisonBenchmark.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/MultiLanguageStemmerComparisonBenchmark.java
--- a/src/jmh/java/org/egothor/stemmer/benchmark/QualityStemmerMatrix.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/QualityStemmerMatrix.java
@@ -0,0 +1,133 @@
+package org.egothor.stemmer.benchmark;
+
+import java.io.IOException;
+import java.util.Arrays;
+import java.util.List;
+import java.util.Objects;
+import java.util.ArrayList;
+import java.util.EnumSet;
+
+import org.egothor.stemmer.StemmerPatchTrieLoader.Language;
+
+/** Authoritative analytical view of the candidate matrix defined by the JMH quality benchmark. */
+public final class QualityStemmerMatrix {
+    /** Utility class. */
+    private QualityStemmerMatrix() {
+        throw new AssertionError("No instances.");
+    }
+
+    /**
+     * Returns every currently registered JMH quality candidate in declaration order.
+     * The returned list is immutable and is derived directly from the benchmark enum.
+     *
+     * @return complete immutable candidate list
+     */
+    public static List<Candidate> candidates() {
+        final List<Candidate> candidates = new ArrayList<>();
+        Arrays.stream(StemmerComparisonBenchmarkQuality.QualityCandidate.values())
+                .map(candidate -> new Candidate(candidate.name(), candidate.radixorLanguage(),
+                        () -> adapt(candidate.createStemmer())))
+                .forEach(candidates::add);
+        final EnumSet<Language> registeredRadixorLanguages = candidates.stream()
+                .filter(candidate -> candidate.name().endsWith("_RADIXOR"))
+                .map(Candidate::language).collect(() -> EnumSet.noneOf(Language.class), EnumSet::add, EnumSet::addAll);
+        Arrays.stream(Language.values()).filter(language -> !registeredRadixorLanguages.contains(language))
+                .map(language -> new Candidate(language.name() + "_RADIXOR", language,
+                        () -> adapt(StemmerComparisonBenchmarkQuality.createRadixorQualityStemmer(language))))
+                .forEach(candidates::add);
+        Arrays.stream(HunspellStemmerComparisonBenchmarkQuality.HunspellLanguageCase.values())
+                .map(languageCase -> new Candidate("HUNSPELL_" + languageCase.name() + "_LUCENE_FILTER",
+                        languageCase.radixorLanguage(),
+                        () -> new BatchStemmer() {
+                            /** {@inheritDoc} */
+                            @Override public String[] stem(final String[] forms) throws IOException {
+                                return HunspellStemmerComparisonBenchmarkQuality.stemForQuality(languageCase, forms);
+                            }
+                            /** {@inheritDoc} */
+                            @Override public List<List<String>> stemCandidates(final String[] forms) throws IOException {
+                                return HunspellStemmerComparisonBenchmarkQuality.stemCandidatesForQuality(languageCase, forms);
+                            }
+                            /** {@inheritDoc} */
+                            @Override public boolean supportsMultipleOutputs() { return true; }
+                        }))
+                .forEach(candidates::add);
+        return List.copyOf(candidates);
+    }
+
+    /** Adapts one authoritative general-matrix stemmer without changing capability semantics. */
+    private static BatchStemmer adapt(final StemmerComparisonBenchmarkQuality.CandidateStemmer stemmer) {
+        return new BatchStemmer() {
+            /** {@inheritDoc} */
+            @Override public String[] stem(final String[] forms) throws IOException { return stemmer.stem(forms); }
+            /** {@inheritDoc} */
+            @Override public List<List<String>> stemCandidates(final String[] forms) throws IOException {
+                return stemmer.stemCandidates(forms);
+            }
+            /** {@inheritDoc} */
+            @Override public boolean supportsMultipleOutputs() { return stemmer.supportsMultipleOutputs(); }
+        };
+    }
+
+    /** One JMH candidate and its authoritative dictionary-language mapping. */
+    public static final class Candidate {
+        private final String name;
+        private final Language language;
+        private final StemmerFactory factory;
+
+        /** Creates an immutable facade over one benchmark candidate. */
+        private Candidate(final String name, final Language language,
+                final StemmerFactory factory) {
+            this.name = Objects.requireNonNull(name, "name");
+            this.language = Objects.requireNonNull(language, "language");
+            this.factory = Objects.requireNonNull(factory, "factory");
+        }
+
+        /** @return stable JMH candidate name */
+        public String name() {
+            return this.name;
+        }
+
+        /** @return registered Radixor gold-standard dictionary language */
+        public Language language() {
+            return this.language;
+        }
+
+        /**
+         * Creates a scenario-confined adapter using exactly the JMH factory and preprocessing path.
+         *
+         * @return sequential batch stemmer
+         * @throws IOException if benchmark-only resources cannot be loaded
+         */
+        public BatchStemmer createStemmer() throws IOException {
+            return this.factory.create();
+        }
+    }
+
+    /** Internal checked factory shared by the JMH quality registries. */
+    @FunctionalInterface
+    private interface StemmerFactory {
+        /** @return a scenario-confined adapter @throws IOException if resources fail */
+        BatchStemmer create() throws IOException;
+    }
+
+    /** Sequential, scenario-confined batch stemmer contract. */
+    @FunctionalInterface
+    public interface BatchStemmer {
+        /**
+         * Stems all supplied forms in order.
+         *
+         * @param forms input forms, never {@code null}
+         * @return one non-null output per form
+         * @throws IOException when the JMH adapter fails
+         */
+        String[] stem(String[] forms) throws IOException;
+
+        /** Returns complete candidate sets; single-output adapters return singleton sets. */
+        default List<List<String>> stemCandidates(final String[] forms) throws IOException {
+            return Arrays.stream(stem(forms)).map(List::of).toList();
+        }
+
+        /** @return whether this adapter exposes genuine alternative outputs */
+        default boolean supportsMultipleOutputs() { return false; }
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/RadixorBenchmarkStemmer.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/RadixorBenchmarkStemmer.java
@@ -0,0 +1,103 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.util.LinkedHashSet;
+import java.util.List;
+import java.util.Objects;
+import java.util.Set;
+
+import org.egothor.stemmer.CompiledPatchCommand;
+import org.egothor.stemmer.FrequencyTrie;
+
+/**
+ * Benchmark-only Radixor stemmer adapter for the canonical preferred-result
+ * path over normalized benchmark tokens.
+ *
+ * <p>
+ * The benchmark corpus is normalized during setup, so this adapter uses
+ * {@link FrequencyTrie#getNormalizedString(String)} to avoid measuring
+ * redundant lookup-time normalization. Patch commands are applied with the
+ * traversal direction persisted in the trie metadata.
+ * </p>
+ *
+ * <p>
+ * Instances are mutable and intended for one JMH worker thread.
+ * </p>
+ */
+final class RadixorBenchmarkStemmer {
+
+    /**
+     * Compiled Radixor patch trie with decoded patch-command values.
+     */
+    private final FrequencyTrie<CompiledPatchCommand> trie;
+
+    /**
+     * Creates a benchmark stemmer around one compiled Radixor trie.
+     *
+     * @param trie compiled Radixor patch trie
+     */
+    RadixorBenchmarkStemmer(final FrequencyTrie<CompiledPatchCommand> trie) {
+        this.trie = Objects.requireNonNull(trie, "trie");
+    }
+
+    /**
+     * Stems one benchmark token through the canonical trie lookup API.
+     *
+     * @param token input token
+     * @return Radixor stem or the input token when no patch is stored
+     */
+    String stem(final String token) {
+        final CompiledPatchCommand patch = this.trie.getNormalizedString(token);
+        if (patch == null || patch.preservesAllSources()) {
+            return token;
+        }
+        return patch.apply(token);
+    }
+
+    /**
+     * Returns every distinct candidate stem from the ranked {@code getAll} path,
+     * always including the deterministic primary output.
+     *
+     * @param token original input token
+     * @return immutable candidate list in deterministic ranked order
+     */
+    List<String> stemAll(final String token) {
+        final String primary = stem(token);
+        final Set<String> candidates = new LinkedHashSet<>();
+        candidates.add(primary);
+        final CompiledPatchCommand[] patches = this.trie.getAll(token);
+        for (CompiledPatchCommand patch : patches) {
+            candidates.add(patch.preservesAllSources() ? token : patch.apply(token));
+        }
+        return List.copyOf(candidates);
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/SnowballLanguageCase.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/SnowballLanguageCase.java
@@ -0,0 +1,196 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import org.egothor.stemmer.StemmerPatchTrieLoader;
+import org.egothor.stemmer.benchmark.snowball.ext.danishStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.dutchStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.finnishStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.frenchStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.germanStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.hungarianStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.italianStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.norwegianStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.portugueseStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.russianStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.spanishStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.swedishStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.yiddishStemmer;
+
+/**
+ * Maps Radixor dictionary languages to matching official Snowball algorithms.
+ */
+enum SnowballLanguageCase {
+
+    /**
+     * Danish Snowball stemming over the Radixor Danish dictionary.
+     */
+    DANISH("Danish", StemmerPatchTrieLoader.Language.DA_DK, danishStemmer::new, "Danish"),
+
+    /**
+     * Dutch Snowball stemming over the Radixor Dutch dictionary.
+     */
+    DUTCH("Dutch", StemmerPatchTrieLoader.Language.NL_NL, dutchStemmer::new, "Dutch"),
+
+    /**
+     * Finnish Snowball stemming over the Radixor Finnish dictionary.
+     */
+    FINNISH("Finnish", StemmerPatchTrieLoader.Language.FI_FI, finnishStemmer::new, "Finnish"),
+
+    /**
+     * French Snowball stemming over the Radixor French dictionary.
+     */
+    FRENCH("French", StemmerPatchTrieLoader.Language.FR_FR, frenchStemmer::new, "French"),
+
+    /**
+     * German Snowball stemming over the Radixor German dictionary.
+     */
+    GERMAN("German", StemmerPatchTrieLoader.Language.DE_DE, germanStemmer::new, "German"),
+
+    /**
+     * Hungarian Snowball stemming over the Radixor Hungarian dictionary.
+     */
+    HUNGARIAN("Hungarian", StemmerPatchTrieLoader.Language.HU_HU, hungarianStemmer::new, "Hungarian"),
+
+    /**
+     * Italian Snowball stemming over the Radixor Italian dictionary.
+     */
+    ITALIAN("Italian", StemmerPatchTrieLoader.Language.IT_IT, italianStemmer::new, "Italian"),
+
+    /**
+     * Norwegian Snowball stemming over the Radixor Bokmal dictionary.
+     */
+    NORWEGIAN_BOKMAL("Norwegian Bokmal", StemmerPatchTrieLoader.Language.NB_NO, norwegianStemmer::new,
+            "Norwegian"),
+
+    /**
+     * Norwegian Snowball stemming over the Radixor Nynorsk dictionary.
+     */
+    NORWEGIAN_NYNORSK("Norwegian Nynorsk", StemmerPatchTrieLoader.Language.NN_NO, norwegianStemmer::new,
+            "Norwegian"),
+
+    /**
+     * Portuguese Snowball stemming over the Radixor Portuguese dictionary.
+     */
+    PORTUGUESE("Portuguese", StemmerPatchTrieLoader.Language.PT_PT, portugueseStemmer::new, "Portuguese"),
+
+    /**
+     * Russian Snowball stemming over the Radixor Russian dictionary.
+     */
+    RUSSIAN("Russian", StemmerPatchTrieLoader.Language.RU_RU, russianStemmer::new, "Russian"),
+
+    /**
+     * Spanish Snowball stemming over the Radixor Spanish dictionary.
+     */
+    SPANISH("Spanish", StemmerPatchTrieLoader.Language.ES_ES, spanishStemmer::new, "Spanish"),
+
+    /**
+     * Swedish Snowball stemming over the Radixor Swedish dictionary.
+     */
+    SWEDISH("Swedish", StemmerPatchTrieLoader.Language.SV_SE, swedishStemmer::new, "Swedish"),
+
+    /**
+     * Yiddish Snowball stemming over the Radixor Yiddish dictionary.
+     */
+    YIDDISH("Yiddish", StemmerPatchTrieLoader.Language.YI, yiddishStemmer::new, "Yiddish");
+
+    /**
+     * Human-readable language name.
+     */
+    private final String displayLanguage;
+
+    /**
+     * Matching Radixor language resource.
+     */
+    private final StemmerPatchTrieLoader.Language radixorLanguage;
+
+    /**
+     * Factory for the isolated benchmark-only Snowball implementation.
+     */
+    private final SnowballStemmerAdapter.Factory directFactory;
+
+    /**
+     * Lucene SnowballFilter algorithm name.
+     */
+    private final String luceneSnowballName;
+
+    /**
+     * Creates a language case.
+     *
+     * @param displayLanguage   human-readable language name
+     * @param radixorLanguage   matching Radixor language resource
+     * @param directFactory     direct Snowball stemmer factory
+     * @param luceneSnowballName Lucene SnowballFilter algorithm name
+     */
+    SnowballLanguageCase(final String displayLanguage, final StemmerPatchTrieLoader.Language radixorLanguage,
+            final SnowballStemmerAdapter.Factory directFactory, final String luceneSnowballName) {
+        this.displayLanguage = displayLanguage;
+        this.radixorLanguage = radixorLanguage;
+        this.directFactory = directFactory;
+        this.luceneSnowballName = luceneSnowballName;
+    }
+
+    /**
+     * Returns the human-readable language name.
+     *
+     * @return display language
+     */
+    String displayLanguage() {
+        return this.displayLanguage;
+    }
+
+    /**
+     * Returns the matching Radixor dictionary language.
+     *
+     * @return Radixor language
+     */
+    StemmerPatchTrieLoader.Language radixorLanguage() {
+        return this.radixorLanguage;
+    }
+
+    /**
+     * Creates a direct Snowball stemmer adapter.
+     *
+     * @return direct Snowball adapter
+     */
+    SnowballStemmerAdapter createDirectStemmer() {
+        return new SnowballStemmerAdapter(this.directFactory);
+    }
+
+    /**
+     * Returns the Lucene SnowballFilter algorithm name.
+     *
+     * @return Lucene SnowballFilter algorithm name
+     */
+    String luceneSnowballName() {
+        return this.luceneSnowballName;
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/SnowballLanguageStemmerComparisonBenchmark.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/SnowballLanguageStemmerComparisonBenchmark.java
@@ -0,0 +1,246 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.io.IOException;
+import java.util.concurrent.TimeUnit;
+
+import org.apache.lucene.analysis.LowerCaseFilter;
+import org.apache.lucene.analysis.TokenStream;
+import org.apache.lucene.analysis.snowball.SnowballFilter;
+import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;
+import org.egothor.stemmer.FrequencyTrie;
+import org.egothor.stemmer.ReductionMode;
+import org.egothor.stemmer.StemmerPatchTrieLoader;
+import org.openjdk.jmh.annotations.Benchmark;
+import org.openjdk.jmh.annotations.BenchmarkMode;
+import org.openjdk.jmh.annotations.Level;
+import org.openjdk.jmh.annotations.Measurement;
+import org.openjdk.jmh.annotations.Mode;
+import org.openjdk.jmh.annotations.OutputTimeUnit;
+import org.openjdk.jmh.annotations.Param;
+import org.openjdk.jmh.annotations.Scope;
+import org.openjdk.jmh.annotations.Setup;
+import org.openjdk.jmh.annotations.State;
+import org.openjdk.jmh.annotations.Warmup;
+import org.openjdk.jmh.infra.Blackhole;
+
+/**
+ * Compares Radixor with official Snowball algorithms for every Radixor language
+ * that has a matching Snowball Java stemmer.
+ *
+ * <p>
+ * Each benchmark operation processes the same changed-token Radixor
+ * dictionary-derived language corpus, repeated only when the changed-token
+ * resource contains fewer than 5,000 token fields. The direct Snowball method
+ * measures the isolated benchmark-only Snowball source. The Lucene
+ * SnowballFilter method measures Lucene's TokenStream integration path,
+ * including lower-case normalization and token attribute overhead.
+ * </p>
+ */
+@BenchmarkMode(Mode.AverageTime)
+@OutputTimeUnit(TimeUnit.NANOSECONDS)
+@Warmup(iterations = 3, time = 1, timeUnit = TimeUnit.SECONDS)
+@Measurement(iterations = 5, time = 1, timeUnit = TimeUnit.SECONDS)
+public class SnowballLanguageStemmerComparisonBenchmark {
+
+    /**
+     * Shared language corpus and Radixor trie state.
+     */
+    @State(Scope.Benchmark)
+    public static class SharedState {
+
+        /**
+         * Language/algorithm case under comparison.
+         */
+        @Param({ "DANISH", "DUTCH", "FINNISH", "FRENCH", "GERMAN", "HUNGARIAN", "ITALIAN",
+                "NORWEGIAN_BOKMAL", "NORWEGIAN_NYNORSK", "PORTUGUESE", "RUSSIAN", "SPANISH", "SWEDISH",
+                "YIDDISH" })
+        public String languageCaseName;
+
+        /**
+         * Resolved language/algorithm case.
+         */
+        private SnowballLanguageCase languageCase;
+
+        /**
+         * Shared deterministic changed-token dictionary corpus.
+         */
+        private String[] tokens;
+
+        /**
+         * Compiled Radixor trie for the selected language.
+         */
+        private RadixorBenchmarkStemmer radixorStemmer;
+
+        /**
+         * Initializes shared language resources before measurement.
+         *
+         * @throws IOException if the corpus or trie cannot be loaded
+         */
+        @Setup(Level.Trial)
+        public void setUp() throws IOException {
+            this.languageCase = SnowballLanguageCase.valueOf(this.languageCaseName);
+            this.tokens = LanguageBenchmarkCorpus.createTokens(this.languageCase.radixorLanguage());
+            this.radixorStemmer = new RadixorBenchmarkStemmer(StemmerPatchTrieLoader.loadCompiled(
+                    this.languageCase.radixorLanguage(), true,
+                    ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS));
+        }
+    }
+
+    /**
+     * Per-thread direct Snowball state.
+     */
+    @State(Scope.Thread)
+    public static class DirectState {
+
+        /**
+         * Reusable direct Snowball adapter.
+         */
+        private SnowballStemmerAdapter snowballStemmer;
+
+        /**
+         * Initializes direct Snowball state for the selected language.
+         *
+         * @param sharedState selected language state
+         */
+        @Setup(Level.Trial)
+        public void setUp(final SharedState sharedState) {
+            this.snowballStemmer = sharedState.languageCase.createDirectStemmer();
+        }
+    }
+
+    /**
+     * Per-thread Lucene SnowballFilter state.
+     */
+    @State(Scope.Thread)
+    public static class LuceneSnowballState {
+
+        /**
+         * Reusable benchmark input stream.
+         */
+        private BenchmarkTokenStream input;
+
+        /**
+         * Reusable Lucene SnowballFilter output stream.
+         */
+        private TokenStream output;
+
+        /**
+         * Reusable term attribute.
+         */
+        private CharTermAttribute termAttribute;
+
+        /**
+         * Initializes Lucene SnowballFilter state for the selected language.
+         *
+         * @param sharedState selected language state
+         */
+        @Setup(Level.Trial)
+        public void setUp(final SharedState sharedState) {
+            this.input = new BenchmarkTokenStream(new String[0]);
+            final TokenStream normalizedInput = new LowerCaseFilter(this.input);
+            this.output = new SnowballFilter(normalizedInput, sharedState.languageCase.luceneSnowballName());
+            this.termAttribute = this.output.addAttribute(CharTermAttribute.class);
+        }
+
+        /**
+         * Runs the reusable Lucene SnowballFilter over one corpus.
+         *
+         * <p>
+         * The benchmark intentionally rebinds the {@code String[]} corpus on every
+         * measured operation so the adaptation cost from the canonical string input to
+         * Lucene's mutable character attributes is included.
+         * </p>
+         *
+         * @param tokens    token corpus
+         * @param blackhole result sink
+         * @throws IOException if Lucene token streaming fails
+         */
+        void run(final String[] tokens, final Blackhole blackhole) throws IOException {
+            this.input.setTokens(tokens);
+            this.output.reset();
+            while (this.output.incrementToken()) {
+                blackhole.consume(this.termAttribute.toString());
+            }
+            this.output.end();
+        }
+    }
+
+    /**
+     * Runs Radixor over the selected Snowball-language corpus.
+     *
+     * @param sharedState shared benchmark state
+     * @param blackhole   result sink
+     */
+    @Benchmark
+    public void radixor(final SharedState sharedState, final Blackhole blackhole) {
+        final String[] tokens = sharedState.tokens;
+        final RadixorBenchmarkStemmer stemmer = sharedState.radixorStemmer;
+
+        for (String token : tokens) {
+            blackhole.consume(stemmer.stem(token));
+        }
+    }
+
+    /**
+     * Runs the official Snowball direct Java implementation over the selected
+     * language corpus.
+     *
+     * @param sharedState shared benchmark state
+     * @param directState reusable direct Snowball state
+     * @param blackhole   result sink
+     */
+    @Benchmark
+    public void snowballDirect(final SharedState sharedState, final DirectState directState,
+            final Blackhole blackhole) {
+        final String[] tokens = sharedState.tokens;
+        final SnowballStemmerAdapter stemmer = directState.snowballStemmer;
+
+        for (String token : tokens) {
+            blackhole.consume(stemmer.stem(token));
+        }
+    }
+
+    /**
+     * Runs Lucene SnowballFilter over the selected language corpus.
+     *
+     * @param sharedState shared benchmark state
+     * @param luceneState reusable Lucene Snowball state
+     * @param blackhole   result sink
+     * @throws IOException if Lucene token streaming fails
+     */
+    @Benchmark
+    public void luceneSnowballFilter(final SharedState sharedState, final LuceneSnowballState luceneState,
+            final Blackhole blackhole) throws IOException {
+        luceneState.run(sharedState.tokens, blackhole);
+    }
+}
--- a/src/jmh/java/org/egothor/stemmer/benchmark/SnowballStemmerAdapter.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/SnowballStemmerAdapter.java
@@ -32,7 +32,7 @@ package org.egothor.stemmer.benchmark;

 import java.util.Objects;

-import org.tartarus.snowball.SnowballStemmer;
+import org.egothor.stemmer.benchmark.snowball.SnowballStemmer;

 /**
 * Small adapter around a Snowball stemmer instance used by benchmarks.
--- a/src/jmh/java/org/egothor/stemmer/benchmark/StemmerComparisonBenchmarkQuality.java
+++ b/src/jmh/java/org/egothor/stemmer/benchmark/StemmerComparisonBenchmarkQuality.java
@@ -0,0 +1,923 @@
+/*******************************************************************************
+ * Copyright (C) 2026, Leo Galambos
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ *
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * 3. Neither the name of the copyright holder nor the names of its contributors
+ *    may be used to endorse or promote products derived from this software
+ *    without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ ******************************************************************************/
+package org.egothor.stemmer.benchmark;
+
+import java.io.IOException;
+import java.net.URL;
+import java.util.List;
+import java.util.Objects;
+import java.util.concurrent.TimeUnit;
+import java.util.function.Function;
+
+import org.apache.lucene.analysis.LowerCaseFilter;
+import org.apache.lucene.analysis.TokenStream;
+import org.apache.lucene.analysis.ar.ArabicNormalizationFilter;
+import org.apache.lucene.analysis.core.DecimalDigitFilter;
+import org.apache.lucene.analysis.cz.CzechStemFilter;
+import org.apache.lucene.analysis.de.GermanLightStemFilter;
+import org.apache.lucene.analysis.de.GermanMinimalStemFilter;
+import org.apache.lucene.analysis.de.GermanNormalizationFilter;
+import org.apache.lucene.analysis.de.GermanStemFilter;
+import org.apache.lucene.analysis.en.EnglishMinimalStemFilter;
+import org.apache.lucene.analysis.en.EnglishPossessiveFilter;
+import org.apache.lucene.analysis.en.KStemFilter;
+import org.apache.lucene.analysis.en.PorterStemFilter;
+import org.apache.lucene.analysis.es.SpanishLightStemFilter;
+import org.apache.lucene.analysis.es.SpanishMinimalStemFilter;
+import org.apache.lucene.analysis.es.SpanishPluralStemFilter;
+import org.apache.lucene.analysis.fa.PersianNormalizationFilter;
+import org.apache.lucene.analysis.fa.PersianStemFilter;
+import org.apache.lucene.analysis.fi.FinnishLightStemFilter;
+import org.apache.lucene.analysis.fr.FrenchLightStemFilter;
+import org.apache.lucene.analysis.fr.FrenchMinimalStemFilter;
+import org.apache.lucene.analysis.hu.HungarianLightStemFilter;
+import org.apache.lucene.analysis.it.ItalianLightStemFilter;
+import org.apache.lucene.analysis.morfologik.MorfologikFilter;
+import org.apache.lucene.analysis.no.NorwegianLightStemFilter;
+import org.apache.lucene.analysis.no.NorwegianMinimalStemFilter;
+import org.apache.lucene.analysis.pl.PolishAnalyzer;
+import org.apache.lucene.analysis.pt.PortugueseLightStemFilter;
+import org.apache.lucene.analysis.pt.PortugueseMinimalStemFilter;
+import org.apache.lucene.analysis.pt.PortugueseStemFilter;
+import org.apache.lucene.analysis.ru.RussianLightStemFilter;
+import org.apache.lucene.analysis.snowball.SnowballFilter;
+import org.apache.lucene.analysis.stempel.StempelFilter;
+import org.apache.lucene.analysis.stempel.StempelStemmer;
+import org.apache.lucene.analysis.sv.SwedishLightStemFilter;
+import org.apache.lucene.analysis.sv.SwedishMinimalStemFilter;
+import org.apache.lucene.analysis.tokenattributes.CharTermAttribute;
+import org.apache.lucene.analysis.tokenattributes.PositionIncrementAttribute;
+import org.egothor.stemmer.FrequencyTrie;
+import org.egothor.stemmer.ReductionMode;
+import org.egothor.stemmer.StemmerPatchTrieLoader;
+import org.egothor.stemmer.benchmark.snowball.ext.englishStemmer;
+import org.egothor.stemmer.benchmark.snowball.ext.porterStemmer;
+import org.openjdk.jmh.annotations.AuxCounters;
+import org.openjdk.jmh.annotations.Benchmark;
+import org.openjdk.jmh.annotations.BenchmarkMode;
+import org.openjdk.jmh.annotations.Fork;
+import org.openjdk.jmh.annotations.Level;
+import org.openjdk.jmh.annotations.Measurement;
+import org.openjdk.jmh.annotations.Mode;
+import org.openjdk.jmh.annotations.OutputTimeUnit;
+import org.openjdk.jmh.annotations.Param;
+import org.openjdk.jmh.annotations.Scope;
+import org.openjdk.jmh.annotations.Setup;
+import org.openjdk.jmh.annotations.State;
+import org.openjdk.jmh.annotations.Warmup;
+import org.openjdk.jmh.infra.Blackhole;
+
+import morfologik.stemming.Dictionary;
+import morfologik.stemming.DictionaryLookup;
+import morfologik.stemming.WordData;
+
+/**
+ * Emits exact-root agreement metrics through standard JMH result files.
+ *
+ * <p>
+ * This benchmark is a quality pass, not a throughput competitor. Each operation
+ * evaluates one stemmer against the complete Radixor dictionary resource for
+ * the matching language. The useful outputs are the JMH auxiliary counters
+ * {@code correctMatches}, {@code evaluatedTokens},
+ * {@code changedCorrectMatches}, {@code changedEvaluatedTokens},
+ * {@code rootPreservedMatches}, and {@code rootEvaluatedTokens}.
+ * </p>
+ */
+@BenchmarkMode(Mode.AverageTime)
+@OutputTimeUnit(TimeUnit.NANOSECONDS)
+@Warmup(iterations = 0)
+@Measurement(iterations = 1, time = 1, timeUnit = TimeUnit.MILLISECONDS)
+@Fork(0)
+public class StemmerComparisonBenchmarkQuality {
+
+    /**
+     * Shared quality state for one candidate stemmer.
+     */
+    @State(Scope.Benchmark)
+    public static class QualityState {
+
+        /**
+         * Candidate stemmer whose exact-root agreement is measured.
+         */
+        @Param({
+                "ENGLISH_RADIXOR",
+                "ENGLISH_SNOWBALL_ORIGINAL_PORTER",
+                "ENGLISH_SNOWBALL_PORTER2",
+                "ENGLISH_LUCENE_PORTER_COPIED",
+                "ENGLISH_LUCENE_PORTER_FILTER",
+                "ENGLISH_LUCENE_KSTEM_FILTER",
+                "ENGLISH_LUCENE_MINIMAL_FILTER",
+                "ENGLISH_LUCENE_POSSESSIVE_FILTER",
+                "ENGLISH_PAICE_HUSK_LANCASTER",
+                "ENGLISH_OPENNLP_PORTER",
+                "CZECH_RADIXOR",
+                "CZECH_LUCENE_CZECH_STEM_FILTER",
+                "GERMAN_RADIXOR",
+                "GERMAN_LUCENE_GERMAN_STEM_FILTER",
+                "GERMAN_LUCENE_GERMAN_LIGHT_STEM_FILTER",
+                "GERMAN_LUCENE_GERMAN_MINIMAL_STEM_FILTER",
+                "GERMAN_CISTEM",
+                "SPANISH_RADIXOR",
+                "SPANISH_LUCENE_SPANISH_LIGHT_STEM_FILTER",
+                "SPANISH_LUCENE_SPANISH_MINIMAL_STEM_FILTER",
+                "SPANISH_LUCENE_SPANISH_PLURAL_STEM_FILTER",
+                "PERSIAN_RADIXOR",
+                "PERSIAN_LUCENE_PERSIAN_STEM_FILTER",
+                "FINNISH_RADIXOR",
+                "FINNISH_LUCENE_FINNISH_LIGHT_STEM_FILTER",
+                "FRENCH_RADIXOR",
+                "FRENCH_LUCENE_FRENCH_LIGHT_STEM_FILTER",
+                "FRENCH_LUCENE_FRENCH_MINIMAL_STEM_FILTER",
+                "HUNGARIAN_RADIXOR",
+                "HUNGARIAN_LUCENE_HUNGARIAN_LIGHT_STEM_FILTER",
+                "ITALIAN_RADIXOR",
+                "ITALIAN_LUCENE_ITALIAN_LIGHT_STEM_FILTER",
+                "NORWEGIAN_BOKMAL_RADIXOR",
+                "NORWEGIAN_BOKMAL_LUCENE_NORWEGIAN_LIGHT_STEM_FILTER",
+                "NORWEGIAN_BOKMAL_LUCENE_NORWEGIAN_MINIMAL_STEM_FILTER",
+                "POLISH_RADIXOR",
+                "POLISH_LUCENE_STEMPEL_DIRECT",
+                "POLISH_LUCENE_STEMPEL_FILTER",
+                "POLISH_LUCENE_MORFOLOGIK_FILTER",
+                "PORTUGUESE_RADIXOR",
+                "PORTUGUESE_LUCENE_PORTUGUESE_STEM_FILTER",
+                "PORTUGUESE_LUCENE_PORTUGUESE_LIGHT_STEM_FILTER",
+                "PORTUGUESE_LUCENE_PORTUGUESE_MINIMAL_STEM_FILTER",
+                "RUSSIAN_RADIXOR",
+                "RUSSIAN_LUCENE_RUSSIAN_LIGHT_STEM_FILTER",
+                "SWEDISH_RADIXOR",
+                "SWEDISH_LUCENE_SWEDISH_LIGHT_STEM_FILTER",
+                "SWEDISH_LUCENE_SWEDISH_MINIMAL_STEM_FILTER",
+                "UKRAINIAN_RADIXOR",
+                "UKRAINIAN_MORFOLOGIK_DIRECT",
+                "UKRAINIAN_LUCENE_MORFOLOGIK_FILTER",
+                "SNOWBALL_DANISH_DIRECT",
+                "SNOWBALL_DANISH_LUCENE_FILTER",
+                "SNOWBALL_DUTCH_DIRECT",
+                "SNOWBALL_DUTCH_LUCENE_FILTER",
+                "SNOWBALL_FINNISH_DIRECT",
+                "SNOWBALL_FINNISH_LUCENE_FILTER",
+                "SNOWBALL_FRENCH_DIRECT",
+                "SNOWBALL_FRENCH_LUCENE_FILTER",
+                "SNOWBALL_GERMAN_DIRECT",
+                "SNOWBALL_GERMAN_LUCENE_FILTER",
+                "SNOWBALL_HUNGARIAN_DIRECT",
+                "SNOWBALL_HUNGARIAN_LUCENE_FILTER",
+                "SNOWBALL_ITALIAN_DIRECT",
+                "SNOWBALL_ITALIAN_LUCENE_FILTER",
+                "SNOWBALL_NORWEGIAN_BOKMAL_DIRECT",
+                "SNOWBALL_NORWEGIAN_BOKMAL_LUCENE_FILTER",
+                "SNOWBALL_NORWEGIAN_NYNORSK_DIRECT",
+                "SNOWBALL_NORWEGIAN_NYNORSK_LUCENE_FILTER",
+                "SNOWBALL_PORTUGUESE_DIRECT",
+                "SNOWBALL_PORTUGUESE_LUCENE_FILTER",
+                "SNOWBALL_RUSSIAN_DIRECT",
+                "SNOWBALL_RUSSIAN_LUCENE_FILTER",
+                "SNOWBALL_SPANISH_DIRECT",
+                "SNOWBALL_SPANISH_LUCENE_FILTER",
+                "SNOWBALL_SWEDISH_DIRECT",
+                "SNOWBALL_SWEDISH_LUCENE_FILTER",
+                "SNOWBALL_YIDDISH_DIRECT",
+                "SNOWBALL_YIDDISH_LUCENE_FILTER"
+        })
+        public String candidateName;
+
+        /**
+         * Full dictionary corpus for the selected language.
+         */
+        private LanguageBenchmarkCorpus.Corpus corpus;
+
+        /**
+         * Candidate evaluator.
+         */
+        private QualityEvaluator evaluator;
+
+        /**
+         * Initializes corpus and evaluator before measurement.
+         *
+         * @throws IOException if dictionary or stemmer resources cannot be loaded
+         */
+        @Setup(Level.Trial)
+        public void setUp() throws IOException {
+            final QualityCandidate candidate = QualityCandidate.valueOf(this.candidateName);
+            this.corpus = LanguageBenchmarkCorpus.createFullCorpus(candidate.radixorLanguage());
+            this.evaluator = candidate.createEvaluator();
+        }
+    }
+
+    /**
+     * JMH auxiliary counters for exact-root agreement.
+     */
+    @State(Scope.Thread)
+    @AuxCounters(AuxCounters.Type.EVENTS)
+    public static class QualityCounters {
+
+        /**
+         * Number of outputs equal to the dictionary root.
+         */
+        public long correctMatches;
+
+        /**
+         * Number of evaluated input tokens.
+         */
+        public long evaluatedTokens;
+
+        /**
+         * Number of exact-root matches where the input token differs from the
+         * expected root.
+         */
+        public long changedCorrectMatches;
+
+        /**
+         * Number of evaluated tokens where the input token differs from the expected
+         * root.
+         */
+        public long changedEvaluatedTokens;
+
+        /**
+         * Number of exact-root matches where the input token is already the expected
+         * root.
+         */
+        public long rootPreservedMatches;
+
+        /**
+         * Number of evaluated tokens where the input token is already the expected
+         * root.
+         */
+        public long rootEvaluatedTokens;
+
+        /**
+         * Resets counters before each measured iteration.
+         */
+        @Setup(Level.Iteration)
+        public void reset() {
+            this.correctMatches = 0L;
+            this.evaluatedTokens = 0L;
+            this.changedCorrectMatches = 0L;
+            this.changedEvaluatedTokens = 0L;
+            this.rootPreservedMatches = 0L;
+            this.rootEvaluatedTokens = 0L;
+        }
+    }
+
+    /**
+     * Runs exact-root agreement over the full dictionary corpus.
+     *
+     * @param state quality state
+     * @param counters auxiliary JMH counters
+     * @param blackhole result sink
+     * @return exact-root match count for this operation
+     * @throws IOException if Lucene streaming fails
+     */
+    @Benchmark
+    public int exactRootAgreement(final QualityState state, final QualityCounters counters, final Blackhole blackhole)
+            throws IOException {
+        final QualityResult result = state.evaluator.evaluate(state.corpus, blackhole);
+        counters.correctMatches += result.correctMatches();
+        counters.evaluatedTokens += result.evaluatedTokens();
+        counters.changedCorrectMatches += result.changedCorrectMatches();
+        counters.changedEvaluatedTokens += result.changedEvaluatedTokens();
+        counters.rootPreservedMatches += result.rootPreservedMatches();
+        counters.rootEvaluatedTokens += result.rootEvaluatedTokens();
+        return result.correctMatches();
+    }
+
+    /**
+     * Candidate stemmers that can be evaluated against a Radixor resource.
+     */
+    enum QualityCandidate {
+        ENGLISH_RADIXOR(StemmerPatchTrieLoader.Language.US_UK),
+        ENGLISH_SNOWBALL_ORIGINAL_PORTER(StemmerPatchTrieLoader.Language.US_UK),
+        ENGLISH_SNOWBALL_PORTER2(StemmerPatchTrieLoader.Language.US_UK),
+        ENGLISH_LUCENE_PORTER_COPIED(StemmerPatchTrieLoader.Language.US_UK),
+        ENGLISH_LUCENE_PORTER_FILTER(StemmerPatchTrieLoader.Language.US_UK),
+        ENGLISH_LUCENE_KSTEM_FILTER(StemmerPatchTrieLoader.Language.US_UK),
+        ENGLISH_LUCENE_MINIMAL_FILTER(StemmerPatchTrieLoader.Language.US_UK),
+        ENGLISH_LUCENE_POSSESSIVE_FILTER(StemmerPatchTrieLoader.Language.US_UK),
+        ENGLISH_PAICE_HUSK_LANCASTER(StemmerPatchTrieLoader.Language.US_UK),
+        ENGLISH_OPENNLP_PORTER(StemmerPatchTrieLoader.Language.US_UK),
+        CZECH_RADIXOR(StemmerPatchTrieLoader.Language.CS_CZ),
+        CZECH_LUCENE_CZECH_STEM_FILTER(StemmerPatchTrieLoader.Language.CS_CZ),
+        GERMAN_RADIXOR(StemmerPatchTrieLoader.Language.DE_DE),
+        GERMAN_LUCENE_GERMAN_STEM_FILTER(StemmerPatchTrieLoader.Language.DE_DE),
+        GERMAN_LUCENE_GERMAN_LIGHT_STEM_FILTER(StemmerPatchTrieLoader.Language.DE_DE),
+        GERMAN_LUCENE_GERMAN_MINIMAL_STEM_FILTER(StemmerPatchTrieLoader.Language.DE_DE),
+        GERMAN_CISTEM(StemmerPatchTrieLoader.Language.DE_DE),
+        SPANISH_RADIXOR(StemmerPatchTrieLoader.Language.ES_ES),
+        SPANISH_LUCENE_SPANISH_LIGHT_STEM_FILTER(StemmerPatchTrieLoader.Language.ES_ES),
+        SPANISH_LUCENE_SPANISH_MINIMAL_STEM_FILTER(StemmerPatchTrieLoader.Language.ES_ES),
+        SPANISH_LUCENE_SPANISH_PLURAL_STEM_FILTER(StemmerPatchTrieLoader.Language.ES_ES),
+        PERSIAN_RADIXOR(StemmerPatchTrieLoader.Language.FA_IR),
+        PERSIAN_LUCENE_PERSIAN_STEM_FILTER(StemmerPatchTrieLoader.Language.FA_IR),
+        FINNISH_RADIXOR(StemmerPatchTrieLoader.Language.FI_FI),
+        FINNISH_LUCENE_FINNISH_LIGHT_STEM_FILTER(StemmerPatchTrieLoader.Language.FI_FI),
+        FRENCH_RADIXOR(StemmerPatchTrieLoader.Language.FR_FR),
+        FRENCH_LUCENE_FRENCH_LIGHT_STEM_FILTER(StemmerPatchTrieLoader.Language.FR_FR),
+        FRENCH_LUCENE_FRENCH_MINIMAL_STEM_FILTER(StemmerPatchTrieLoader.Language.FR_FR),
+        HUNGARIAN_RADIXOR(StemmerPatchTrieLoader.Language.HU_HU),
+        HUNGARIAN_LUCENE_HUNGARIAN_LIGHT_STEM_FILTER(StemmerPatchTrieLoader.Language.HU_HU),
+        ITALIAN_RADIXOR(StemmerPatchTrieLoader.Language.IT_IT),
+        ITALIAN_LUCENE_ITALIAN_LIGHT_STEM_FILTER(StemmerPatchTrieLoader.Language.IT_IT),
+        NORWEGIAN_BOKMAL_RADIXOR(StemmerPatchTrieLoader.Language.NB_NO),
+        NORWEGIAN_BOKMAL_LUCENE_NORWEGIAN_LIGHT_STEM_FILTER(StemmerPatchTrieLoader.Language.NB_NO),
+        NORWEGIAN_BOKMAL_LUCENE_NORWEGIAN_MINIMAL_STEM_FILTER(StemmerPatchTrieLoader.Language.NB_NO),
+        POLISH_RADIXOR(StemmerPatchTrieLoader.Language.PL_PL),
+        POLISH_LUCENE_STEMPEL_DIRECT(StemmerPatchTrieLoader.Language.PL_PL),
+        POLISH_LUCENE_STEMPEL_FILTER(StemmerPatchTrieLoader.Language.PL_PL),
+        POLISH_LUCENE_MORFOLOGIK_FILTER(StemmerPatchTrieLoader.Language.PL_PL),
+        PORTUGUESE_RADIXOR(StemmerPatchTrieLoader.Language.PT_PT),
+        PORTUGUESE_LUCENE_PORTUGUESE_STEM_FILTER(StemmerPatchTrieLoader.Language.PT_PT),
+        PORTUGUESE_LUCENE_PORTUGUESE_LIGHT_STEM_FILTER(StemmerPatchTrieLoader.Language.PT_PT),
+        PORTUGUESE_LUCENE_PORTUGUESE_MINIMAL_STEM_FILTER(StemmerPatchTrieLoader.Language.PT_PT),
+        RUSSIAN_RADIXOR(StemmerPatchTrieLoader.Language.RU_RU),
+        RUSSIAN_LUCENE_RUSSIAN_LIGHT_STEM_FILTER(StemmerPatchTrieLoader.Language.RU_RU),
+        SWEDISH_RADIXOR(StemmerPatchTrieLoader.Language.SV_SE),
+        SWEDISH_LUCENE_SWEDISH_LIGHT_STEM_FILTER(StemmerPatchTrieLoader.Language.SV_SE),
+        SWEDISH_LUCENE_SWEDISH_MINIMAL_STEM_FILTER(StemmerPatchTrieLoader.Language.SV_SE),
+        UKRAINIAN_RADIXOR(StemmerPatchTrieLoader.Language.UK_UA),
+        UKRAINIAN_MORFOLOGIK_DIRECT(StemmerPatchTrieLoader.Language.UK_UA),
+        UKRAINIAN_LUCENE_MORFOLOGIK_FILTER(StemmerPatchTrieLoader.Language.UK_UA),
+        SNOWBALL_DANISH_DIRECT(StemmerPatchTrieLoader.Language.DA_DK, SnowballLanguageCase.DANISH),
+        SNOWBALL_DANISH_LUCENE_FILTER(StemmerPatchTrieLoader.Language.DA_DK, SnowballLanguageCase.DANISH),
+        SNOWBALL_DUTCH_DIRECT(StemmerPatchTrieLoader.Language.NL_NL, SnowballLanguageCase.DUTCH),
+        SNOWBALL_DUTCH_LUCENE_FILTER(StemmerPatchTrieLoader.Language.NL_NL, SnowballLanguageCase.DUTCH),
+        SNOWBALL_FINNISH_DIRECT(StemmerPatchTrieLoader.Language.FI_FI, SnowballLanguageCase.FINNISH),
+        SNOWBALL_FINNISH_LUCENE_FILTER(StemmerPatchTrieLoader.Language.FI_FI, SnowballLanguageCase.FINNISH),
+        SNOWBALL_FRENCH_DIRECT(StemmerPatchTrieLoader.Language.FR_FR, SnowballLanguageCase.FRENCH),
+        SNOWBALL_FRENCH_LUCENE_FILTER(StemmerPatchTrieLoader.Language.FR_FR, SnowballLanguageCase.FRENCH),
+        SNOWBALL_GERMAN_DIRECT(StemmerPatchTrieLoader.Language.DE_DE, SnowballLanguageCase.GERMAN),
+        SNOWBALL_GERMAN_LUCENE_FILTER(StemmerPatchTrieLoader.Language.DE_DE, SnowballLanguageCase.GERMAN),
+        SNOWBALL_HUNGARIAN_DIRECT(StemmerPatchTrieLoader.Language.HU_HU, SnowballLanguageCase.HUNGARIAN),
+        SNOWBALL_HUNGARIAN_LUCENE_FILTER(StemmerPatchTrieLoader.Language.HU_HU, SnowballLanguageCase.HUNGARIAN),
+        SNOWBALL_ITALIAN_DIRECT(StemmerPatchTrieLoader.Language.IT_IT, SnowballLanguageCase.ITALIAN),
+        SNOWBALL_ITALIAN_LUCENE_FILTER(StemmerPatchTrieLoader.Language.IT_IT, SnowballLanguageCase.ITALIAN),
+        SNOWBALL_NORWEGIAN_BOKMAL_DIRECT(StemmerPatchTrieLoader.Language.NB_NO,
+                SnowballLanguageCase.NORWEGIAN_BOKMAL),
+        SNOWBALL_NORWEGIAN_BOKMAL_LUCENE_FILTER(StemmerPatchTrieLoader.Language.NB_NO,
+                SnowballLanguageCase.NORWEGIAN_BOKMAL),
+        SNOWBALL_NORWEGIAN_NYNORSK_DIRECT(StemmerPatchTrieLoader.Language.NN_NO,
+                SnowballLanguageCase.NORWEGIAN_NYNORSK),
+        SNOWBALL_NORWEGIAN_NYNORSK_LUCENE_FILTER(StemmerPatchTrieLoader.Language.NN_NO,
+                SnowballLanguageCase.NORWEGIAN_NYNORSK),
+        SNOWBALL_PORTUGUESE_DIRECT(StemmerPatchTrieLoader.Language.PT_PT, SnowballLanguageCase.PORTUGUESE),
+        SNOWBALL_PORTUGUESE_LUCENE_FILTER(StemmerPatchTrieLoader.Language.PT_PT, SnowballLanguageCase.PORTUGUESE),
+        SNOWBALL_RUSSIAN_DIRECT(StemmerPatchTrieLoader.Language.RU_RU, SnowballLanguageCase.RUSSIAN),
+        SNOWBALL_RUSSIAN_LUCENE_FILTER(StemmerPatchTrieLoader.Language.RU_RU, SnowballLanguageCase.RUSSIAN),
+        SNOWBALL_SPANISH_DIRECT(StemmerPatchTrieLoader.Language.ES_ES, SnowballLanguageCase.SPANISH),
+        SNOWBALL_SPANISH_LUCENE_FILTER(StemmerPatchTrieLoader.Language.ES_ES, SnowballLanguageCase.SPANISH),
+        SNOWBALL_SWEDISH_DIRECT(StemmerPatchTrieLoader.Language.SV_SE, SnowballLanguageCase.SWEDISH),
+        SNOWBALL_SWEDISH_LUCENE_FILTER(StemmerPatchTrieLoader.Language.SV_SE, SnowballLanguageCase.SWEDISH),
+        SNOWBALL_YIDDISH_DIRECT(StemmerPatchTrieLoader.Language.YI, SnowballLanguageCase.YIDDISH),
+        SNOWBALL_YIDDISH_LUCENE_FILTER(StemmerPatchTrieLoader.Language.YI, SnowballLanguageCase.YIDDISH);
+
+        /**
+         * Radixor dictionary language used as truth.
+         */
+        private final StemmerPatchTrieLoader.Language radixorLanguage;
+
+        /**
+         * Optional Snowball language mapping.
+         */
+        private final SnowballLanguageCase snowballLanguageCase;
+
+        /**
+         * Creates a candidate.
+         *
+         * @param radixorLanguage Radixor dictionary language
+         */
+        QualityCandidate(final StemmerPatchTrieLoader.Language radixorLanguage) {
+            this(radixorLanguage, null);
+        }
+
+        /**
+         * Creates a candidate.
+         *
+         * @param radixorLanguage Radixor dictionary language
+         * @param snowballLanguageCase matching Snowball case
+         */
+        QualityCandidate(final StemmerPatchTrieLoader.Language radixorLanguage,
+                final SnowballLanguageCase snowballLanguageCase) {
+            this.radixorLanguage = radixorLanguage;
+            this.snowballLanguageCase = snowballLanguageCase;
+        }
+
+        /**
+         * Returns the Radixor dictionary language.
+         *
+         * @return Radixor language
+         */
+        StemmerPatchTrieLoader.Language radixorLanguage() {
+            return this.radixorLanguage;
+        }
+
+        /**
+         * Creates the evaluator for this candidate.
+         *
+         * @return quality evaluator
+         * @throws IOException if stemmer resources cannot be loaded
+         */
+        CandidateStemmer createStemmer() throws IOException {
+            if (name().endsWith("_RADIXOR")) {
+                return radixor(createRadixorStemmer(this.radixorLanguage));
+            }
+            if (name().endsWith("_DIRECT") && this.snowballLanguageCase != null) {
+                return direct(this.snowballLanguageCase.createDirectStemmer()::stem);
+            }
+            if (name().endsWith("_LUCENE_FILTER") && this.snowballLanguageCase != null) {
+                return tokenFilter(input -> new SnowballFilter(new LowerCaseFilter(input),
+                        this.snowballLanguageCase.luceneSnowballName()));
+            }
+
+            return switch (this) {
+                case ENGLISH_SNOWBALL_ORIGINAL_PORTER -> direct(new SnowballStemmerAdapter(porterStemmer::new)::stem);
+                case ENGLISH_SNOWBALL_PORTER2 -> direct(new SnowballStemmerAdapter(englishStemmer::new)::stem);
+                case ENGLISH_LUCENE_PORTER_COPIED -> direct(new LucenePorterStemmerCopied()::stem);
+                case ENGLISH_LUCENE_PORTER_FILTER -> tokenFilter(PorterStemFilter::new);
+                case ENGLISH_LUCENE_KSTEM_FILTER -> tokenFilter(KStemFilter::new);
+                case ENGLISH_LUCENE_MINIMAL_FILTER -> tokenFilter(EnglishMinimalStemFilter::new);
+                case ENGLISH_LUCENE_POSSESSIVE_FILTER -> tokenFilter(EnglishPossessiveFilter::new);
+                case ENGLISH_PAICE_HUSK_LANCASTER -> direct(new PaiceHuskLancasterStemmer()::stem);
+                case ENGLISH_OPENNLP_PORTER -> {
+                    final opennlp.tools.stemmer.PorterStemmer stemmer =
+                            new opennlp.tools.stemmer.PorterStemmer();
+                    yield direct(token -> stemmer.stem(token).toString());
+                }
+                case CZECH_LUCENE_CZECH_STEM_FILTER -> tokenFilter(input -> new CzechStemFilter(lowercase(input)));
+                case GERMAN_LUCENE_GERMAN_STEM_FILTER -> tokenFilter(input -> new GermanStemFilter(lowercase(input)));
+                case GERMAN_LUCENE_GERMAN_LIGHT_STEM_FILTER ->
+                    tokenFilter(input -> new GermanLightStemFilter(germanNormalize(input)));
+                case GERMAN_LUCENE_GERMAN_MINIMAL_STEM_FILTER ->
+                    tokenFilter(input -> new GermanMinimalStemFilter(germanNormalize(input)));
+                case GERMAN_CISTEM -> direct(createGermanCistemStemmer());
+                case SPANISH_LUCENE_SPANISH_LIGHT_STEM_FILTER ->
+                    tokenFilter(input -> new SpanishLightStemFilter(lowercase(input)));
+                case SPANISH_LUCENE_SPANISH_MINIMAL_STEM_FILTER ->
+                    tokenFilter(input -> new SpanishMinimalStemFilter(lowercase(input)));
+                case SPANISH_LUCENE_SPANISH_PLURAL_STEM_FILTER ->
+                    tokenFilter(input -> new SpanishPluralStemFilter(lowercase(input)));
+                case PERSIAN_LUCENE_PERSIAN_STEM_FILTER ->
+                    tokenFilter(input -> new PersianStemFilter(persianNormalize(input)));
+                case FINNISH_LUCENE_FINNISH_LIGHT_STEM_FILTER ->
+                    tokenFilter(input -> new FinnishLightStemFilter(lowercase(input)));
+                case FRENCH_LUCENE_FRENCH_LIGHT_STEM_FILTER ->
+                    tokenFilter(input -> new FrenchLightStemFilter(lowercase(input)));
+                case FRENCH_LUCENE_FRENCH_MINIMAL_STEM_FILTER ->
+                    tokenFilter(input -> new FrenchMinimalStemFilter(lowercase(input)));
+                case HUNGARIAN_LUCENE_HUNGARIAN_LIGHT_STEM_FILTER ->
+                    tokenFilter(input -> new HungarianLightStemFilter(lowercase(input)));
+                case ITALIAN_LUCENE_ITALIAN_LIGHT_STEM_FILTER ->
+                    tokenFilter(input -> new ItalianLightStemFilter(lowercase(input)));
+                case NORWEGIAN_BOKMAL_LUCENE_NORWEGIAN_LIGHT_STEM_FILTER ->
+                    tokenFilter(input -> new NorwegianLightStemFilter(lowercase(input)));
+                case NORWEGIAN_BOKMAL_LUCENE_NORWEGIAN_MINIMAL_STEM_FILTER ->
+                    tokenFilter(input -> new NorwegianMinimalStemFilter(lowercase(input)));
+                case POLISH_LUCENE_STEMPEL_DIRECT -> {
+                    final StempelStemmer stemmer = new StempelStemmer(PolishAnalyzer.getDefaultTable());
+                    yield direct(token -> {
+                        final StringBuilder stem = stemmer.stem(token);
+                        return stem == null ? token : stem.toString();
+                    });
+                }
+                case POLISH_LUCENE_STEMPEL_FILTER ->
+                    tokenFilter(input -> new StempelFilter(input, new StempelStemmer(PolishAnalyzer.getDefaultTable())));
+                case POLISH_LUCENE_MORFOLOGIK_FILTER -> tokenFilter(MorfologikFilter::new, true);
+                case PORTUGUESE_LUCENE_PORTUGUESE_STEM_FILTER ->
+                    tokenFilter(input -> new PortugueseStemFilter(lowercase(input)));
+                case PORTUGUESE_LUCENE_PORTUGUESE_LIGHT_STEM_FILTER ->
+                    tokenFilter(input -> new PortugueseLightStemFilter(lowercase(input)));
+                case PORTUGUESE_LUCENE_PORTUGUESE_MINIMAL_STEM_FILTER ->
+                    tokenFilter(input -> new PortugueseMinimalStemFilter(lowercase(input)));
+                case RUSSIAN_LUCENE_RUSSIAN_LIGHT_STEM_FILTER ->
+                    tokenFilter(input -> new RussianLightStemFilter(lowercase(input)));
+                case SWEDISH_LUCENE_SWEDISH_LIGHT_STEM_FILTER ->
+                    tokenFilter(input -> new SwedishLightStemFilter(lowercase(input)));
+                case SWEDISH_LUCENE_SWEDISH_MINIMAL_STEM_FILTER ->
+                    tokenFilter(input -> new SwedishMinimalStemFilter(lowercase(input)));
+                case UKRAINIAN_MORFOLOGIK_DIRECT -> {
+                    final DictionaryLookup lookup = new DictionaryLookup(loadUkrainianMorfologikDictionary());
+                    yield morphologik(lookup);
+                }
+                case UKRAINIAN_LUCENE_MORFOLOGIK_FILTER -> {
+                    final Dictionary dictionary = loadUkrainianMorfologikDictionary();
+                    yield tokenFilter(input -> new MorfologikFilter(input, dictionary), true);
+                }
+                default -> throw new IllegalStateException("No evaluator for " + this + ".");
+            };
+        }
+
+        /** Creates the exact-root evaluator used by the JMH quality benchmark. */
+        QualityEvaluator createEvaluator() throws IOException {
+            return exactRootEvaluator(createStemmer());
+        }
+    }
+
+    /**
+     * Creates a CISTEM stemmer adapter.
+     *
+     * @return German stem function
+     */
+    private static Stemmer createGermanCistemStemmer() {
+        return Cistem::stem;
+    }
+
+    /**
+     * Direct stemmer function.
+     */
+    @FunctionalInterface
+    private interface Stemmer {
+
+        /**
+         * Produces one stem.
+         *
+         * @param token input token
+         * @return produced stem
+         */
+        String stem(String token);
+    }
+
+    /**
+     * Quality evaluator for one candidate.
+     */
+    @FunctionalInterface
+    private interface QualityEvaluator {
+
+        /**
+         * Evaluates exact-root agreement for one corpus.
+         *
+         * @param corpus token/root corpus
+         * @param blackhole result sink
+         * @return exact-root match count
+         * @throws IOException if Lucene streaming fails
+         */
+        QualityResult evaluate(LanguageBenchmarkCorpus.Corpus corpus, Blackhole blackhole) throws IOException;
+    }
+
+    /** Stateful candidate adapter confined to one sequential evaluation scenario. */
+    @FunctionalInterface
+    interface CandidateStemmer {
+
+        /**
+         * Stems a deterministic batch through the authoritative JMH invocation path.
+         *
+         * @param tokens input tokens, never {@code null}
+         * @return one non-null output for every input token
+         * @throws IOException if a token-stream implementation fails
+         */
+        String[] stem(String[] tokens) throws IOException;
+
+        /**
+         * Returns complete distinct candidate sets, each containing its primary output.
+         * Single-output adapters expose singleton lists.
+         *
+         * @param tokens input tokens
+         * @return immutable candidate list for every token
+         * @throws IOException if adapter processing fails
+         */
+        default List<List<String>> stemCandidates(final String[] tokens) throws IOException {
+            final String[] primary = stem(tokens);
+            return java.util.Arrays.stream(primary).map(List::of).toList();
+        }
+
+        /** @return whether the adapter exposes genuine alternative outputs */
+        default boolean supportsMultipleOutputs() {
+            return false;
+        }
+    }
+
+    /** Creates the candidate-capable Radixor adapter backed by ranked {@code getAll}. */
+    private static CandidateStemmer radixor(final RadixorBenchmarkStemmer stemmer) {
+        return new CandidateStemmer() {
+            /** {@inheritDoc} */
+            @Override public String[] stem(final String[] tokens) {
+                final String[] outputs = new String[tokens.length];
+                for (int index = 0; index < tokens.length; index++) { outputs[index] = stemmer.stem(tokens[index]); }
+                return outputs;
+            }
+            /** {@inheritDoc} */
+            @Override public List<List<String>> stemCandidates(final String[] tokens) {
+                return java.util.Arrays.stream(tokens).map(stemmer::stemAll).toList();
+            }
+            /** {@inheritDoc} */
+            @Override public boolean supportsMultipleOutputs() { return true; }
+        };
+    }
+
+    /**
+     * Creates the authoritative multi-output Radixor adapter for a validated dictionary language.
+     *
+     * @param language bundled Radixor language
+     * @return scenario-confined adapter using the JMH invocation path
+     * @throws IOException if the compiled dictionary cannot be loaded
+     */
+    static CandidateStemmer createRadixorQualityStemmer(final StemmerPatchTrieLoader.Language language)
+            throws IOException {
+        return radixor(createRadixorStemmer(language));
+    }
+
+    /**
+     * Exact-root agreement counters for one quality operation.
+     *
+     * @param correctMatches total exact-root matches
+     * @param evaluatedTokens total evaluated tokens
+     * @param changedCorrectMatches exact-root matches where token differs from root
+     * @param changedEvaluatedTokens evaluated tokens where token differs from root
+     * @param rootPreservedMatches exact-root matches where token already equals root
+     * @param rootEvaluatedTokens evaluated tokens where token already equals root
+     */
+    private record QualityResult(int correctMatches, int evaluatedTokens, int changedCorrectMatches,
+            int changedEvaluatedTokens, int rootPreservedMatches, int rootEvaluatedTokens) {
+    }
+
+    /**
+     * Creates a direct candidate adapter.
+     *
+     * @param stemmer direct stemmer
+     * @return sequential batch adapter
+     */
+    private static CandidateStemmer direct(final Stemmer stemmer) {
+        Objects.requireNonNull(stemmer, "stemmer");
+        return tokens -> {
+            final String[] outputs = new String[tokens.length];
+            for (int index = 0; index < tokens.length; index++) {
+                outputs[index] = stemmer.stem(tokens[index]);
+            }
+            return outputs;
+        };
+    }
+
+    /**
+     * Creates a TokenFilter candidate adapter.
+     *
+     * @param factory token stream factory
+     * @return sequential batch adapter
+     */
+    private static CandidateStemmer tokenFilter(final Function<TokenStream, TokenStream> factory) {
+        Objects.requireNonNull(factory, "factory");
+        return tokens -> firstTokenFilterOutputs(tokens, factory, null);
+    }
+
+    /** Creates a TokenFilter adapter that preserves all terms emitted per position. */
+    private static CandidateStemmer tokenFilter(final Function<TokenStream, TokenStream> factory,
+            final boolean multipleOutputs) {
+        if (!multipleOutputs) { return tokenFilter(factory); }
+        return new CandidateStemmer() {
+            /** {@inheritDoc} */
+            @Override public String[] stem(final String[] tokens) throws IOException {
+                return firstTokenFilterOutputs(tokens, factory, null);
+            }
+            /** {@inheritDoc} */
+            @Override public List<List<String>> stemCandidates(final String[] tokens) throws IOException {
+                return allTokenFilterOutputs(tokens, factory);
+            }
+            /** {@inheritDoc} */
+            @Override public boolean supportsMultipleOutputs() { return true; }
+        };
+    }
+
+    /** Creates a multi-analysis Morphologik direct adapter. */
+    private static CandidateStemmer morphologik(final DictionaryLookup lookup) {
+        return new CandidateStemmer() {
+            /** {@inheritDoc} */
+            @Override public String[] stem(final String[] tokens) {
+                final String[] outputs = new String[tokens.length];
+                for (int index = 0; index < tokens.length; index++) { outputs[index] = firstMorfologikStem(tokens[index], lookup); }
+                return outputs;
+            }
+            /** {@inheritDoc} */
+            @Override public List<List<String>> stemCandidates(final String[] tokens) {
+                return java.util.Arrays.stream(tokens).map(token -> allMorfologikStems(token, lookup)).toList();
+            }
+            /** {@inheritDoc} */
+            @Override public boolean supportsMultipleOutputs() { return true; }
+        };
+    }
+
+    /**
+     * Creates exact-root accounting around an authoritative candidate adapter.
+     *
+     * @param stemmer candidate adapter
+     * @return JMH exact-root evaluator
+     */
+    private static QualityEvaluator exactRootEvaluator(final CandidateStemmer stemmer) {
+        Objects.requireNonNull(stemmer, "stemmer");
+        return (corpus, blackhole) -> {
+            final String[] actualStems = stemmer.stem(corpus.tokens());
+            final String[] expectedRoots = corpus.expectedRoots();
+            final String[] tokens = corpus.tokens();
+            int correct = 0;
+            int changedCorrect = 0;
+            int changedEvaluated = 0;
+            int rootPreserved = 0;
+            int rootEvaluated = 0;
+            for (int index = 0; index < actualStems.length; index++) {
+                final String token = tokens[index];
+                final String expectedRoot = expectedRoots[index];
+                blackhole.consume(actualStems[index]);
+                final boolean exact = Objects.equals(expectedRoot, actualStems[index]);
+                if (exact) {
+                    correct++;
+                }
+                if (Objects.equals(token, expectedRoot)) {
+                    rootEvaluated++;
+                    if (exact) {
+                        rootPreserved++;
+                    }
+                } else {
+                    changedEvaluated++;
+                    if (exact) {
+                        changedCorrect++;
+                    }
+                }
+            }
+            return new QualityResult(correct, actualStems.length, changedCorrect, changedEvaluated, rootPreserved,
+                    rootEvaluated);
+        };
+    }
+
+    /**
+     * Creates a direct Radixor stemmer.
+     *
+     * @param language Radixor dictionary language
+     * @return direct stemmer
+     * @throws IOException if the trie cannot be loaded
+     */
+    private static RadixorBenchmarkStemmer createRadixorStemmer(final StemmerPatchTrieLoader.Language language) throws IOException {
+        return new RadixorBenchmarkStemmer(StemmerPatchTrieLoader.loadCompiled(
+                language, true, ReductionMode.MERGE_SUBTREES_WITH_EQUIVALENT_RANKED_GET_ALL_RESULTS));
+    }
+
+    /**
+     * Loads the benchmark-only Ukrainian Morfologik dictionary.
+     *
+     * @return Ukrainian Morfologik dictionary
+     * @throws IOException if the dictionary cannot be loaded
+     */
+    private static Dictionary loadUkrainianMorfologikDictionary() throws IOException {
+        final URL dictionaryUrl = StemmerComparisonBenchmarkQuality.class.getClassLoader()
+                .getResource("ua/net/nlp/ukrainian.dict");
+        if (dictionaryUrl == null) {
+            throw new IllegalStateException("Missing Ukrainian Morfologik dictionary resource.");
+        }
+        return Dictionary.read(dictionaryUrl);
+    }
+
+    /**
+     * Returns the first Morfologik stem for one token.
+     *
+     * @param token input token
+     * @param lookup dictionary lookup
+     * @return first Morfologik stem, or the input token when no analysis exists
+     */
+    private static String firstMorfologikStem(final String token, final DictionaryLookup lookup) {
+        final List<WordData> analyses = lookup.lookup(token);
+        if (analyses.isEmpty()) {
+            return token;
+        }
+        return analyses.get(0).getStem().toString();
+    }
+
+    /** Returns all distinct Morphologik lemma strings and always includes the primary output. */
+    private static List<String> allMorfologikStems(final String token, final DictionaryLookup lookup) {
+        final java.util.LinkedHashSet<String> stems = new java.util.LinkedHashSet<>();
+        stems.add(firstMorfologikStem(token, lookup));
+        for (WordData analysis : lookup.lookup(token)) { stems.add(analysis.getStem().toString()); }
+        return List.copyOf(stems);
+    }
+
+    /**
+     * Extracts the first emitted term for each input token from a TokenFilter
+     * pipeline.
+     *
+     * @param tokens token corpus
+     * @param factory token stream factory
+     * @param blackhole result sink
+     * @return first emitted term per input token
+     * @throws IOException if Lucene streaming fails
+     */
+    private static String[] firstTokenFilterOutputs(final String[] tokens, final Function<TokenStream, TokenStream> factory,
+            final Blackhole blackhole) throws IOException {
+        final String[] outputs = new String[tokens.length];
+        final BenchmarkTokenStream input = new BenchmarkTokenStream(tokens);
+        final TokenStream output = factory.apply(input);
+        final CharTermAttribute termAttribute = output.addAttribute(CharTermAttribute.class);
+        final PositionIncrementAttribute positionAttribute = output.addAttribute(PositionIncrementAttribute.class);
+        int inputIndex = -1;
+        boolean recordedForPosition = false;
+
+        output.reset();
+        while (output.incrementToken()) {
+            final int positionIncrement = positionAttribute.getPositionIncrement();
+            if (positionIncrement > 0) {
+                inputIndex += positionIncrement;
+                recordedForPosition = false;
+            }
+            if (inputIndex >= 0 && inputIndex < outputs.length && !recordedForPosition) {
+                outputs[inputIndex] = termAttribute.toString();
+                recordedForPosition = true;
+            }
+            if (blackhole != null) {
+                blackhole.consume(termAttribute);
+            }
+        }
+        output.end();
+        output.close();
+
+        for (int index = 0; index < outputs.length; index++) {
+            if (outputs[index] == null) {
+                outputs[index] = tokens[index];
+            }
+        }
+        return outputs;
+    }
+
+    /**
+     * Extracts every distinct emitted term for each input position and includes the
+     * deterministic primary output even when a filter omits it.
+     */
+    private static List<List<String>> allTokenFilterOutputs(final String[] tokens,
+            final Function<TokenStream, TokenStream> factory) throws IOException {
+        final List<java.util.LinkedHashSet<String>> candidates = new java.util.ArrayList<>(tokens.length);
+        for (int index = 0; index < tokens.length; index++) { candidates.add(new java.util.LinkedHashSet<>()); }
+        final BenchmarkTokenStream input = new BenchmarkTokenStream(tokens);
+        final TokenStream output = factory.apply(input);
+        final CharTermAttribute term = output.addAttribute(CharTermAttribute.class);
+        final PositionIncrementAttribute position = output.addAttribute(PositionIncrementAttribute.class);
+        int inputIndex = -1;
+        output.reset();
+        while (output.incrementToken()) {
+            if (position.getPositionIncrement() > 0) { inputIndex += position.getPositionIncrement(); }
+            if (inputIndex >= 0 && inputIndex < candidates.size()) { candidates.get(inputIndex).add(term.toString()); }
+        }
+        output.end();
+        output.close();
+        final String[] primary = firstTokenFilterOutputs(tokens, factory, null);
+        final List<List<String>> result = new java.util.ArrayList<>(tokens.length);
+        for (int index = 0; index < tokens.length; index++) {
+            candidates.get(index).add(primary[index]);
+            result.add(List.copyOf(candidates.get(index)));
+        }
+        return List.copyOf(result);
+    }
+
+    /**
+     * Adds Lucene lower-case normalization.
+     *
+     * @param input input token stream
+     * @return normalized stream
+     */
+    private static TokenStream lowercase(final TokenStream input) {
+        return new LowerCaseFilter(input);
+    }
+
+    /**
+     * Adds Lucene German normalization.
+     *
+     * @param input input token stream
+     * @return normalized stream
+     */
+    private static TokenStream germanNormalize(final TokenStream input) {
+        return new GermanNormalizationFilter(lowercase(input));
+    }
+
+    /**
+     * Adds Lucene Persian normalization.
+     *
+     * @param input input token stream
+     * @return normalized stream
+     */
+    private static TokenStream persianNormalize(final TokenStream input) {
+        TokenStream result = lowercase(input);
+        result = new DecimalDigitFilter(result);
+        result = new ArabicNormalizationFilter(result);
+        result = new PersianNormalizationFilter(result);
+        return result;
+    }
+}
--- a/src/main/java/module-info.java
+++ b/src/main/java/module-info.java
@@ -0,0 +1,5 @@
+module org.egothor.radixor {
+    requires java.logging;
+
+    exports org.egothor.stemmer;
+}
--- a/src/main/java/org/egothor/stemmer/Compile.java
+++ b/src/main/java/org/egothor/stemmer/Compile.java
@@ -147,6 +147,7 @@ public final class Compile {
     * @param arguments parsed command-line arguments
     * @throws IOException if compilation or output writing fails
     */
+    @SuppressWarnings("deprecation")
    private static void compile(final Arguments arguments) throws IOException {
        final ReductionSettings reductionSettings = new ReductionSettings(arguments.reductionMode(),
                arguments.dominantWinnerMinPercent(), arguments.dominantWinnerOverSecondRatio());
--- a/src/main/java/org/egothor/stemmer/CompiledPatchCommand.java
+++ b/src/main/java/org/egothor/stemmer/CompiledPatchCommand.java
--- a/src/main/java/org/egothor/stemmer/FrequencyTrie.java
+++ b/src/main/java/org/egothor/stemmer/FrequencyTrie.java
@@ -35,6 +35,9 @@ import java.io.DataOutputStream;
 import java.io.IOException;
 import java.io.InputStream;
 import java.io.OutputStream;
+import java.nio.charset.StandardCharsets;
+import java.security.MessageDigest;
+import java.security.NoSuchAlgorithmException;
 import java.util.ArrayList;
 import java.util.Arrays;
 import java.util.Collections;
@@ -86,6 +89,7 @@ import org.egothor.stemmer.trie.ReductionSignature;
 *
 * @param <V> value type
 */
+@SuppressWarnings({ "PMD.CyclomaticComplexity", "PMD.CouplingBetweenObjects" })
 public final class FrequencyTrie<V> {

    /**
@@ -93,6 +97,16 @@ public final class FrequencyTrie<V> {
     */
    private static final Logger LOGGER = Logger.getLogger(FrequencyTrie.class.getName());

+    /**
+     * Domain separator used by the trie fingerprint canonical input.
+     */
+    private static final String FINGERPRINT_DOMAIN = "RADIXOR-FREQUENCY-TRIE-FINGERPRINT";
+
+    /**
+     * Version of the canonical fingerprint input format.
+     */
+    private static final int FINGERPRINT_FORMAT_VERSION = 2;
+
    /**
     * Root node of the compiled read-only trie.
     */
@@ -103,6 +117,12 @@ public final class FrequencyTrie<V> {
     */
    private final TrieMetadata metadata;

+    /**
+     * Canonical SHA-256 fingerprint bytes. The internal array is never exposed
+     * directly to callers.
+     */
+    private final byte[] fingerprintBytes;
+
    /**
     * Cached traversal direction used for key lookup.
     */
@@ -149,7 +169,7 @@ public final class FrequencyTrie<V> {
    /**
     * Binary format version.
     */
-    private static final int STREAM_VERSION = 5;
+    private static final int STREAM_VERSION = 6;

    /**
     * Version where traversal-direction ordinal is persisted.
@@ -166,6 +186,16 @@ public final class FrequencyTrie<V> {
     */
    private static final int CASE_VERSION = 4;

+    /**
+     * Version where the persisted metadata switched to a text block.
+     */
+    private static final int TEXT_METADATA_VERSION = 5;
+
+    /**
+     * Version where contracted accepting nodes are persisted.
+     */
+    private static final int ACCEPTING_NODE_VERSION = 6;
+
    /**
     * Argument name for lookup keys.
     */
@@ -233,12 +263,28 @@ public final class FrequencyTrie<V> {
            final TrieMetadata metadata) {
        this.root = Objects.requireNonNull(root, "root");
        this.metadata = Objects.requireNonNull(metadata, "metadata");
+        this.fingerprintBytes = computeFingerprintBytes(root, metadata);
        this.lookupTraversalDirection = metadata.traversalDirection();
        this.lowercasesLookupKeys = metadata.caseProcessingMode() == CaseProcessingMode.LOWERCASE_WITH_LOCALE_ROOT;
        this.removeDiacritics = metadata.diacriticProcessingMode() == DiacriticProcessingMode.REMOVE;
        this.emptyValues = arrayFactory.apply(0);
    }

+    /**
+     * Creates a trie from an already compiled root.
+     *
+     * @param arrayFactory array factory
+     * @param root         compiled root
+     * @param metadata     trie metadata
+     * @param <V>          value type
+     * @return trie instance
+     */
+    /* default */ static <V> FrequencyTrie<V> fromCompiled(final IntFunction<V[]> arrayFactory,
+            final CompiledNode<V> root,
+            final TrieMetadata metadata) {
+        return new FrequencyTrie<>(arrayFactory, root, metadata);
+    }
+
    /**
     * Returns the most frequent value stored at the node addressed by the supplied
     * key.
@@ -271,6 +317,63 @@ public final class FrequencyTrie<V> {
        return orderedValues[0];
    }

+    /**
+     * Returns the preferred value for an already-normalized key.
+     *
+     * <p>
+     * This method bypasses {@link TrieMetadata#caseProcessingMode()} and
+     * {@link TrieMetadata#diacriticProcessingMode()}. Callers must supply input
+     * normalized exactly as required by this trie's metadata. It is intended for
+     * hot paths where normalization is guaranteed by an upstream tokenizer or
+     * benchmark corpus and repeated lookup-time normalization would be redundant.
+     * </p>
+     *
+     * @param key already-normalized key to resolve
+     * @return most frequent value, or {@code null} if the key does not exist or no
+     *         value is stored at the addressed node
+     * @throws NullPointerException if {@code key} is {@code null}
+     */
+    public V getNormalized(final CharSequence key) {
+        Objects.requireNonNull(key, ARG_KEY);
+        final CompiledNode<V> node = findNode(key);
+        if (node == null) {
+            return null;
+        }
+        final V[] orderedValues = node.orderedValues();
+        if (orderedValues.length == 0) {
+            return null;
+        }
+        return orderedValues[0];
+    }
+
+    /**
+     * Returns the preferred value for an already-normalized {@link String} key.
+     *
+     * <p>
+     * This overload keeps high-volume string lookup on a monomorphic path and
+     * avoids the {@link CharSequence} dispatch used by the general overload.
+     * Callers must supply input normalized exactly as required by this trie's
+     * metadata.
+     * </p>
+     *
+     * @param key already-normalized key to resolve
+     * @return most frequent value, or {@code null} if the key does not exist or no
+     *         value is stored at the addressed node
+     * @throws NullPointerException if {@code key} is {@code null}
+     */
+    public V getNormalizedString(final String key) {
+        Objects.requireNonNull(key, ARG_KEY);
+        final CompiledNode<V> node = findNode(key);
+        if (node == null) {
+            return null;
+        }
+        final V[] orderedValues = node.orderedValues();
+        if (orderedValues.length == 0) {
+            return null;
+        }
+        return orderedValues[0];
+    }
+
    /**
     * Returns all values stored at the node addressed by the supplied key, ordered
     * by descending frequency.
@@ -504,6 +607,60 @@ public final class FrequencyTrie<V> {
        return this.metadata;
    }

+    /**
+     * Returns the deterministic SHA-256 fingerprint of this trie.
+     *
+     * <p>
+     * The fingerprint is a canonical model identity, not a Java object identity. It
+     * includes a fingerprint-domain marker, the fingerprint input format version,
+     * persisted metadata, and the complete compiled-node structure reachable from
+     * the root, including edges, child references, local values, and local counts.
+     * </p>
+     *
+     * <p>
+     * The returned value is stable across JVM runs for equivalent trie content and
+     * metadata. It does not include object identity, memory layout, runtime cache
+     * state, absolute file paths, timestamps, or other process-local state.
+     * </p>
+     *
+     * @return 64-character lowercase hexadecimal SHA-256 fingerprint
+     */
+    public String getFingerprint() {
+        return toLowerHex(this.fingerprintBytes);
+    }
+
+    /**
+     * Returns a defensive copy of the raw SHA-256 fingerprint bytes.
+     *
+     * <p>
+     * The returned array has length {@code 32}. Mutating it does not affect this
+     * trie.
+     * </p>
+     *
+     * @return defensive copy of the 32-byte SHA-256 fingerprint
+     */
+    public byte[] copyFingerprintBytes() {
+        return Arrays.copyOf(this.fingerprintBytes, this.fingerprintBytes.length);
+    }
+
+    private static <V> byte[] computeFingerprintBytes(final CompiledNode<V> root, final TrieMetadata metadata) {
+        final MessageDigest messageDigest = newSha256Digest();
+        updateUtf8(messageDigest, FINGERPRINT_DOMAIN);
+        updateInt(messageDigest, FINGERPRINT_FORMAT_VERSION);
+        updateUtf8(messageDigest, metadata.toTextBlock());
+
+        final Map<CompiledNode<V>, Integer> nodeIds = new IdentityHashMap<>();
+        final List<CompiledNode<V>> orderedNodes = new ArrayList<>();
+        assignNodeIds(root, nodeIds, orderedNodes);
+
+        updateInt(messageDigest, nodeIds.get(root));
+        updateInt(messageDigest, orderedNodes.size());
+        for (CompiledNode<V> node : orderedNodes) {
+            updateNodeFingerprint(messageDigest, node, nodeIds);
+        }
+        return messageDigest.digest();
+    }
+
    /**
     * Returns the root node mainly for diagnostics and tests within the package.
     *
@@ -670,6 +827,7 @@ public final class FrequencyTrie<V> {
     */
    private static <V> void writeNode(final DataOutputStream dataOutput, final ValueStreamCodec<V> valueCodec,
            final CompiledNode<V> node, final Map<CompiledNode<V>, Integer> nodeIds) throws IOException {
+        dataOutput.writeBoolean(node.acceptsRemainingInput());
        dataOutput.writeInt(node.edgeLabels().length);
        for (int index = 0; index < node.edgeLabels().length; index++) {
            dataOutput.writeChar(node.edgeLabels()[index]);
@@ -687,6 +845,64 @@ public final class FrequencyTrie<V> {
        }
    }

+    private static MessageDigest newSha256Digest() {
+        try {
+            return MessageDigest.getInstance("SHA-256");
+        } catch (NoSuchAlgorithmException exception) {
+            throw new IllegalStateException("SHA-256 digest is not available.", exception);
+        }
+    }
+
+    private static <V> void updateNodeFingerprint(final MessageDigest messageDigest, final CompiledNode<V> node,
+            final Map<CompiledNode<V>, Integer> nodeIds) {
+        final char[] edgeLabels = node.edgeLabels();
+        final CompiledNode<V>[] children = node.children();
+        final V[] values = node.orderedValues();
+        final int[] counts = node.orderedCounts();
+
+        updateInt(messageDigest, node.acceptsRemainingInput() ? 1 : 0);
+        updateInt(messageDigest, edgeLabels.length);
+        for (char edgeLabel : edgeLabels) {
+            updateInt(messageDigest, edgeLabel);
+        }
+        for (CompiledNode<V> child : children) {
+            final Integer childNodeId = nodeIds.get(child);
+            if (childNodeId == null) {
+                throw new IllegalStateException("Missing child node identifier during trie fingerprinting.");
+            }
+            updateInt(messageDigest, childNodeId);
+        }
+
+        updateInt(messageDigest, values.length);
+        for (V value : values) {
+            updateUtf8(messageDigest, String.valueOf(value));
+        }
+        for (int count : counts) {
+            updateInt(messageDigest, count);
+        }
+    }
+
+    private static void updateUtf8(final MessageDigest messageDigest, final String value) {
+        final byte[] encoded = value.getBytes(StandardCharsets.UTF_8);
+        updateInt(messageDigest, encoded.length);
+        messageDigest.update(encoded);
+    }
+
+    private static void updateInt(final MessageDigest messageDigest, final int value) {
+        messageDigest.update((byte) (value >>> 24));
+        messageDigest.update((byte) (value >>> 16));
+        messageDigest.update((byte) (value >>> 8));
+        messageDigest.update((byte) value);
+    }
+
+    private static String toLowerHex(final byte[] digest) {
+        final StringBuilder builder = new StringBuilder(digest.length * 2);
+        for (byte item : digest) {
+            builder.append(Character.forDigit((item >>> 4) & 0x0F, 16)).append(Character.forDigit(item & 0x0F, 16));
+        }
+        return builder.toString();
+    }
+
    /**
     * Internal helper that materializes serialized trie data.
     *
@@ -731,7 +947,7 @@ public final class FrequencyTrie<V> {
            final TrieMetadata sourceMetadata = readMetadata(dataInput, version);
            final int effectiveMaxExpandedIndex = maxExpandedIndex >= 0 ? maxExpandedIndex : DEFAULT_MAX_EXPANDED_INDEX;
            final CompiledNode<V>[] nodes = readNodes(dataInput, arrayFactory, valueCodec, nodeCount,
-                    effectiveMaxExpandedIndex);
+                    effectiveMaxExpandedIndex, version);
            final CompiledNode<V> rootNode = nodes[rootNodeId];

            if (LOGGER.isLoggable(Level.FINE)) {
@@ -748,8 +964,8 @@ public final class FrequencyTrie<V> {

        private static TrieMetadata readMetadata(final DataInputStream dataInput, final int version)
                throws IOException {
-            if (version == STREAM_VERSION) {
-                return readTextMetadata(dataInput);
+            if (version >= TEXT_METADATA_VERSION) {
+                return readTextMetadata(dataInput, version);
            }

            final WordTraversalDirection traversalDirection = readTraversalDirection(dataInput, version);
@@ -766,9 +982,10 @@ public final class FrequencyTrie<V> {
                    caseProcessingMode);
        }

-        private static TrieMetadata readTextMetadata(final DataInputStream dataInput) throws IOException {
+        private static TrieMetadata readTextMetadata(final DataInputStream dataInput, final int version)
+                throws IOException {
            try {
-                return TrieMetadata.fromTextBlock(STREAM_VERSION, dataInput.readUTF());
+                return TrieMetadata.fromTextBlock(version, dataInput.readUTF());
            } catch (IllegalArgumentException exception) {
                throw new IOException("Invalid metadata block.", exception);
            }
@@ -804,14 +1021,19 @@ public final class FrequencyTrie<V> {

        private static <V> CompiledNode<V>[] readNodes(final DataInputStream dataInput,
                final IntFunction<V[]> arrayFactory, final ValueStreamCodec<V> valueCodec, final int nodeCount,
-                final int maxExpandedIndex) throws IOException {
+                final int maxExpandedIndex, final int version) throws IOException {
            final char[][] edgeLabelsByNode = new char[nodeCount][];
            final int[][] childNodeIdsByNode = new int[nodeCount][];
            @SuppressWarnings("unchecked")
            final V[][] orderedValuesByNode = (V[][]) new Object[nodeCount][];
            final int[][] orderedCountsByNode = new int[nodeCount][];
+            final boolean[] acceptsRemainingInputByNode = new boolean[nodeCount];

            for (int nodeIndex = 0; nodeIndex < nodeCount; nodeIndex++) {
+                if (version >= ACCEPTING_NODE_VERSION) {
+                    acceptsRemainingInputByNode[nodeIndex] = dataInput.readBoolean();
+                }
+
                final int edgeCount = dataInput.readInt();
                if (edgeCount < 0) {
                    throw new IOException("Negative edge count at node " + nodeIndex + ": " + edgeCount);
@@ -831,6 +1053,12 @@ public final class FrequencyTrie<V> {
                if (valueCount < 0) {
                    throw new IOException("Negative value count at node " + nodeIndex + ": " + valueCount);
                }
+                if (acceptsRemainingInputByNode[nodeIndex] && edgeCount != 0) {
+                    throw new IOException("Accepting node " + nodeIndex + " cannot have child edges.");
+                }
+                if (acceptsRemainingInputByNode[nodeIndex] && valueCount == 0) {
+                    throw new IOException("Accepting node " + nodeIndex + " must store at least one value.");
+                }

                orderedValuesByNode[nodeIndex] = arrayFactory.apply(valueCount);
                orderedCountsByNode[nodeIndex] = new int[valueCount];
@@ -851,7 +1079,7 @@ public final class FrequencyTrie<V> {

            for (int nodeIndex = 0; nodeIndex < nodeCount; nodeIndex++) {
                nodes[nodeIndex] = resolveNode(nodeIndex, edgeLabelsByNode, childNodeIdsByNode, orderedValuesByNode,
-                        orderedCountsByNode, nodes, inProgress, maxExpandedIndex);
+                        orderedCountsByNode, acceptsRemainingInputByNode, nodes, inProgress, maxExpandedIndex);
            }

            return nodes;
@@ -859,8 +1087,8 @@ public final class FrequencyTrie<V> {

        private static <V> CompiledNode<V> resolveNode(final int nodeIndex, final char[][] edgeLabelsByNode,
                final int[][] childNodeIdsByNode, final V[][] orderedValuesByNode, final int[][] orderedCountsByNode,
-                final CompiledNode<V>[] nodes, final boolean[] inProgress, final int maxExpandedIndex)
-                throws IOException {
+                final boolean[] acceptsRemainingInputByNode, final CompiledNode<V>[] nodes,
+                final boolean[] inProgress, final int maxExpandedIndex) throws IOException {
            final CompiledNode<V> cachedNode = nodes[nodeIndex];
            if (cachedNode != null) {
                return cachedNode;
@@ -885,11 +1113,12 @@ public final class FrequencyTrie<V> {
                                + ": " + childNodeId);
                    }
                    children[edgeIndex] = resolveNode(childNodeId, edgeLabelsByNode, childNodeIdsByNode,
-                            orderedValuesByNode, orderedCountsByNode, nodes, inProgress, maxExpandedIndex);
+                            orderedValuesByNode, orderedCountsByNode, acceptsRemainingInputByNode, nodes, inProgress,
+                            maxExpandedIndex);
                }

                final CompiledNode<V> node = new CompiledNode<>(edgeLabels, children, orderedValuesByNode[nodeIndex],
-                        maxExpandedIndex, orderedCountsByNode[nodeIndex]);
+                        acceptsRemainingInputByNode[nodeIndex], maxExpandedIndex, orderedCountsByNode[nodeIndex]);
                nodes[nodeIndex] = node;
                return node;
            } finally {
@@ -915,7 +1144,30 @@ public final class FrequencyTrie<V> {
     * @return compiled node, or {@code null} if the path does not exist
     */
    private CompiledNode<V> findNode(final String key) {
-        return findNode((CharSequence) key);
+        CompiledNode<V> current = this.root;
+        if (this.lookupTraversalDirection == WordTraversalDirection.BACKWARD) {
+            for (int traversalOffset = key.length() - 1; traversalOffset >= 0; traversalOffset--) {
+                if (current.acceptsRemainingInput()) {
+                    return current;
+                }
+                current = current.findChild(key.charAt(traversalOffset));
+                if (current == null) {
+                    return null;
+                }
+            }
+            return current;
+        }
+
+        for (int traversalOffset = 0; traversalOffset < key.length(); traversalOffset++) {
+            if (current.acceptsRemainingInput()) {
+                return current;
+            }
+            current = current.findChild(key.charAt(traversalOffset));
+            if (current == null) {
+                return null;
+            }
+        }
+        return current;
    }

    /**
@@ -928,6 +1180,9 @@ public final class FrequencyTrie<V> {
        CompiledNode<V> current = this.root;
        if (this.lookupTraversalDirection == WordTraversalDirection.BACKWARD) {
            for (int traversalOffset = key.length() - 1; traversalOffset >= 0; traversalOffset--) {
+                if (current.acceptsRemainingInput()) {
+                    return current;
+                }
                current = current.findChild(key.charAt(traversalOffset));
                if (current == null) {
                    return null;
@@ -937,6 +1192,9 @@ public final class FrequencyTrie<V> {
        }

        for (int traversalOffset = 0; traversalOffset < key.length(); traversalOffset++) {
+            if (current.acceptsRemainingInput()) {
+                return current;
+            }
            current = current.findChild(key.charAt(traversalOffset));
            if (current == null) {
                return null;
@@ -957,6 +1215,9 @@ public final class FrequencyTrie<V> {
        CompiledNode<V> current = this.root;
        if (this.lookupTraversalDirection == WordTraversalDirection.BACKWARD) {
            for (int traversalOffset = offset + length - 1; traversalOffset >= offset; traversalOffset--) {
+                if (current.acceptsRemainingInput()) {
+                    return current;
+                }
                current = current.findChild(key[traversalOffset]);
                if (current == null) {
                    return null;
@@ -967,6 +1228,9 @@ public final class FrequencyTrie<V> {

        final int endExclusive = offset + length;
        for (int traversalOffset = offset; traversalOffset < endExclusive; traversalOffset++) {
+            if (current.acceptsRemainingInput()) {
+                return current;
+            }
            current = current.findChild(key[traversalOffset]);
            if (current == null) {
                return null;
@@ -1403,21 +1667,31 @@ public final class FrequencyTrie<V> {
         * @return canonical reduced node
         */
        private ReducedNode<V> reduce(final MutableNode<V> source, final ReductionContext<V> context) {
-            final Map<Character, ReducedNode<V>> reducedChildren = new LinkedHashMap<>();
+            Map<Character, ReducedNode<V>> reducedChildren = new LinkedHashMap<>();

            for (Map.Entry<Character, MutableNode<V>> childEntry : source.children().entrySet()) {
                final ReducedNode<V> reducedChild = reduce(childEntry.getValue(), context);
                reducedChildren.put(childEntry.getKey(), reducedChild);
            }

-            final Map<V, Integer> localCounts = copyCounts(source.valueCounts());
+            Map<V, Integer> localCounts = copyCounts(source.valueCounts());
+            boolean acceptsRemainingInput = false;
+            if (context.settings().contractUniformSubtrees()) {
+                final Map<V, Integer> contractedCounts = contractUniformSubtree(localCounts, reducedChildren);
+                if (!contractedCounts.isEmpty()) {
+                    localCounts = contractedCounts;
+                    reducedChildren = Collections.emptyMap();
+                    acceptsRemainingInput = true;
+                }
+            }
+
            final LocalValueSummary<V> localSummary = LocalValueSummary.of(localCounts, this.arrayFactory);
            final ReductionSignature<V> signature = ReductionSignature.create(localSummary, reducedChildren,
-                    context.settings());
+                    context.settings(), acceptsRemainingInput);

            ReducedNode<V> canonical = context.lookup(signature);
            if (canonical == null) {
-                canonical = new ReducedNode<>(signature, localCounts, reducedChildren);
+                canonical = new ReducedNode<>(signature, localCounts, reducedChildren, acceptsRemainingInput);
                context.register(signature, canonical);
                return canonical;
            }
@@ -1428,6 +1702,64 @@ public final class FrequencyTrie<V> {
            return canonical;
        }

+        /**
+         * Returns aggregated local counts when the supplied internal subtree contains
+         * one uniform value, otherwise {@code null}.
+         *
+         * @param localCounts     local counts at the current node
+         * @param reducedChildren already reduced children
+         * @return single-value aggregate for a uniform non-leaf subtree, otherwise an
+         *         empty map
+         */
+        private Map<V, Integer> contractUniformSubtree(final Map<V, Integer> localCounts,
+                final Map<Character, ReducedNode<V>> reducedChildren) {
+            if (reducedChildren.isEmpty()) {
+                return Collections.emptyMap();
+            }
+
+            V uniformValue = null;
+            boolean valueSeen = false;
+
+            if (!localCounts.isEmpty()) {
+                if (localCounts.size() != SINGLE_VALUE_COUNT) {
+                    return Collections.emptyMap();
+                }
+                final Map.Entry<V, Integer> localEntry = localCounts.entrySet().iterator().next();
+                uniformValue = localEntry.getKey();
+                valueSeen = true;
+            }
+
+            for (ReducedNode<V> child : reducedChildren.values()) {
+                if (!isSingleValueLeaf(child)) {
+                    return Collections.emptyMap();
+                }
+                final Map.Entry<V, Integer> childEntry = child.localCounts().entrySet().iterator().next();
+                if (valueSeen && !Objects.equals(uniformValue, childEntry.getKey())) {
+                    return Collections.emptyMap();
+                }
+                uniformValue = childEntry.getKey();
+                valueSeen = true;
+            }
+
+            if (!valueSeen) {
+                return Collections.emptyMap();
+            }
+
+            final Map<V, Integer> contractedCounts = new LinkedHashMap<>(SINGLE_VALUE_COUNT);
+            contractedCounts.put(uniformValue, SINGLE_VALUE_COUNT);
+            return contractedCounts;
+        }
+
+        /**
+         * Returns whether the reduced node is a leaf with exactly one stored value.
+         *
+         * @param node node to inspect
+         * @return {@code true} when the node can participate in uniform contraction
+         */
+        private boolean isSingleValueLeaf(final ReducedNode<V> node) {
+            return node.children().isEmpty() && node.localCounts().size() == SINGLE_VALUE_COUNT;
+        }
+
        /**
         * Freezes a reduced node into an immutable compiled node.
         *
@@ -1460,7 +1792,7 @@ public final class FrequencyTrie<V> {
            }

            final CompiledNode<V> frozen = new CompiledNode<>(edges, childNodes, localSummary.orderedValues(),
-                    this.maxExpandedIndex, localSummary.orderedCounts());
+                    reducedNode.acceptsRemainingInput(), this.maxExpandedIndex, localSummary.orderedCounts());
            cache.put(reducedNode, frozen);
            return frozen;
        }
--- a/src/main/java/org/egothor/stemmer/FrequencyTrieBuilders.java
+++ b/src/main/java/org/egothor/stemmer/FrequencyTrieBuilders.java
@@ -30,7 +30,10 @@
 ******************************************************************************/
 package org.egothor.stemmer;

+import java.util.IdentityHashMap;
+import java.util.Map;
 import java.util.Objects;
+import java.util.function.Function;
 import java.util.function.IntFunction;
 import java.util.logging.Level;
 import java.util.logging.Logger;
@@ -115,6 +118,63 @@ public final class FrequencyTrieBuilders {
        return copyOf(source, arrayFactory, ReductionSettings.withDefaults(reductionMode));
    }

+    /**
+     * Reconstructs a compiled trie with every stored value transformed to another
+     * value type.
+     *
+     * <p>
+     * The method preserves logical keys, local value counts, trie metadata, and the
+     * supplied reduction settings. It is intended for runtime specialization, such
+     * as replacing serialized patch-command strings with precompiled patch command
+     * objects without changing the persisted binary trie format.
+     * </p>
+     *
+     * @param source            source compiled trie
+     * @param arrayFactory      array factory for mapped values
+     * @param reductionSettings reduction settings for the mapped trie
+     * @param valueMapper       value mapping function
+     * @param <S>               source value type
+     * @param <T>               target value type
+     * @return compiled trie containing mapped values
+     * @throws NullPointerException if any argument is {@code null}
+     */
+    public static <S, T> FrequencyTrie<T> mapValues(final FrequencyTrie<S> source,
+            final IntFunction<T[]> arrayFactory, final ReductionSettings reductionSettings,
+            final Function<? super S, ? extends T> valueMapper) {
+        Objects.requireNonNull(source, "source");
+        Objects.requireNonNull(arrayFactory, "arrayFactory");
+        Objects.requireNonNull(reductionSettings, "reductionSettings");
+        Objects.requireNonNull(valueMapper, "valueMapper");
+
+        final Map<CompiledNode<S>, CompiledNode<T>> cache = new IdentityHashMap<>();
+        final CompiledNode<T> mappedRoot = mapCompiledNode(source.root(), arrayFactory, valueMapper, cache);
+        final TrieMetadata metadata = TrieMetadata.forCompilation(source.traversalDirection(), reductionSettings,
+                source.metadata().diacriticProcessingMode(), source.metadata().caseProcessingMode());
+
+        LOGGER.log(Level.FINE, "Mapped compiled trie values to a specialized value type.");
+        return FrequencyTrie.fromCompiled(arrayFactory, mappedRoot, metadata);
+    }
+
+    /**
+     * Reconstructs a compiled trie with every stored value transformed to another
+     * value type using default settings for the supplied reduction mode.
+     *
+     * @param source        source compiled trie
+     * @param arrayFactory  array factory for mapped values
+     * @param reductionMode reduction mode for the mapped trie
+     * @param valueMapper   value mapping function
+     * @param <S>           source value type
+     * @param <T>           target value type
+     * @return compiled trie containing mapped values
+     * @throws NullPointerException if any argument is {@code null}
+     */
+    public static <S, T> FrequencyTrie<T> mapValues(final FrequencyTrie<S> source,
+            final IntFunction<T[]> arrayFactory, final ReductionMode reductionMode,
+            final Function<? super S, ? extends T> valueMapper) {
+        Objects.requireNonNull(reductionMode, "reductionMode");
+        return mapValues(source, arrayFactory, ReductionSettings.withDefaults(reductionMode), valueMapper);
+    }
+
    /**
     * Copies one compiled node and all reachable descendants into the target
     * builder.
@@ -138,4 +198,43 @@ public final class FrequencyTrieBuilders {
            keyBuilder.setLength(keyBuilder.length() - 1);
        }
    }
+
+    /**
+     * Maps one compiled node graph while preserving canonical sharing and accepting
+     * leaf semantics.
+     *
+     * @param node         source node
+     * @param arrayFactory target value array factory
+     * @param valueMapper  value mapper
+     * @param cache        identity cache for shared compiled nodes
+     * @param <S>          source value type
+     * @param <T>          target value type
+     * @return mapped compiled node
+     */
+    private static <S, T> CompiledNode<T> mapCompiledNode(final CompiledNode<S> node,
+            final IntFunction<T[]> arrayFactory, final Function<? super S, ? extends T> valueMapper,
+            final Map<CompiledNode<S>, CompiledNode<T>> cache) {
+        final CompiledNode<T> existing = cache.get(node);
+        if (existing != null) {
+            return existing;
+        }
+
+        final CompiledNode<S>[] sourceChildren = node.children();
+        @SuppressWarnings("unchecked")
+        final CompiledNode<T>[] mappedChildren = new CompiledNode[sourceChildren.length];
+        for (int childIndex = 0; childIndex < sourceChildren.length; childIndex++) {
+            mappedChildren[childIndex] = mapCompiledNode(sourceChildren[childIndex], arrayFactory, valueMapper, cache);
+        }
+
+        final S[] sourceValues = node.orderedValues();
+        final T[] mappedValues = arrayFactory.apply(sourceValues.length);
+        for (int valueIndex = 0; valueIndex < sourceValues.length; valueIndex++) {
+            mappedValues[valueIndex] = valueMapper.apply(sourceValues[valueIndex]);
+        }
+
+        final CompiledNode<T> mapped = new CompiledNode<>(node.edgeLabels().clone(), mappedChildren, mappedValues,
+                node.acceptsRemainingInput(), CompiledNode.DEFAULT_MAX_EXPANDED_INDEX, node.orderedCounts().clone());
+        cache.put(node, mapped);
+        return mapped;
+    }
 }
--- a/src/main/java/org/egothor/stemmer/PatchCommandEncoder.java
+++ b/src/main/java/org/egothor/stemmer/PatchCommandEncoder.java
@@ -70,16 +70,6 @@ import java.util.concurrent.locks.ReentrantLock;
@SuppressWarnings({ "PMD.AvoidLiteralsInIfCondition", "PMD.CyclomaticComplexity", "PMD.ForLoopVariableCount" })
 public final class PatchCommandEncoder {

-    /**
-     * Backward direction apply strategy with no runtime direction branching.
-     */
-    private static final ApplyStrategy BACKWARD_APPLY_STRATEGY = PatchCommandEncoder::applyBackward;
-
-    /**
-     * Forward direction apply strategy with no runtime direction branching.
-     */
-    private static final ApplyStrategy FORWARD_APPLY_STRATEGY = PatchCommandEncoder::applyForward;
-
    /**
     * Serialized opcode for deleting one or more characters.
     */
@@ -175,9 +165,9 @@ public final class PatchCommandEncoder {
    private final WordTraversalDirection traversalDirection;

    /**
-     * Direction-specialized patch apply strategy.
+     * Whether this instance applies patch commands in backward traversal order.
     */
-    private final ApplyStrategy applyStrategy;
+    private final boolean backwardTraversal;

    /**
     * Currently allocated source dimension of reusable matrices.
@@ -222,21 +212,6 @@ public final class PatchCommandEncoder {
        MATCH
    }

-    /**
-     * Direction-specialized patch application strategy.
-     */
-    @FunctionalInterface
-    private interface ApplyStrategy {
-        /**
-         * Applies the command.
-         * 
-         * @param source       original text
-         * @param patchCommand patch command
-         * @return final text after applying the command
-         */
-        String apply(String source, String patchCommand);
-    }
-
    private PatchCommandEncoder(final Builder builder) {
        this.traversalDirection = Objects.requireNonNull(builder.traversalDirection, "traversalDirection");
        final int insertCost = builder.insertCost;
@@ -260,7 +235,7 @@ public final class PatchCommandEncoder {
        this.deleteCost = deleteCost;
        this.replaceCost = replaceCost;
        this.matchCost = matchCost;
-        this.applyStrategy = applyStrategyFor(this.traversalDirection);
+        this.backwardTraversal = this.traversalDirection == WordTraversalDirection.BACKWARD;
        this.sourceCapacity = 0;
        this.targetCapacity = 0;
        this.costMatrix = new int[0][0];
@@ -304,19 +279,43 @@ public final class PatchCommandEncoder {
     * direction.
     *
     * <p>
-     * This is the branch-free instance-level fast path for repeated patch
-     * application in a known traversal direction.
+     * This is the instance-level fast path for repeated patch application in a
+     * known traversal direction. It avoids the static API null and direction
+     * validation path and calls the selected decoder directly.
     * </p>
     *
     * @param source       original source word
     * @param patchCommand compact patch command
     * @return transformed word, or {@code null} when {@code source} is {@code null}
+     * @deprecated Since 2.3.0. Runtime stemming should compile
+     *             {@code patchCommand} once through {@link #compile(String)} and
+     *             reuse {@link CompiledPatchCommand#apply(String)}. The
+     *             String-based application path reparses the patch command on every
+     *             call and is kept only for source compatibility before the 3.0.0
+     *             migration.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public String applyWithConfiguredDirection(final String source, final String patchCommand) {
        if (source == null) {
            return null;
        }
-        return this.applyStrategy.apply(source, patchCommand);
+        if (this.backwardTraversal) {
+            return applyBackwardNonNull(source, patchCommand);
+        }
+        return applyForwardNonNull(source, patchCommand);
+    }
+
+    /**
+     * Compiles a patch command for repeated application with this encoder
+     * instance traversal direction.
+     *
+     * @param patchCommand compact patch command
+     * @return immutable compiled patch command
+     * @throws IllegalArgumentException if the serialized command contains an
+     *                                  unsupported opcode or invalid NOOP argument
+     */
+    public CompiledPatchCommand compile(final String patchCommand) {
+        return CompiledPatchCommand.compile(patchCommand, this.traversalDirection);
    }

    /**
@@ -326,7 +325,14 @@ public final class PatchCommandEncoder {
     * @param source       original source word
     * @param patchCommand compact patch command
     * @return transformed word, or {@code null} when {@code source} is {@code null}
+     * @deprecated Since 2.3.0. Runtime stemming should use
+     *             {@link CompiledPatchCommand#compile(String, WordTraversalDirection)}
+     *             once and then reuse {@link CompiledPatchCommand#apply(String)}.
+     *             This method repeatedly interprets the serialized patch-command
+     *             string and is retained only for compatibility before the 3.0.0
+     *             migration.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static String apply(final String source, final String patchCommand) {
        return apply(source, patchCommand, WordTraversalDirection.BACKWARD);
    }
@@ -343,14 +349,41 @@ public final class PatchCommandEncoder {
     * @param patchCommand       compact patch command
     * @param traversalDirection traversal direction used by the patch command
     * @return transformed word, or {@code null} when {@code source} is {@code null}
+     * @deprecated Since 2.3.0. Runtime stemming should use
+     *             {@link CompiledPatchCommand#compile(String, WordTraversalDirection)}
+     *             once and then reuse {@link CompiledPatchCommand#apply(String)}.
+     *             This method repeatedly interprets the serialized patch-command
+     *             string and is retained only for compatibility before the 3.0.0
+     *             migration.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static String apply(final String source, final String patchCommand,
            final WordTraversalDirection traversalDirection) {
        Objects.requireNonNull(traversalDirection, "traversalDirection");
        if (source == null) {
            return null;
        }
-        return applyStrategyFor(traversalDirection).apply(source, patchCommand);
+        if (traversalDirection == WordTraversalDirection.BACKWARD) {
+            return applyBackwardNonNull(source, patchCommand);
+        }
+        return applyForwardNonNull(source, patchCommand);
+    }
+
+    /**
+     * Compiles a patch command for repeated application with the supplied
+     * traversal direction.
+     *
+     * @param patchCommand       compact patch command
+     * @param traversalDirection traversal direction used by the patch command
+     * @return immutable compiled patch command
+     * @throws NullPointerException     if {@code traversalDirection} is
+     *                                  {@code null}
+     * @throws IllegalArgumentException if the serialized command contains an
+     *                                  unsupported opcode or invalid NOOP argument
+     */
+    public static CompiledPatchCommand compile(final String patchCommand,
+            final WordTraversalDirection traversalDirection) {
+        return CompiledPatchCommand.compile(patchCommand, traversalDirection);
    }

    /**
@@ -371,7 +404,15 @@ public final class PatchCommandEncoder {
     * @param outputLength       writable output capacity
     * @return produced character count, or {@link #APPLY_INSUFFICIENT_CAPACITY}
     *         when {@code outputLength} is too small
+     * @deprecated Since 2.3.0. Compile {@code patchCommand} once through
+     *             {@link #compile(String, WordTraversalDirection)} and call
+     *             {@link CompiledPatchCommand#applyTo(CharSequence, char[], int, int)}
+     *             or
+     *             {@link CompiledPatchCommand#applyTo(CharSequence, int, int, char[], int, int)}.
+     *             This String-based method reparses patch commands on every call and
+     *             is kept only for compatibility before the 3.0.0 migration.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static int applyTo(final CharSequence source, final String patchCommand,
            final WordTraversalDirection traversalDirection, final char[] output, final int outputOffset,
            final int outputLength) {
@@ -405,7 +446,13 @@ public final class PatchCommandEncoder {
     *         when {@code outputLength} is too small
     * @throws IllegalArgumentException when source and output ranges overlap in the
     *                                  same array
+     * @deprecated Since 2.3.0. Compile {@code patchCommand} once through
+     *             {@link #compile(String, WordTraversalDirection)} and call
+     *             {@link CompiledPatchCommand#applyTo(char[], int, int, char[], int, int)}.
+     *             This String-based method reparses patch commands on every call and
+     *             is kept only for compatibility before the 3.0.0 migration.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static int applyTo(final char[] source, final int sourceOffset, final int sourceLength,
            final String patchCommand, final WordTraversalDirection traversalDirection, final char[] output,
            final int outputOffset, final int outputLength) {
@@ -484,36 +531,33 @@ public final class PatchCommandEncoder {
    /**
     * Applies a patch command using the historical backward Egothor semantics.
     *
-     * @param source       original source word in legacy backward logical space
+     * @param source       non-null original source word in legacy backward logical
+     *                     space
     * @param patchCommand compact patch command
-     * @return transformed word, or {@code null} when {@code source} is {@code null}
+     * @return transformed word
     */
-    private static String applyBackward(final String source, final String patchCommand) {
-        if (source == null) {
-            return null;
-        }
-        if (patchCommand == null || patchCommand.isEmpty()) {
+    private static String applyBackwardNonNull(final String source, final String patchCommand) {
+        if (patchCommand == null) {
            return source;
        }
-        if (NOOP_PATCH.equals(patchCommand)) {
+        final int patchLength = patchCommand.length();
+        if (patchLength == 0 || (patchLength & 1) != 0) {
            return source;
        }
-        if ((patchCommand.length() & 1) != 0) {
-            return source;
-        }
-        if (patchCommand.length() == 2) {
+        if (patchLength == 2) {
            return applySingleBackwardInstruction(source, patchCommand.charAt(0), patchCommand.charAt(1));
        }

-        final StringBuilder result = new StringBuilder(source);
-        if (result.isEmpty()) {
-            return applyBackwardToEmptySource(result, patchCommand);
+        if (source.isEmpty()) {
+            return applyBackwardToEmptySource(patchCommand);
        }

+        final StringBuilder result = new StringBuilder(source);
+
        int position = result.length() - 1;

        try {
-            for (int patchIndex = 0, patchLength = patchCommand.length(); patchIndex < patchLength; patchIndex += 2) {
+            for (int patchIndex = 0; patchIndex < patchLength; patchIndex += 2) {
                final char opcode = patchCommand.charAt(patchIndex);
                final char argument = patchCommand.charAt(patchIndex + 1);

@@ -547,12 +591,12 @@ public final class PatchCommandEncoder {

                    case NOOP_OPCODE:
                        if (argument != NOOP_ARGUMENT) {
-                            throw new IllegalArgumentException("Unsupported NOOP patch argument: " + argument);
+                            throw new IllegalArgumentException(MSG_NOOP + argument);
                        }
                        return source;

                    default:
-                        throw new IllegalArgumentException("Unsupported patch opcode: " + opcode);
+                        throw new IllegalArgumentException(MSG_OPCODE + opcode);
                }

                position--;
@@ -567,36 +611,32 @@ public final class PatchCommandEncoder {
    /**
     * Applies a patch command using forward traversal semantics.
     *
-     * @param source       original source word
+     * @param source       non-null original source word
     * @param patchCommand compact patch command
-     * @return transformed word, or {@code null} when {@code source} is {@code null}
+     * @return transformed word
     */
-    private static String applyForward(final String source, final String patchCommand) {
-        if (source == null) {
-            return null;
-        }
-        if (patchCommand == null || patchCommand.isEmpty()) {
+    private static String applyForwardNonNull(final String source, final String patchCommand) {
+        if (patchCommand == null) {
            return source;
        }
-        if (NOOP_PATCH.equals(patchCommand)) {
+        final int patchLength = patchCommand.length();
+        if (patchLength == 0 || (patchLength & 1) != 0) {
            return source;
        }
-        if ((patchCommand.length() & 1) != 0) {
-            return source;
-        }
-        if (patchCommand.length() == 2) {
+        if (patchLength == 2) {
            return applySingleForwardInstruction(source, patchCommand.charAt(0), patchCommand.charAt(1));
        }

-        final StringBuilder result = new StringBuilder(source);
-        if (result.isEmpty()) {
-            return applyForwardToEmptySource(result, patchCommand);
+        if (source.isEmpty()) {
+            return applyForwardToEmptySource(patchCommand);
        }

+        final StringBuilder result = new StringBuilder(source);
+
        int position = 0;

        try {
-            for (int patchIndex = 0, patchLength = patchCommand.length(); patchIndex < patchLength; patchIndex += 2) {
+            for (int patchIndex = 0; patchIndex < patchLength; patchIndex += 2) {
                final char opcode = patchCommand.charAt(patchIndex);
                final char argument = patchCommand.charAt(patchIndex + 1);

@@ -628,12 +668,12 @@ public final class PatchCommandEncoder {

                    case NOOP_OPCODE:
                        if (argument != NOOP_ARGUMENT) {
-                            throw new IllegalArgumentException("Unsupported NOOP patch argument: " + argument);
+                            throw new IllegalArgumentException(MSG_NOOP + argument);
                        }
                        return source;

                    default:
-                        throw new IllegalArgumentException("Unsupported patch opcode: " + opcode);
+                        throw new IllegalArgumentException(MSG_OPCODE + opcode);
                }

                position++;
@@ -751,12 +791,12 @@ public final class PatchCommandEncoder {
     * behavior for index-invalid commands.
     * </p>
     *
-     * @param result       empty result builder
     * @param patchCommand compact patch command
     * @return transformed word, or the original empty word when the patch is
     *         malformed
     */
-    private static String applyBackwardToEmptySource(final StringBuilder result, final String patchCommand) {
+    private static String applyBackwardToEmptySource(final String patchCommand) {
+        final StringBuilder result = new StringBuilder(patchCommand.length() >> 1);
        try {
            for (int patchIndex = 0, patchLength = patchCommand.length(); patchIndex < patchLength; patchIndex += 2) {
                final char opcode = patchCommand.charAt(patchIndex);
@@ -774,12 +814,12 @@ public final class PatchCommandEncoder {

                    case NOOP_OPCODE:
                        if (argument != NOOP_ARGUMENT) {
-                            throw new IllegalArgumentException("Unsupported NOOP patch argument: " + argument);
+                            throw new IllegalArgumentException(MSG_NOOP + argument);
                        }
                        return "";

                    default:
-                        throw new IllegalArgumentException("Unsupported patch opcode: " + opcode);
+                        throw new IllegalArgumentException(MSG_OPCODE + opcode);
                }
            }
        } catch (IndexOutOfBoundsException exception) {
@@ -792,12 +832,12 @@ public final class PatchCommandEncoder {
    /**
     * Applies a forward patch command to an empty source word.
     *
-     * @param result       empty result builder
     * @param patchCommand compact patch command
     * @return transformed word, or the original empty word when the patch is
     *         malformed
     */
-    private static String applyForwardToEmptySource(final StringBuilder result, final String patchCommand) {
+    private static String applyForwardToEmptySource(final String patchCommand) {
+        final StringBuilder result = new StringBuilder(patchCommand.length() >> 1);
        try {
            for (int patchIndex = 0, patchLength = patchCommand.length(); patchIndex < patchLength; patchIndex += 2) {
                final char opcode = patchCommand.charAt(patchIndex);
@@ -815,12 +855,12 @@ public final class PatchCommandEncoder {

                    case NOOP_OPCODE:
                        if (argument != NOOP_ARGUMENT) {
-                            throw new IllegalArgumentException("Unsupported NOOP patch argument: " + argument);
+                            throw new IllegalArgumentException(MSG_NOOP + argument);
                        }
                        return "";

                    default:
-                        throw new IllegalArgumentException("Unsupported patch opcode: " + opcode);
+                        throw new IllegalArgumentException(MSG_OPCODE + opcode);
                }
            }
        } catch (IndexOutOfBoundsException exception) {
@@ -1535,16 +1575,6 @@ public final class PatchCommandEncoder {
        }
    }

-    /**
-     * Returns the direction-specialized apply strategy.
-     *
-     * @param traversalDirection requested traversal direction
-     * @return branch-free apply strategy for that direction
-     */
-    private static ApplyStrategy applyStrategyFor(final WordTraversalDirection traversalDirection) {
-        return traversalDirection == WordTraversalDirection.BACKWARD ? BACKWARD_APPLY_STRATEGY : FORWARD_APPLY_STRATEGY;
-    }
-
    /**
     * Decodes a compact count argument used by skip and delete instructions.
     *
--- a/src/main/java/org/egothor/stemmer/ReductionSettings.java
+++ b/src/main/java/org/egothor/stemmer/ReductionSettings.java
@@ -42,10 +42,13 @@ import java.util.Objects;
 * @param reductionMode                 reduction mode
 * @param dominantWinnerMinPercent      minimum dominant winner percentage
 * @param dominantWinnerOverSecondRatio minimum winner-over-second ratio
+ * @param contractUniformSubtrees       whether compilation may contract a subtree
+ *                                      whose reachable terminal values all contain
+ *                                      the same single value
 */
@SuppressWarnings("PMD.LongVariable")
 public record ReductionSettings(ReductionMode reductionMode, int dominantWinnerMinPercent,
-        int dominantWinnerOverSecondRatio) {
+        int dominantWinnerOverSecondRatio, boolean contractUniformSubtrees) {

    /**
     * Default minimum dominant winner percentage.
@@ -65,12 +68,14 @@ public record ReductionSettings(ReductionMode reductionMode, int dominantWinnerM
     *                                      the inclusive range {@code 1..100}
     * @param dominantWinnerOverSecondRatio minimum winner-over-second ratio, must
     *                                      be at least {@code 1}
+     * @param contractUniformSubtrees       whether uniform subtrees may be
+     *                                      contracted into accepting leaves
     * @throws NullPointerException     if {@code reductionMode} is {@code null}
     * @throws IllegalArgumentException if any numeric value is outside the valid
     *                                  range
     */
    public ReductionSettings(final ReductionMode reductionMode, final int dominantWinnerMinPercent,
-            final int dominantWinnerOverSecondRatio) {
+            final int dominantWinnerOverSecondRatio, final boolean contractUniformSubtrees) {
        this.reductionMode = Objects.requireNonNull(reductionMode, "reductionMode");
        if (dominantWinnerMinPercent < 1 || dominantWinnerMinPercent > 100) {
            throw new IllegalArgumentException("dominantWinnerMinPercent must be in range 1..100.");
@@ -80,6 +85,19 @@ public record ReductionSettings(ReductionMode reductionMode, int dominantWinnerM
        }
        this.dominantWinnerMinPercent = dominantWinnerMinPercent;
        this.dominantWinnerOverSecondRatio = dominantWinnerOverSecondRatio;
+        this.contractUniformSubtrees = contractUniformSubtrees;
+    }
+
+    /**
+     * Creates a new instance without uniform-subtree contraction.
+     *
+     * @param reductionMode                 reduction mode
+     * @param dominantWinnerMinPercent      minimum dominant winner percentage
+     * @param dominantWinnerOverSecondRatio minimum winner-over-second ratio
+     */
+    public ReductionSettings(final ReductionMode reductionMode, final int dominantWinnerMinPercent,
+            final int dominantWinnerOverSecondRatio) {
+        this(reductionMode, dominantWinnerMinPercent, dominantWinnerOverSecondRatio, false);
    }

    /**
@@ -93,4 +111,23 @@ public record ReductionSettings(ReductionMode reductionMode, int dominantWinnerM
        return new ReductionSettings(reductionMode, DEFAULT_DOMINANT_WINNER_MIN_PERCENT,
                DEFAULT_DOMINANT_WINNER_OVER_SECOND_RATIO);
    }
+
+    /**
+     * Returns settings that run uniform-subtree contraction before the configured
+     * subtree-merging mode.
+     *
+     * <p>
+     * This is intended for Radixor patch-command tries, where a contracted accepting
+     * leaf can safely represent a subtree whose reachable entries all use the same
+     * patch command.
+     * </p>
+     *
+     * @param settings base settings
+     * @return equivalent settings with uniform-subtree contraction enabled
+     */
+    /* default */ static ReductionSettings withUniformSubtreeContraction(final ReductionSettings settings) {
+        Objects.requireNonNull(settings, "settings");
+        return new ReductionSettings(settings.reductionMode(), settings.dominantWinnerMinPercent(),
+                settings.dominantWinnerOverSecondRatio(), true);
+    }
 }
--- a/src/main/java/org/egothor/stemmer/StemmerKnowledgeExperiment.java
+++ b/src/main/java/org/egothor/stemmer/StemmerKnowledgeExperiment.java
@@ -349,6 +349,7 @@ public final class StemmerKnowledgeExperiment {
     * @param trie         compiled trie under test
     * @return immutable counts for this single input
     */
+    @SuppressWarnings("deprecation")
    private static EvaluationCounts evaluateInput(final String input, final String expectedStem,
            final FrequencyTrie<String> trie) {
        long getCorrect = 0L;
--- a/src/main/java/org/egothor/stemmer/StemmerPatchTrieLoader.java
+++ b/src/main/java/org/egothor/stemmer/StemmerPatchTrieLoader.java
@@ -39,6 +39,8 @@ import java.io.PushbackInputStream;
 import java.nio.charset.StandardCharsets;
 import java.nio.file.Files;
 import java.nio.file.Path;
+import java.util.HashMap;
+import java.util.Map;
 import java.util.Objects;
 import java.util.logging.Level;
 import java.util.logging.Logger;
@@ -68,6 +70,7 @@ import java.util.zip.GZIPInputStream;
 * items containing Unicode whitespace characters while reporting them through
 * aggregated warning log records.
 */
+@SuppressWarnings({ "PMD.ExcessivePublicCount", "PMD.TooManyMethods" })
 public final class StemmerPatchTrieLoader {

    /* default */ static final String FILENAME_REQUIRED = "fileName required";
@@ -293,7 +296,12 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the dictionary cannot be found or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(Language, boolean, ReductionSettings)} so
+     *             patch commands are represented as {@link CompiledPatchCommand}
+     *             values instead of reparsed {@link String} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final Language language, final boolean storeOriginal,
            final ReductionSettings reductionSettings) throws IOException {
        Objects.requireNonNull(language, "language");
@@ -303,6 +311,30 @@ public final class StemmerPatchTrieLoader {
        return load(language, storeOriginal, metadata);
    }

+    /**
+     * Loads a bundled dictionary and returns a runtime-specialized trie whose
+     * values are compiled patch commands.
+     *
+     * <p>
+     * The text dictionary is still compiled through the canonical serialized
+     * patch-command representation. The returned trie replaces each stored
+     * serialized patch command with a {@link CompiledPatchCommand} so repeated
+     * runtime stemming does not parse patch-command strings.
+     * </p>
+     *
+     * @param language          bundled language dictionary
+     * @param storeOriginal     whether the stem itself should be inserted using the
+     *                          canonical no-op patch command
+     * @param reductionSettings reduction settings
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the dictionary cannot be found or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final Language language,
+            final boolean storeOriginal, final ReductionSettings reductionSettings) throws IOException {
+        return compilePatchTrie(load(language, storeOriginal, reductionSettings));
+    }
+
    /**
     * Loads a bundled dictionary using explicit trie compilation metadata.
     *
@@ -320,7 +352,11 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the dictionary cannot be found or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(Language, boolean, TrieMetadata)} so patch
+     *             commands are represented as {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final Language language, final boolean storeOriginal,
            final TrieMetadata metadata) throws IOException {
        Objects.requireNonNull(language, "language");
@@ -335,6 +371,23 @@ public final class StemmerPatchTrieLoader {
        }
    }

+    /**
+     * Loads a bundled dictionary using explicit trie compilation metadata and
+     * returns a runtime-specialized trie whose values are compiled patch commands.
+     *
+     * @param language      bundled language dictionary
+     * @param storeOriginal whether the stem itself should be inserted using the
+     *                      canonical no-op patch command
+     * @param metadata      trie metadata describing the compilation configuration
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the dictionary cannot be found or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final Language language,
+            final boolean storeOriginal, final TrieMetadata metadata) throws IOException {
+        return compilePatchTrie(load(language, storeOriginal, metadata));
+    }
+
    /**
     * Loads a bundled dictionary using default settings for the supplied reduction
     * mode.
@@ -354,13 +407,35 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the dictionary cannot be found or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(Language, boolean, ReductionMode)} so patch
+     *             commands are represented as {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final Language language, final boolean storeOriginal,
            final ReductionMode reductionMode) throws IOException {
        Objects.requireNonNull(reductionMode, "reductionMode");
        return load(language, storeOriginal, ReductionSettings.withDefaults(reductionMode));
    }

+    /**
+     * Loads a bundled dictionary using default settings for the supplied reduction
+     * mode and returns a runtime-specialized trie whose values are compiled patch
+     * commands.
+     *
+     * @param language      bundled language dictionary
+     * @param storeOriginal whether the stem itself should be inserted using the
+     *                      canonical no-op patch command
+     * @param reductionMode reduction mode
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the dictionary cannot be found or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final Language language,
+            final boolean storeOriginal, final ReductionMode reductionMode) throws IOException {
+        return compilePatchTrie(load(language, storeOriginal, reductionMode));
+    }
+
    /**
     * Loads a dictionary from a filesystem path using explicit reduction settings.
     *
@@ -379,13 +454,35 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the file cannot be opened or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(Path, boolean, ReductionSettings)} so patch
+     *             commands are represented as {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final Path path, final boolean storeOriginal,
            final ReductionSettings reductionSettings) throws IOException {
        return load(path, storeOriginal, reductionSettings, WordTraversalDirection.BACKWARD,
                CaseProcessingMode.LOWERCASE_WITH_LOCALE_ROOT, DiacriticProcessingMode.AS_IS);
    }

+    /**
+     * Loads a dictionary from a filesystem path using explicit reduction settings
+     * and returns a runtime-specialized trie whose values are compiled patch
+     * commands.
+     *
+     * @param path              path to the dictionary file
+     * @param storeOriginal     whether the stem itself should be inserted using the
+     *                          canonical no-op patch command
+     * @param reductionSettings reduction settings
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the file cannot be opened or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final Path path,
+            final boolean storeOriginal, final ReductionSettings reductionSettings) throws IOException {
+        return compilePatchTrie(load(path, storeOriginal, reductionSettings));
+    }
+
    /**
     * Loads a dictionary from a filesystem path using explicit reduction settings
     * and explicit traversal direction.
@@ -405,7 +502,12 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the file cannot be opened or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(Path, boolean, ReductionSettings, WordTraversalDirection)}
+     *             so patch commands are represented as
+     *             {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final Path path, final boolean storeOriginal,
            final ReductionSettings reductionSettings, final WordTraversalDirection traversalDirection)
            throws IOException {
@@ -413,6 +515,26 @@ public final class StemmerPatchTrieLoader {
                CaseProcessingMode.LOWERCASE_WITH_LOCALE_ROOT, DiacriticProcessingMode.AS_IS);
    }

+    /**
+     * Loads a dictionary from a filesystem path using explicit reduction settings
+     * and traversal direction, returning runtime-specialized compiled patch values.
+     *
+     * @param path               path to the dictionary file
+     * @param storeOriginal      whether the stem itself should be inserted using
+     *                           the canonical no-op patch command
+     * @param reductionSettings  reduction settings
+     * @param traversalDirection traversal direction used for both trie keys and
+     *                           patch commands
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the file cannot be opened or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final Path path,
+            final boolean storeOriginal, final ReductionSettings reductionSettings,
+            final WordTraversalDirection traversalDirection) throws IOException {
+        return compilePatchTrie(load(path, storeOriginal, reductionSettings, traversalDirection));
+    }
+
    /**
     * Loads a dictionary from a filesystem path using explicit reduction settings,
     * explicit traversal direction, and explicit case processing mode.
@@ -432,7 +554,12 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the file cannot be opened or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(Path, boolean, ReductionSettings, WordTraversalDirection, CaseProcessingMode)}
+     *             so patch commands are represented as
+     *             {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final Path path, final boolean storeOriginal,
            final ReductionSettings reductionSettings, final WordTraversalDirection traversalDirection,
            final CaseProcessingMode caseProcessingMode) throws IOException {
@@ -440,6 +567,29 @@ public final class StemmerPatchTrieLoader {
                DiacriticProcessingMode.AS_IS);
    }

+    /**
+     * Loads a dictionary from a filesystem path using explicit reduction settings,
+     * traversal direction, and case processing mode, returning runtime-specialized
+     * compiled patch values.
+     *
+     * @param path               path to the dictionary file
+     * @param storeOriginal      whether the stem itself should be inserted using
+     *                           the canonical no-op patch command
+     * @param reductionSettings  reduction settings
+     * @param traversalDirection traversal direction used for both trie keys and
+     *                           patch commands
+     * @param caseProcessingMode case processing mode used during dictionary parsing
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the file cannot be opened or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final Path path,
+            final boolean storeOriginal, final ReductionSettings reductionSettings,
+            final WordTraversalDirection traversalDirection, final CaseProcessingMode caseProcessingMode)
+            throws IOException {
+        return compilePatchTrie(load(path, storeOriginal, reductionSettings, traversalDirection, caseProcessingMode));
+    }
+
    /**
     * Loads a dictionary from a filesystem path using explicit reduction settings,
     * traversal direction, case processing mode, and diacritic processing mode.
@@ -457,7 +607,12 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the file cannot be opened or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(Path, boolean, ReductionSettings, WordTraversalDirection, CaseProcessingMode, DiacriticProcessingMode)}
+     *             so patch commands are represented as
+     *             {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final Path path, final boolean storeOriginal,
            final ReductionSettings reductionSettings, final WordTraversalDirection traversalDirection,
            final CaseProcessingMode caseProcessingMode, final DiacriticProcessingMode diacriticProcessingMode)
@@ -468,6 +623,32 @@ public final class StemmerPatchTrieLoader {
        return load(path, storeOriginal, metadata);
    }

+    /**
+     * Loads a dictionary from a filesystem path using explicit semantic metadata
+     * dimensions, returning runtime-specialized compiled patch values.
+     *
+     * @param path                    path to the dictionary file
+     * @param storeOriginal           whether the stem itself should be inserted
+     *                                using the canonical no-op patch command
+     * @param reductionSettings       reduction settings
+     * @param traversalDirection      traversal direction used for both trie keys
+     *                                and patch commands
+     * @param caseProcessingMode      case processing mode used during dictionary
+     *                                parsing
+     * @param diacriticProcessingMode diacritic processing mode used during
+     *                                dictionary parsing
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the file cannot be opened or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final Path path,
+            final boolean storeOriginal, final ReductionSettings reductionSettings,
+            final WordTraversalDirection traversalDirection, final CaseProcessingMode caseProcessingMode,
+            final DiacriticProcessingMode diacriticProcessingMode) throws IOException {
+        return compilePatchTrie(load(path, storeOriginal, reductionSettings, traversalDirection, caseProcessingMode,
+                diacriticProcessingMode));
+    }
+
    /**
     * Loads a dictionary from a filesystem path using explicit trie compilation
     * metadata.
@@ -485,7 +666,11 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the file cannot be opened or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(Path, boolean, TrieMetadata)} so patch
+     *             commands are represented as {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final Path path, final boolean storeOriginal, final TrieMetadata metadata)
            throws IOException {
        Objects.requireNonNull(path, PARAMETER_PATH);
@@ -498,6 +683,24 @@ public final class StemmerPatchTrieLoader {
        }
    }

+    /**
+     * Loads a dictionary from a filesystem path using explicit trie compilation
+     * metadata and returns a runtime-specialized trie whose values are compiled
+     * patch commands.
+     *
+     * @param path          path to the dictionary file
+     * @param storeOriginal whether the stem itself should be inserted using the
+     *                      canonical no-op patch command
+     * @param metadata      trie metadata describing the compilation configuration
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the file cannot be opened or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final Path path,
+            final boolean storeOriginal, final TrieMetadata metadata) throws IOException {
+        return compilePatchTrie(load(path, storeOriginal, metadata));
+    }
+
    /**
     * Loads a dictionary from a filesystem path using default settings for the
     * supplied reduction mode.
@@ -518,13 +721,35 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the file cannot be opened or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(Path, boolean, ReductionMode)} so patch
+     *             commands are represented as {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final Path path, final boolean storeOriginal,
            final ReductionMode reductionMode) throws IOException {
        Objects.requireNonNull(reductionMode, "reductionMode");
        return load(path, storeOriginal, ReductionSettings.withDefaults(reductionMode));
    }

+    /**
+     * Loads a dictionary from a filesystem path using default settings for the
+     * supplied reduction mode and returns runtime-specialized compiled patch
+     * values.
+     *
+     * @param path          path to the dictionary file
+     * @param storeOriginal whether the stem itself should be inserted using the
+     *                      canonical no-op patch command
+     * @param reductionMode reduction mode
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the file cannot be opened or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final Path path,
+            final boolean storeOriginal, final ReductionMode reductionMode) throws IOException {
+        return compilePatchTrie(load(path, storeOriginal, reductionMode));
+    }
+
    /**
     * Loads a dictionary from a filesystem path string using explicit reduction
     * settings.
@@ -543,13 +768,36 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the file cannot be opened or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(String, boolean, ReductionSettings)} so
+     *             patch commands are represented as
+     *             {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final String fileName, final boolean storeOriginal,
            final ReductionSettings reductionSettings) throws IOException {
        Objects.requireNonNull(fileName, FILENAME_REQUIRED);
        return load(Path.of(fileName), storeOriginal, reductionSettings);
    }

+    /**
+     * Loads a dictionary from a filesystem path string using explicit reduction
+     * settings and returns runtime-specialized compiled patch values.
+     *
+     * @param fileName          file name or path string
+     * @param storeOriginal     whether the stem itself should be inserted using the
+     *                          canonical no-op patch command
+     * @param reductionSettings reduction settings
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the file cannot be opened or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final String fileName,
+            final boolean storeOriginal, final ReductionSettings reductionSettings) throws IOException {
+        Objects.requireNonNull(fileName, FILENAME_REQUIRED);
+        return loadCompiled(Path.of(fileName), storeOriginal, reductionSettings);
+    }
+
    /**
     * Loads a dictionary from a filesystem path string using explicit reduction
     * settings and explicit traversal direction.
@@ -571,7 +819,12 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the file cannot be opened or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(String, boolean, ReductionSettings, WordTraversalDirection)}
+     *             so patch commands are represented as
+     *             {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final String fileName, final boolean storeOriginal,
            final ReductionSettings reductionSettings, final WordTraversalDirection traversalDirection)
            throws IOException {
@@ -580,6 +833,28 @@ public final class StemmerPatchTrieLoader {
                CaseProcessingMode.LOWERCASE_WITH_LOCALE_ROOT);
    }

+    /**
+     * Loads a dictionary from a filesystem path string using explicit reduction
+     * settings and traversal direction, returning runtime-specialized compiled
+     * patch values.
+     *
+     * @param fileName           file name or path string
+     * @param storeOriginal      whether the stem itself should be inserted using
+     *                           the canonical no-op patch command
+     * @param reductionSettings  reduction settings
+     * @param traversalDirection traversal direction used for both trie keys and
+     *                           patch commands
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the file cannot be opened or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final String fileName,
+            final boolean storeOriginal, final ReductionSettings reductionSettings,
+            final WordTraversalDirection traversalDirection) throws IOException {
+        Objects.requireNonNull(fileName, FILENAME_REQUIRED);
+        return loadCompiled(Path.of(fileName), storeOriginal, reductionSettings, traversalDirection);
+    }
+
    /**
     * Loads a dictionary from a filesystem path string using explicit reduction
     * settings, explicit traversal direction, and explicit case processing mode.
@@ -600,7 +875,12 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the file cannot be opened or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(String, boolean, ReductionSettings, WordTraversalDirection, CaseProcessingMode)}
+     *             so patch commands are represented as
+     *             {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final String fileName, final boolean storeOriginal,
            final ReductionSettings reductionSettings, final WordTraversalDirection traversalDirection,
            final CaseProcessingMode caseProcessingMode) throws IOException {
@@ -609,6 +889,31 @@ public final class StemmerPatchTrieLoader {
                DiacriticProcessingMode.AS_IS);
    }

+    /**
+     * Loads a dictionary from a filesystem path string using explicit reduction
+     * settings, traversal direction, and case processing mode, returning
+     * runtime-specialized compiled patch values.
+     *
+     * @param fileName           file name or path string
+     * @param storeOriginal      whether the stem itself should be inserted using
+     *                           the canonical no-op patch command
+     * @param reductionSettings  reduction settings
+     * @param traversalDirection traversal direction used for both trie keys and
+     *                           patch commands
+     * @param caseProcessingMode case processing mode used during dictionary parsing
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the file cannot be opened or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final String fileName,
+            final boolean storeOriginal, final ReductionSettings reductionSettings,
+            final WordTraversalDirection traversalDirection, final CaseProcessingMode caseProcessingMode)
+            throws IOException {
+        Objects.requireNonNull(fileName, FILENAME_REQUIRED);
+        return loadCompiled(Path.of(fileName), storeOriginal, reductionSettings, traversalDirection,
+                caseProcessingMode);
+    }
+
    /**
     * Loads a dictionary from a filesystem path string using explicit reduction
     * settings, explicit traversal direction, explicit case processing mode, and
@@ -627,7 +932,12 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the file cannot be opened or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(String, boolean, ReductionSettings, WordTraversalDirection, CaseProcessingMode, DiacriticProcessingMode)}
+     *             so patch commands are represented as
+     *             {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final String fileName, final boolean storeOriginal,
            final ReductionSettings reductionSettings, final WordTraversalDirection traversalDirection,
            final CaseProcessingMode caseProcessingMode, final DiacriticProcessingMode diacriticProcessingMode)
@@ -637,6 +947,33 @@ public final class StemmerPatchTrieLoader {
                diacriticProcessingMode);
    }

+    /**
+     * Loads a dictionary from a filesystem path string using explicit semantic
+     * metadata dimensions, returning runtime-specialized compiled patch values.
+     *
+     * @param fileName                file name or path string
+     * @param storeOriginal           whether the stem itself should be inserted
+     *                                using the canonical no-op patch command
+     * @param reductionSettings       reduction settings
+     * @param traversalDirection      traversal direction used for both trie keys
+     *                                and patch commands
+     * @param caseProcessingMode      case processing mode used during dictionary
+     *                                parsing
+     * @param diacriticProcessingMode diacritic processing mode used during
+     *                                dictionary parsing
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the file cannot be opened or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final String fileName,
+            final boolean storeOriginal, final ReductionSettings reductionSettings,
+            final WordTraversalDirection traversalDirection, final CaseProcessingMode caseProcessingMode,
+            final DiacriticProcessingMode diacriticProcessingMode) throws IOException {
+        Objects.requireNonNull(fileName, FILENAME_REQUIRED);
+        return loadCompiled(Path.of(fileName), storeOriginal, reductionSettings, traversalDirection,
+                caseProcessingMode, diacriticProcessingMode);
+    }
+
    /**
     * Loads a dictionary from a filesystem path string using explicit trie
     * compilation metadata.
@@ -652,13 +989,35 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the file cannot be opened or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(String, boolean, TrieMetadata)} so patch
+     *             commands are represented as {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final String fileName, final boolean storeOriginal,
            final TrieMetadata metadata) throws IOException {
        Objects.requireNonNull(fileName, FILENAME_REQUIRED);
        return load(Path.of(fileName), storeOriginal, metadata);
    }

+    /**
+     * Loads a dictionary from a filesystem path string using explicit trie
+     * compilation metadata and returns runtime-specialized compiled patch values.
+     *
+     * @param fileName      file name or path string
+     * @param storeOriginal whether the stem itself should be inserted using the
+     *                      canonical no-op patch command
+     * @param metadata      trie metadata describing the compilation configuration
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the file cannot be opened or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final String fileName,
+            final boolean storeOriginal, final TrieMetadata metadata) throws IOException {
+        Objects.requireNonNull(fileName, FILENAME_REQUIRED);
+        return loadCompiled(Path.of(fileName), storeOriginal, metadata);
+    }
+
    /**
     * Loads a dictionary from a filesystem path string using default settings for
     * the supplied reduction mode.
@@ -677,13 +1036,36 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if any argument is {@code null}
     * @throws IOException          if the file cannot be opened or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadCompiled(String, boolean, ReductionMode)} so patch
+     *             commands are represented as {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> load(final String fileName, final boolean storeOriginal,
            final ReductionMode reductionMode) throws IOException {
        Objects.requireNonNull(fileName, FILENAME_REQUIRED);
        return load(Path.of(fileName), storeOriginal, reductionMode);
    }

+    /**
+     * Loads a dictionary from a filesystem path string using default settings for
+     * the supplied reduction mode and returns runtime-specialized compiled patch
+     * values.
+     *
+     * @param fileName      file name or path string
+     * @param storeOriginal whether the stem itself should be inserted using the
+     *                      canonical no-op patch command
+     * @param reductionMode reduction mode
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if any argument is {@code null}
+     * @throws IOException          if the file cannot be opened or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadCompiled(final String fileName,
+            final boolean storeOriginal, final ReductionMode reductionMode) throws IOException {
+        Objects.requireNonNull(fileName, FILENAME_REQUIRED);
+        return loadCompiled(Path.of(fileName), storeOriginal, reductionMode);
+    }
+
    /**
     * Parses one dictionary and builds the compiled trie.
     *
@@ -736,7 +1118,9 @@ public final class StemmerPatchTrieLoader {
        Objects.requireNonNull(reductionSettings, "reductionSettings");
        Objects.requireNonNull(caseProcessingMode, "caseProcessingMode");
        Objects.requireNonNull(diacriticProcessingMode, "diacriticProcessingMode");
-        return TrieMetadata.forCompilation(traversalDirection, reductionSettings, diacriticProcessingMode,
+        final ReductionSettings patchReductionSettings = ReductionSettings
+                .withUniformSubtreeContraction(reductionSettings);
+        return TrieMetadata.forCompilation(traversalDirection, patchReductionSettings, diacriticProcessingMode,
                caseProcessingMode);
    }

@@ -750,6 +1134,28 @@ public final class StemmerPatchTrieLoader {
        return language.isRightToLeft() ? WordTraversalDirection.FORWARD : WordTraversalDirection.BACKWARD;
    }

+    /**
+     * Maps textual patch commands to runtime-specialized compiled patch commands.
+     *
+     * <p>
+     * Equal textual patch commands are compiled once and shared by all trie values
+     * that reference them. The returned trie preserves the source trie keys,
+     * metadata, traversal direction, counts, and reduction settings.
+     * </p>
+     *
+     * @param trie source trie containing textual patch commands
+     * @return equivalent trie containing compiled patch commands
+     * @throws NullPointerException if {@code trie} is {@code null}
+     */
+    private static FrequencyTrie<CompiledPatchCommand> compilePatchTrie(final FrequencyTrie<String> trie) {
+        final FrequencyTrie<String> sourceTrie = Objects.requireNonNull(trie, "trie");
+        final Map<String, CompiledPatchCommand> compiledPatches = new HashMap<>(4096);
+        return FrequencyTrieBuilders.mapValues(sourceTrie, CompiledPatchCommand[]::new,
+                sourceTrie.metadata().reductionSettings(),
+                patch -> compiledPatches.computeIfAbsent(patch,
+                        value -> CompiledPatchCommand.compile(value, sourceTrie.traversalDirection())));
+    }
+
    /**
     * Loads a GZip-compressed binary patch-command trie from a filesystem path.
     *
@@ -758,12 +1164,30 @@ public final class StemmerPatchTrieLoader {
     * @throws NullPointerException if {@code path} is {@code null}
     * @throws IOException          if the file cannot be opened, decompressed, or
     *                              read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadBinaryCompiled(Path)} so patch commands are
+     *             represented as {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> loadBinary(final Path path) throws IOException {
        Objects.requireNonNull(path, PARAMETER_PATH);
        return StemmerPatchTrieBinaryIO.read(path);
    }

+    /**
+     * Loads a GZip-compressed binary patch-command trie from a filesystem path and
+     * returns runtime-specialized compiled patch values.
+     *
+     * @param path path to the compressed binary trie file
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if {@code path} is {@code null}
+     * @throws IOException          if the file cannot be opened, decompressed, or
+     *                              read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadBinaryCompiled(final Path path) throws IOException {
+        return compilePatchTrie(loadBinary(path));
+    }
+
    /**
     * Loads a GZip-compressed binary patch-command trie from a filesystem path
     * using a custom dense lookup span override.
@@ -779,12 +1203,34 @@ public final class StemmerPatchTrieLoader {
     * @throws NullPointerException if {@code path} is {@code null}
     * @throws IOException          if the file cannot be opened, decompressed, or
     *                              read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadBinaryCompiled(Path, int)} so patch commands are
+     *             represented as {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> loadBinary(final Path path, final int maxExpandedIndex) throws IOException {
        Objects.requireNonNull(path, PARAMETER_PATH);
        return StemmerPatchTrieBinaryIO.read(path, maxExpandedIndex);
    }

+    /**
+     * Loads a GZip-compressed binary patch-command trie from a filesystem path using
+     * a custom dense lookup span override and returns runtime-specialized compiled
+     * patch values.
+     *
+     * @param path             path to the compressed binary trie file
+     * @param maxExpandedIndex dense lookup span override; negative values use
+     *                         {@link FrequencyTrie#DEFAULT_MAX_EXPANDED_INDEX}
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if {@code path} is {@code null}
+     * @throws IOException          if the file cannot be opened, decompressed, or
+     *                              read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadBinaryCompiled(final Path path,
+            final int maxExpandedIndex) throws IOException {
+        return compilePatchTrie(loadBinary(path, maxExpandedIndex));
+    }
+
    /**
     * Loads a GZip-compressed binary patch-command trie from a filesystem path
     * string.
@@ -794,12 +1240,30 @@ public final class StemmerPatchTrieLoader {
     * @throws NullPointerException if {@code fileName} is {@code null}
     * @throws IOException          if the file cannot be opened, decompressed, or
     *                              read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadBinaryCompiled(String)} so patch commands are
+     *             represented as {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> loadBinary(final String fileName) throws IOException {
        Objects.requireNonNull(fileName, FILENAME_REQUIRED);
        return StemmerPatchTrieBinaryIO.read(fileName);
    }

+    /**
+     * Loads a GZip-compressed binary patch-command trie from a filesystem path
+     * string and returns runtime-specialized compiled patch values.
+     *
+     * @param fileName file name or path string
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if {@code fileName} is {@code null}
+     * @throws IOException          if the file cannot be opened, decompressed, or
+     *                              read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadBinaryCompiled(final String fileName) throws IOException {
+        return compilePatchTrie(loadBinary(fileName));
+    }
+
    /**
     * Loads a GZip-compressed binary patch-command trie from a filesystem path
     * string using a custom dense lookup span override.
@@ -815,13 +1279,35 @@ public final class StemmerPatchTrieLoader {
     * @throws NullPointerException if {@code fileName} is {@code null}
     * @throws IOException          if the file cannot be opened, decompressed, or
     *                              read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadBinaryCompiled(String, int)} so patch commands are
+     *             represented as {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> loadBinary(final String fileName, final int maxExpandedIndex)
            throws IOException {
        Objects.requireNonNull(fileName, FILENAME_REQUIRED);
        return StemmerPatchTrieBinaryIO.read(fileName, maxExpandedIndex);
    }

+    /**
+     * Loads a GZip-compressed binary patch-command trie from a filesystem path string
+     * using a custom dense lookup span override and returns runtime-specialized
+     * compiled patch values.
+     *
+     * @param fileName         file name or path string
+     * @param maxExpandedIndex dense lookup span override; negative values use
+     *                         {@link FrequencyTrie#DEFAULT_MAX_EXPANDED_INDEX}
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if {@code fileName} is {@code null}
+     * @throws IOException          if the file cannot be opened, decompressed, or
+     *                              read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadBinaryCompiled(final String fileName,
+            final int maxExpandedIndex) throws IOException {
+        return compilePatchTrie(loadBinary(fileName, maxExpandedIndex));
+    }
+
    /**
     * Loads a GZip-compressed binary patch-command trie from an input stream.
     *
@@ -829,12 +1315,30 @@ public final class StemmerPatchTrieLoader {
     * @return compiled patch-command trie
     * @throws NullPointerException if {@code inputStream} is {@code null}
     * @throws IOException          if the stream cannot be decompressed or read
+     * @deprecated Since 2.3.0 for runtime stemming. Use
+     *             {@link #loadBinaryCompiled(InputStream)} so patch commands are
+     *             represented as {@link CompiledPatchCommand} values.
     */
+    @Deprecated(since = "2.3.0", forRemoval = false)
    public static FrequencyTrie<String> loadBinary(final InputStream inputStream) throws IOException {
        Objects.requireNonNull(inputStream, "inputStream");
        return StemmerPatchTrieBinaryIO.read(inputStream);
    }

+    /**
+     * Loads a GZip-compressed binary patch-command trie from an input stream and
+     * returns runtime-specialized compiled patch values.
+     *
+     * @param inputStream source input stream
+     * @return compiled patch-command trie with runtime-specialized values
+     * @throws NullPointerException if {@code inputStream} is {@code null}
+     * @throws IOException          if the stream cannot be decompressed or read
+     */
+    public static FrequencyTrie<CompiledPatchCommand> loadBinaryCompiled(final InputStream inputStream)
+            throws IOException {
+        return compilePatchTrie(loadBinary(inputStream));
+    }
+
    /**
     * Loads only persisted metadata from a GZip-compressed binary patch-command
     * trie file.
--- a/src/main/java/org/egothor/stemmer/TrieMetadata.java
+++ b/src/main/java/org/egothor/stemmer/TrieMetadata.java
@@ -167,6 +167,8 @@ public record TrieMetadata(int formatVersion, WordTraversalDirection traversalDi
                .append("dominantWinnerOverSecondRatio=").append(this.reductionSettings.dominantWinnerOverSecondRatio())
                .append('\n')
                //
+                .append("contractUniformSubtrees=").append(this.reductionSettings.contractUniformSubtrees()).append('\n')
+                //
                .append("diacriticProcessingMode=").append(this.diacriticProcessingMode.name()).append('\n')
                //
                .append("caseProcessingMode=").append(this.caseProcessingMode.name()).append('\n');
@@ -207,13 +209,16 @@ public record TrieMetadata(int formatVersion, WordTraversalDirection traversalDi
        final int dominantWinnerMinPercent = Integer.parseInt(requireEntry(entries, "dominantWinnerMinPercent"));
        final int dominantWinnerOverSecondRatio = Integer // NOPMD
                .parseInt(requireEntry(entries, "dominantWinnerOverSecondRatio"));
+        final boolean contractUniformSubtrees = Boolean
+                .parseBoolean(entries.getOrDefault("contractUniformSubtrees", "false"));
        final DiacriticProcessingMode diacriticProcessingMode = DiacriticProcessingMode
                .valueOf(requireEntry(entries, "diacriticProcessingMode"));
        final CaseProcessingMode caseProcessingMode = CaseProcessingMode
                .valueOf(requireEntry(entries, "caseProcessingMode"));

        return new TrieMetadata(formatVersion, traversalDirection,
-                new ReductionSettings(reductionMode, dominantWinnerMinPercent, dominantWinnerOverSecondRatio),
+                new ReductionSettings(reductionMode, dominantWinnerMinPercent, dominantWinnerOverSecondRatio,
+                        contractUniformSubtrees),
                diacriticProcessingMode, caseProcessingMode);
    }

--- a/src/main/java/org/egothor/stemmer/trie/CompiledNode.java
+++ b/src/main/java/org/egothor/stemmer/trie/CompiledNode.java
@@ -93,6 +93,11 @@ public final class CompiledNode<V> {
     */
    private final int[] orderedCounts;

+    /**
+     * Whether this node accepts any remaining lookup input.
+     */
+    private final boolean acceptsRemainingInput;
+
    /**
     * Creates one validated compiled node using {@link #DEFAULT_MAX_EXPANDED_INDEX}
     * for dense lookup sizing.
@@ -120,6 +125,23 @@ public final class CompiledNode<V> {
     */
    public CompiledNode(final char[] edgeLabels, final CompiledNode<V>[] children, final V[] orderedValues,
            final int maxExpandedIndex, final int... orderedCounts) {
+        this(edgeLabels, children, orderedValues, false, maxExpandedIndex, orderedCounts);
+    }
+
+    /**
+     * Creates one validated compiled node.
+     *
+     * @param acceptsRemainingInput whether this node accepts any remaining lookup
+     *                              input
+     * @param maxExpandedIndex      upper bound for the dense lookup interval size
+     * @throws NullPointerException     if any array argument is {@code null}
+     * @throws IllegalArgumentException if the edge-related arrays or value-related
+     *                                  arrays do not have matching lengths, the
+     *                                  dense interval size is negative, or an
+     *                                  accepting node has children
+     */
+    public CompiledNode(final char[] edgeLabels, final CompiledNode<V>[] children, final V[] orderedValues,
+            final boolean acceptsRemainingInput, final int maxExpandedIndex, final int... orderedCounts) {
        Objects.requireNonNull(edgeLabels, "edgeLabels");
        Objects.requireNonNull(children, "children");
        Objects.requireNonNull(orderedValues, "orderedValues");
@@ -135,11 +157,18 @@ public final class CompiledNode<V> {
        if (orderedValues.length != orderedCounts.length) {
            throw new IllegalArgumentException("orderedValues and orderedCounts must have the same length.");
        }
+        if (acceptsRemainingInput && edgeLabels.length != 0) {
+            throw new IllegalArgumentException("Accepting nodes cannot have child edges.");
+        }
+        if (acceptsRemainingInput && orderedValues.length == 0) {
+            throw new IllegalArgumentException("Accepting nodes must store at least one value.");
+        }

        this.edgeLabels = edgeLabels;
        this.children = children;
        this.orderedValues = orderedValues;
        this.orderedCounts = orderedCounts;
+        this.acceptsRemainingInput = acceptsRemainingInput;

        if (edgeLabels.length == 0 || maxExpandedIndex == 0) {
            this.denseChildren = null;
@@ -158,7 +187,7 @@ public final class CompiledNode<V> {
        }

        @SuppressWarnings("unchecked")
-        final CompiledNode<V>[] dense = (CompiledNode<V>[]) new CompiledNode[span + 1];
+        final CompiledNode<V>[] dense = new CompiledNode[span + 1];
        for (int edgeIndex = 0; edgeIndex < edgeLabels.length; edgeIndex++) {
            dense[edgeLabels[edgeIndex] - minEdge] = children[edgeIndex];
        }
@@ -268,6 +297,15 @@ public final class CompiledNode<V> {
        return !hasChildren();
    }

+    /**
+     * Indicates whether this node accepts any remaining lookup input.
+     *
+     * @return {@code true} for a contracted accepting leaf
+     */
+    public boolean acceptsRemainingInput() {
+        return this.acceptsRemainingInput;
+    }
+
    /**
     * Tests whether an edge label is present at this node.
     *
@@ -310,6 +348,7 @@ public final class CompiledNode<V> {
        hash = 31 * hash + Arrays.hashCode(this.orderedValues);
        hash = 31 * hash + Arrays.hashCode(this.orderedCounts);
        hash = 31 * hash + Objects.hash(this.denseEdgeMin);
+        hash = 31 * hash + Boolean.hashCode(this.acceptsRemainingInput);
        hash = 31 * hash + (hasDenseLookup() ? Arrays.hashCode(this.denseChildren) : 0);
        return hash;
    }
@@ -331,6 +370,7 @@ public final class CompiledNode<V> {
        return Arrays.equals(this.edgeLabels, other.edgeLabels) && Arrays.equals(this.children, other.children)
                && Arrays.equals(this.orderedValues, other.orderedValues)
                && Arrays.equals(this.orderedCounts, other.orderedCounts) && this.denseEdgeMin == other.denseEdgeMin
+                && this.acceptsRemainingInput == other.acceptsRemainingInput
                && Arrays.equals(this.denseChildren, other.denseChildren);
    }

@@ -342,12 +382,13 @@ public final class CompiledNode<V> {
    @Override
    public String toString() {
        return "CompiledNode{" + "edgeCount=" + this.edgeLabels.length + ", orderedValueCount="
-                + this.orderedValues.length + ", denseTableLength=" + denseTableLength() + '}';
+                + this.orderedValues.length + ", acceptsRemainingInput=" + this.acceptsRemainingInput
+                + ", denseTableLength=" + denseTableLength() + '}';
    }

    /**
     * Finds a child for the supplied edge character.
-     * <p>
+     * 
     * Lookup order is:
     * <ol>
     * <li>dense array index (if the label interval is compact enough),</li>
@@ -355,7 +396,6 @@ public final class CompiledNode<V> {
     * {@value #LINEAR_CHILD_COUNT_THRESHOLD} or fewer edges,</li>
     * <li>binary search over sorted labels.</li>
     * </ol>
-     * </p>
     *
     * @param edge edge character
     * @return child node, or {@code null} if absent
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Leo Galambos	05f3855b99	feat(benchmarks): expand multilingual stemming quality evaluation * cover all Radixor dictionary languages * add PRIMARY_OUTPUT, ANY_CANDIDATE, and ALL_CANDIDATES policies * measure pairwise over-stemming and under-stemming * add balanced accuracy and complementary quality metrics * compare single-output and multi-output stemmers fairly * improve result validation, reporting, and documentation * move stemming quality tests into the standard test source set * preserve the existing JMH benchmark structure and badge output	2026-07-20 23:20:17 +02:00
Leo Galambos	6d35f01303	fix: retain only the 10 latest GitHub Pages builds	2026-07-20 01:37:17 +02:00
Leo Galambos	049f44e697	Add CISTEM and Hunspell benchmarks and refresh results	2026-07-06 01:51:33 +02:00
Leo Galambos	a52e82933f	feat: hunspell benchmarks	2026-07-04 22:38:58 +02:00
Leo Galambos	5a65de21d9	fix: workflow/benchmark/jmh exceeded the maximum execution time of 30m	2026-07-03 20:03:58 +02:00
Leo Galambos	3ce9cbc84f	chore: update Gradle dependency verification metadata	2026-07-03 18:55:36 +02:00
Leo Galambos	38620d7e71	feat: prepare Radixor 3.0.0 with contracted tries and compiled patch commands Introduce contracted compiled patch tries for faster lookup, make compiled patch commands the primary runtime path, refresh stemmer benchmarks and documentation, and restructure the documentation for 3.0.0 onboarding. BREAKING CHANGE: Radixor 3.0.0 promotes compiled patch-command APIs and new compiled trie artifacts as the primary runtime integration model.	2026-07-03 18:44:39 +02:00
Leo Galambos	df4552b113	fix(javadoc): Remove collision block	2026-05-24 20:14:28 +02:00
Leo Galambos	9a84add263	fix(pmd): PMD errors fixed	2026-05-24 20:09:11 +02:00
Leo Galambos	1a02c41348	feat: Add FrequencyTrie model fingerprints for EGOTHOR v4 analyzer identity The fingerprint covers trie metadata and the compiled node graph, exposes a lowercase hex representation plus defensive raw bytes, and is stable across equivalent trie builds and persistence round-trips.	2026-05-24 20:05:03 +02:00
Leo Galambos	464b580436	feat: Add FrequencyTrie model fingerprints for EGOTHOR v4 analyzer identity The fingerprint covers trie metadata and the compiled node graph, exposes a lowercase hex representation plus defensive raw bytes, and is stable across equivalent trie builds and persistence round-trips.	2026-05-24 19:54:29 +02:00
Leo Galambos	b945902f05	feat: Add JPMS module descriptor for org.egothor.radixor	2026-05-17 17:19:09 +02:00
Leo Galambos	902ad117e8	fix(docs): new header/banner in README	2026-05-17 16:14:23 +02:00
				`@@ -0,0 +1 @@`
				`5a93a6ab60e46489737cd649eb1ac48182114b9038f7f20195ab9d1c1fc0dd28 stemming-quality.csv`