Egothor/Radixor
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
Files
62be4c9127f3c6cba469393eb7d67c4476d81fe1
Radixor/docs/benchmarks/languages/italian.md
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

26 KiB
Raw Blame History

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. 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 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

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.

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

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
Classification metrics
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
Pair-relation metrics
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
Partition metrics (PRIMARY_OUTPUT only)
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
Raw pair counts
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

ANY_CANDIDATE ranking

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
Classification metrics
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
Pair-relation metrics
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
Partition metrics (PRIMARY_OUTPUT only)
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
Raw pair counts
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

ALL_CANDIDATES ranking

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
Classification metrics
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
Pair-relation metrics
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
Partition metrics (PRIMARY_OUTPUT only)
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
Raw pair counts
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

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

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
Classification metrics
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
Pair-relation metrics
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
Partition metrics (PRIMARY_OUTPUT only)
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
Raw pair counts
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

ANY_CANDIDATE ranking

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
Classification metrics
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
Pair-relation metrics
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
Partition metrics (PRIMARY_OUTPUT only)
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
Raw pair counts
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

ALL_CANDIDATES ranking

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
Classification metrics
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
Pair-relation metrics
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
Partition metrics (PRIMARY_OUTPUT only)
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
Raw pair counts
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

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
Reference in New Issue View Git Blame Copy Permalink