README.md

metadata

language: br
language_name: BR
language_family: celtic_brythonic
tags:
  - wikilangs
  - nlp
  - tokenizer
  - embeddings
  - n-gram
  - markov
  - wikipedia
  - monolingual
  - family-celtic_brythonic
license: mit
library_name: wikilangs
pipeline_tag: feature-extraction
datasets:
  - omarkamali/wikipedia-monthly
dataset_info:
  name: wikipedia-monthly
  description: Monthly snapshots of Wikipedia articles across 300+ languages
metrics:
  - name: best_compression_ratio
    type: compression
    value: 3.786
  - name: best_isotropy
    type: isotropy
    value: 0.8171
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-03T00:00:00.000Z

BR - Wikilangs Models

Comprehensive Research Report & Full Ablation Study

This repository contains NLP models trained and evaluated by Wikilangs, specifically on BR Wikipedia data. We analyze tokenizers, n-gram models, Markov chains, vocabulary statistics, and word embeddings.

📋 Repository Contents

Models & Assets

• Tokenizers (8k, 16k, 32k, 64k)
• N-gram models (2, 3, 4, 5-gram)
• Markov chains (context of 1, 2, 3, 4 and 5)
• Subword N-gram and Markov chains
• Embeddings in various sizes and dimensions (aligned and unaligned)
• Language Vocabulary
• Language Statistics

Analysis and Evaluation

• 1. Tokenizer Evaluation
• 2. N-gram Model Evaluation
• 3. Markov Chain Evaluation
• 4. Vocabulary Analysis
• 5. Word Embeddings Evaluation
• 6. Morphological Analysis (Experimental)
• 7. Summary & Recommendations
• Metrics Glossary
• Visualizations Index

---

1. Tokenizer Evaluation

Results

Vocab Size	Compression	Avg Token Len	UNK Rate	Total Tokens
8k	3.235x	3.24	0.4490%	793,680
16k	3.460x	3.46	0.4803%	742,059
32k	3.645x	3.65	0.5060%	704,331
64k	3.786x 🏆	3.79	0.5255%	678,179

Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

Sample 1: Monsano zo ur gumun italian e proviñs Ancona, er Marche. Marche Proviñs Ancona

Vocab	Tokens	Count
8k	▁mon s ano ▁zo ▁ur ▁gumun ▁italian ▁e ▁proviñs ▁anc ... (+9 more)	19
16k	▁mons ano ▁zo ▁ur ▁gumun ▁italian ▁e ▁proviñs ▁ancona , ... (+6 more)	16
32k	▁mons ano ▁zo ▁ur ▁gumun ▁italian ▁e ▁proviñs ▁ancona , ... (+6 more)	16
64k	▁mons ano ▁zo ▁ur ▁gumun ▁italian ▁e ▁proviñs ▁ancona , ... (+6 more)	16

Sample 2: San Asensio zo ur gumun e proviñs La Rioja en Spagn. Rioja

Vocab	Tokens	Count
8k	▁san ▁as ens io ▁zo ▁ur ▁gumun ▁e ▁proviñs ▁la ... (+5 more)	15
16k	▁san ▁as ens io ▁zo ▁ur ▁gumun ▁e ▁proviñs ▁la ... (+5 more)	15
32k	▁san ▁as ens io ▁zo ▁ur ▁gumun ▁e ▁proviñs ▁la ... (+5 more)	15
64k	▁san ▁as ens io ▁zo ▁ur ▁gumun ▁e ▁proviñs ▁la ... (+5 more)	15

Sample 3: Segusino zo ur gumun e proviñs Treviso e Veneto, en Italia.

Vocab	Tokens	Count
8k	▁seg us ino ▁zo ▁ur ▁gumun ▁e ▁proviñs ▁trev iso ... (+6 more)	16
16k	▁seg us ino ▁zo ▁ur ▁gumun ▁e ▁proviñs ▁treviso ▁e ... (+5 more)	15
32k	▁seg us ino ▁zo ▁ur ▁gumun ▁e ▁proviñs ▁treviso ▁e ... (+5 more)	15
64k	▁seg us ino ▁zo ▁ur ▁gumun ▁e ▁proviñs ▁treviso ▁e ... (+5 more)	15

Key Findings

• Best Compression: 64k achieves 3.786x compression
• Lowest UNK Rate: 8k with 0.4490% unknown tokens
• Trade-off: Larger vocabularies improve compression but increase model size
• Recommendation: 32k vocabulary provides optimal balance for production use

---

2. N-gram Model Evaluation

Results

N-gram	Variant	Perplexity	Entropy	Unique N-grams	Top-100 Coverage	Top-1000 Coverage
2-gram	Word	37,349	15.19	296,192	13.7%	32.0%
2-gram	Subword	294 🏆	8.20	11,776	65.3%	98.9%
3-gram	Word	128,487	16.97	570,380	5.9%	19.5%
3-gram	Subword	2,726	11.41	81,142	23.8%	68.1%
4-gram	Word	279,047	18.09	973,376	4.1%	14.8%
4-gram	Subword	17,313	14.08	421,873	10.8%	35.5%

Top 5 N-grams by Size

2-grams (Word):

Rank	N-gram	Count
1	e voe	59,782
2	ar c	55,139
3	a viz	53,711
4	e oa	52,022
5	d ar	47,935

3-grams (Word):

Rank	N-gram	Count
1	zo ur gumun	17,678
2	bro c hall	15,638
3	a zo ur	15,315
4	e oa bet	12,845
5	ur gumun eus	8,897

4-grams (Word):

Rank	N-gram	Count
1	zo ur gumun eus	8,261
2	monumantoù ha traoù heverk	5,435
3	a zo ur gumun	5,065
4	zo ur gumun e	4,314
5	monumant ar re varv	3,991

2-grams (Subword):

Rank	N-gram	Count
1	_ a	1,901,092
2	_ e	1,675,345
3	a n	1,608,231
4	e _	1,592,896
5	r _	1,428,493

3-grams (Subword):

Rank	N-gram	Count
1	a r _	640,562
2	_ e _	639,818
3	e t _	623,800
4	_ a r	555,503
5	e n n	467,995

4-grams (Subword):

Rank	N-gram	Count
1	_ a r _	456,425
2	_ a n _	280,111
3	a n t _	269,237
4	_ g a n	228,714
5	_ h a _	221,977

Key Findings

• Best Perplexity: 2-gram (subword) with 294
• Entropy Trend: Decreases with larger n-grams (more predictable)
• Coverage: Top-1000 patterns cover ~35% of corpus
• Recommendation: 4-gram or 5-gram for best predictive performance

---

3. Markov Chain Evaluation

Results

Context	Variant	Avg Entropy	Perplexity	Branching Factor	Unique Contexts	Predictability
1	Word	0.8900	1.853	7.60	546,189	11.0%
1	Subword	0.8953	1.860	5.85	8,402	10.5%
2	Word	0.3300	1.257	2.04	4,129,549	67.0%
2	Subword	0.6669	1.588	4.21	49,100	33.3%
3	Word	0.1561	1.114	1.34	8,377,190	84.4%
3	Subword	0.6656	1.586	3.74	206,555	33.4%
4	Word	0.0728 🏆	1.052	1.13	11,216,136	92.7%
4	Subword	0.6497	1.569	3.22	772,246	35.0%

Generated Text Samples (Word-based)

Below are text samples generated from each word-based Markov chain model:

Context Size 1:

1. e breizh e vennozhioù diskouez an 18 muhelder bihanañ 34 evel caravan palace da vezañ ivez
2. ar zastava e amzer tredeoged betek da vare a eskemm ha pterosaurus petra a ra an
3. a oa un heñvelster gant dean bounce prison gang crime deep grand prix de france bleu

Context Size 2:

1. e voe embannet an testennoù klasel e vez da 800 000 den pe gant kraf ar revelezh
2. ar c hembraeg mawr bras tolkien avat en doa bet ur gwadliñvadur e barzh ar c haramel
3. a viz eost a oa ul livour hag un impalaer e voe kumun kernitron al lann e

Context Size 3:

1. zo ur gumun en italia e proviñs cuneo etre kumunioù entracque ha valdieri anezhañ unan eus ugent lev...
2. a zo ur bronneg geotdebrer a vev er meurvor atlantel belle isle distaget bɛl ˈaɪl e saozneg zo
3. bro c hall zo un tiern ag ar morioù a zo e kenver ar sonerezh met evel ul

Context Size 4:

1. zo ur gumun eus meurgêr palermo e sikilia un 3 400 a dud zo enni o chom anezhi an
2. monumantoù ha traoù heverk iliz katolik saint eustacheclochers de france douaroniezh emdroadur ar bo...
3. a zo ur gumun eus italia e proviñs piacenza e rannvro emilia romagna ha 525 940 a dud o

Generated Text Samples (Subword-based)

Below are text samples generated from each subword-based Markov chain model:

Context Size 1:

1. _t_onur_b_el_gan
2. er_pakoumiat_g_v
3. adoù-stiz_aleuri

Context Size 2:

1. _a_un_erl_ezenner
2. _e_he_c'hhn_/_niz
3. an_ero_liged_;_ev

Context Size 3:

1. ar_bretek_ilioù,_o
2. _e_pyrrarkva,_leon
3. et_gant_franne,_ga

Context Size 4:

1. _ar_spesadoù_war_ar
2. _an_emsavid_fy_nhad
3. ant_an_alamanentiad

Key Findings

• Best Predictability: Context-4 (word) with 92.7% predictability
• Branching Factor: Decreases with context size (more deterministic)
• Memory Trade-off: Larger contexts require more storage (772,246 contexts)
• Recommendation: Context-3 or Context-4 for text generation

---

4. Vocabulary Analysis

Statistics

Metric	Value
Vocabulary Size	242,115
Total Tokens	15,327,088
Mean Frequency	63.30
Median Frequency	4
Frequency Std Dev	2500.68

Most Common Words

Rank	Word	Frequency
1	e	701,948
2	ar	517,584
3	a	464,667
4	an	326,300
5	ha	228,454
6	gant	189,759
7	c	186,830
8	en	181,309
9	da	170,732
10	ur	158,708

Least Common Words (from vocabulary)

Rank	Word	Frequency
1	nfpb	2
2	konjic	2
3	formoraich	2
4	vsn	2
5	moldavie	2
6	yankovich	2
7	gueydon	2
8	tréhouart	2
9	bouguen	2
10	shimosa	2

Zipf's Law Analysis

Metric	Value
Zipf Coefficient	1.1106
R² (Goodness of Fit)	0.996763
Adherence Quality	excellent

Coverage Analysis

Top N Words	Coverage
Top 100	41.7%
Top 1,000	65.8%
Top 5,000	80.4%
Top 10,000	85.6%

Key Findings

• Zipf Compliance: R²=0.9968 indicates excellent adherence to Zipf's law
• High Frequency Dominance: Top 100 words cover 41.7% of corpus
• Long Tail: 232,115 words needed for remaining 14.4% coverage

---

5. Word Embeddings Evaluation

5.1 Cross-Lingual Alignment

Note: Multilingual alignment visualization not available for this language.

5.2 Model Comparison

Model	Dimension	Isotropy	Semantic Density	Alignment R@1	Alignment R@10
mono_32d	32	0.8120	0.3605	N/A	N/A
mono_64d	64	0.8171 🏆	0.2761	N/A	N/A
mono_128d	128	0.7922	0.2119	N/A	N/A

Key Findings

• Best Isotropy: mono_64d with 0.8171 (more uniform distribution)
• Semantic Density: Average pairwise similarity of 0.2828. Lower values indicate better semantic separation.
• Alignment Quality: No aligned models evaluated in this run.
• Recommendation: 128d aligned for best cross-lingual performance

---

6. Morphological Analysis (Experimental)

⚠️ Warning: This language shows low morphological productivity. The statistical signals used for this analysis may be noisy or less reliable than for morphologically rich languages.

This section presents an automated morphological analysis derived from the statistical divergence between word-level and subword-level models. By analyzing where subword predictability spikes and where word-level coverage fails, we can infer linguistic structures without supervised data.

6.1 Productivity & Complexity

Metric	Value	Interpretation	Recommendation
Productivity Index	0.000	Low morphological productivity	⚠️ Likely unreliable
Idiomaticity Gap	-1.000	Low formulaic content	-

6.2 Affix Inventory (Productive Units)

These are the most productive prefixes and suffixes identified by sampling the vocabulary for global substitutability patterns. A unit is considered an affix if stripping it leaves a valid stem that appears in other contexts.

Productive Prefixes

Prefix	Examples

Productive Suffixes

Suffix	Examples
-s	mariånas, gilgamès, battalions
-er	tufer, hutier, beaver
-où	damkanadoù, barrennoù, heitioù
-es	tarbes, marcondes, cordes
-us	luchinus, menenius, fulmarus
-en	weyden, minchen, wageningen

6.3 Bound Stems (Lexical Roots)

Bound stems are high-frequency subword units that are semantically cohesive but rarely appear as standalone words. These often correspond to the 'core' of a word that requires inflection or derivation to be valid.

Stem	Cohesion	Substitutability	Examples
tion	2.51x	77 contexts	tione, metion, aktion
emba	2.14x	41 contexts	pemba, emban, bemba
adoù	1.80x	74 contexts	dadoù, kadoù, zadoù
nnet	1.78x	70 contexts	annet, bonnet, linnet
iamm	2.35x	24 contexts	liamm, fiamma, fiamme
ouar	1.48x	126 contexts	douar, zouar, mouar
nnad	1.52x	97 contexts	bennad, rannad, hannad
zhio	1.87x	40 contexts	uzhioù, lezhioù, bezhioù
zhañ	1.91x	35 contexts	ezhañ, kozhañ, dizhañ
nnoù	1.84x	39 contexts	tennoù, vannoù, bennoù
hone	1.81x	40 contexts	honeg, khone, dhone
reze	1.46x	94 contexts	breze, dreze, rezet

6.4 Affix Compatibility (Co-occurrence)

This table shows which prefixes and suffixes most frequently co-occur on the same stems, revealing the 'stacking' rules of the language's morphology.

No significant affix co-occurrences detected.

6.5 Recursive Morpheme Segmentation

Using Recursive Hierarchical Substitutability, we decompose complex words into their constituent morphemes. This approach handles nested affixes (e.g., prefix-prefix-root-suffix).

Word	Suggested Split	Confidence	Stem
eildelwennoù	eildelwenn-où	4.5	eildelwenn
hejadennoù	hejadenn-où	4.5	hejadenn
wissenschaften	wissenschaft-en	4.5	wissenschaft
beauvaisen	beauvais-en	4.5	beauvais
wellaennoù	wellaenn-où	4.5	wellaenn
antoninus	antonin-us	4.5	antonin
pluñvennoù	pluñvenn-où	4.5	pluñvenn
kementadoù	kementad-où	4.5	kementad
compositores	compositor-es	4.5	compositor
garidelloù	garidell-où	4.5	garidell
hromozomoù	hromozom-où	4.5	hromozom
reolennoù	reolenn-où	4.5	reolenn
barringer	barring-er	4.5	barring
diamantes	diamant-es	4.5	diamant
stradivarius	stradivari-us	4.5	stradivari

6.6 Linguistic Interpretation

Automated Insight: The language BR appears to be more isolating or has a highly fixed vocabulary. Word-level models perform nearly as well as subword models, indicating fewer productive morphological processes.

---

7. Summary & Recommendations

Production Recommendations

Component	Recommended	Rationale
Tokenizer	64k BPE	Best compression (3.79x)
N-gram	2-gram	Lowest perplexity (294)
Markov	Context-4	Highest predictability (92.7%)
Embeddings	100d	Balanced semantic capture and isotropy

---

Appendix: Metrics Glossary & Interpretation Guide

This section provides definitions, intuitions, and guidance for interpreting the metrics used throughout this report.

Tokenizer Metrics

Compression Ratio

Definition: The ratio of characters to tokens (chars/token). Measures how efficiently the tokenizer represents text.

Intuition: Higher compression means fewer tokens needed to represent the same text, reducing sequence lengths for downstream models. A 3x compression means ~3 characters per token on average.

What to seek: Higher is generally better for efficiency, but extremely high compression may indicate overly aggressive merging that loses morphological information.

Average Token Length (Fertility)

Definition: Mean number of characters per token produced by the tokenizer.

Intuition: Reflects the granularity of tokenization. Longer tokens capture more context but may struggle with rare words; shorter tokens are more flexible but increase sequence length.

What to seek: Balance between 2-5 characters for most languages. Arabic/morphologically-rich languages may benefit from slightly longer tokens.

Unknown Token Rate (OOV Rate)

Definition: Percentage of tokens that map to the unknown/UNK token, indicating words the tokenizer cannot represent.

Intuition: Lower OOV means better vocabulary coverage. High OOV indicates the tokenizer encounters many unseen character sequences.

What to seek: Below 1% is excellent; below 5% is acceptable. BPE tokenizers typically achieve very low OOV due to subword fallback.

N-gram Model Metrics

Perplexity

Definition: Measures how "surprised" the model is by test data. Mathematically: 2^(cross-entropy). Lower values indicate better prediction.

Intuition: If perplexity is 100, the model is as uncertain as if choosing uniformly among 100 options at each step. A perplexity of 10 means effectively choosing among 10 equally likely options.

What to seek: Lower is better. Perplexity decreases with larger n-grams (more context). Values vary widely by language and corpus size.

Entropy

Definition: Average information content (in bits) needed to encode the next token given the context. Related to perplexity: perplexity = 2^entropy.

Intuition: High entropy means high uncertainty/randomness; low entropy means predictable patterns. Natural language typically has entropy between 1-4 bits per character.

What to seek: Lower entropy indicates more predictable text patterns. Entropy should decrease as n-gram size increases.

Coverage (Top-K)

Definition: Percentage of corpus occurrences explained by the top K most frequent n-grams.

Intuition: High coverage with few patterns indicates repetitive/formulaic text; low coverage suggests diverse vocabulary usage.

What to seek: Depends on use case. For language modeling, moderate coverage (40-60% with top-1000) is typical for natural text.

Markov Chain Metrics

Average Entropy

Definition: Mean entropy across all contexts, measuring average uncertainty in next-word prediction.

Intuition: Lower entropy means the model is more confident about what comes next. Context-1 has high entropy (many possible next words); Context-4 has low entropy (few likely continuations).

What to seek: Decreasing entropy with larger context sizes. Very low entropy (<0.1) indicates highly deterministic transitions.

Branching Factor

Definition: Average number of unique next tokens observed for each context.

Intuition: High branching = many possible continuations (flexible but uncertain); low branching = few options (predictable but potentially repetitive).

What to seek: Branching factor should decrease with context size. Values near 1.0 indicate nearly deterministic chains.

Predictability

Definition: Derived metric: (1 - normalized_entropy) × 100%. Indicates how deterministic the model's predictions are.

Intuition: 100% predictability means the next word is always certain; 0% means completely random. Real text falls between these extremes.

What to seek: Higher predictability for text generation quality, but too high (>98%) may produce repetitive output.

Vocabulary & Zipf's Law Metrics

Zipf's Coefficient

Definition: The slope of the log-log plot of word frequency vs. rank. Zipf's law predicts this should be approximately -1.

Intuition: A coefficient near -1 indicates the corpus follows natural language patterns where a few words are very common and most words are rare.

What to seek: Values between -0.8 and -1.2 indicate healthy natural language distribution. Deviations may suggest domain-specific or artificial text.

R² (Coefficient of Determination)

Definition: Measures how well the linear fit explains the frequency-rank relationship. Ranges from 0 to 1.

Intuition: R² near 1.0 means the data closely follows Zipf's law; lower values indicate deviation from expected word frequency patterns.

What to seek: R² > 0.95 is excellent; > 0.99 indicates near-perfect Zipf adherence typical of large natural corpora.

Vocabulary Coverage

Definition: Cumulative percentage of corpus tokens accounted for by the top N words.

Intuition: Shows how concentrated word usage is. If top-100 words cover 50% of text, the corpus relies heavily on common words.

What to seek: Top-100 covering 30-50% is typical. Higher coverage indicates more repetitive text; lower suggests richer vocabulary.

Word Embedding Metrics

Isotropy

Definition: Measures how uniformly distributed vectors are in the embedding space. Computed as the ratio of minimum to maximum singular values.

Intuition: High isotropy (near 1.0) means vectors spread evenly in all directions; low isotropy means vectors cluster in certain directions, reducing expressiveness.

What to seek: Higher isotropy generally indicates better-quality embeddings. Values > 0.1 are reasonable; > 0.3 is good. Lower-dimensional embeddings tend to have higher isotropy.

Average Norm

Definition: Mean magnitude (L2 norm) of word vectors in the embedding space.

Intuition: Indicates the typical "length" of vectors. Consistent norms suggest stable training; high variance may indicate some words are undertrained.

What to seek: Relatively consistent norms across models. The absolute value matters less than consistency (low std deviation).

Cosine Similarity

Definition: Measures angular similarity between vectors, ranging from -1 (opposite) to 1 (identical direction).

Intuition: Words with similar meanings should have high cosine similarity. This is the standard metric for semantic relatedness in embeddings.

What to seek: Semantically related words should score > 0.5; unrelated words should be near 0. Synonyms often score > 0.7.

t-SNE Visualization

Definition: t-Distributed Stochastic Neighbor Embedding - a dimensionality reduction technique that preserves local structure for visualization.

Intuition: Clusters in t-SNE plots indicate groups of semantically related words. Spread indicates vocabulary diversity; tight clusters suggest semantic coherence.

What to seek: Meaningful clusters (e.g., numbers together, verbs together). Avoid over-interpreting distances - t-SNE preserves local, not global, structure.

General Interpretation Guidelines

1. Compare within model families: Metrics are most meaningful when comparing models of the same type (e.g., 8k vs 64k tokenizer).
2. Consider trade-offs: Better performance on one metric often comes at the cost of another (e.g., compression vs. OOV rate).
3. Context matters: Optimal values depend on downstream tasks. Text generation may prioritize different metrics than classification.
4. Corpus influence: All metrics are influenced by corpus characteristics. Wikipedia text differs from social media or literature.
5. Language-specific patterns: Morphologically rich languages (like Arabic) may show different optimal ranges than analytic languages.

Visualizations Index

Visualization	Description
Tokenizer Compression	Compression ratios by vocabulary size
Tokenizer Fertility	Average token length by vocabulary
Tokenizer OOV	Unknown token rates
Tokenizer Total Tokens	Total tokens by vocabulary
N-gram Perplexity	Perplexity by n-gram size
N-gram Entropy	Entropy by n-gram size
N-gram Coverage	Top pattern coverage
N-gram Unique	Unique n-gram counts
Markov Entropy	Entropy by context size
Markov Branching	Branching factor by context
Markov Contexts	Unique context counts
Zipf's Law	Frequency-rank distribution with fit
Vocab Frequency	Word frequency distribution
Top 20 Words	Most frequent words
Vocab Coverage	Cumulative coverage curve
Embedding Isotropy	Vector space uniformity
Embedding Norms	Vector magnitude distribution
Embedding Similarity	Word similarity heatmap
Nearest Neighbors	Similar words for key terms
t-SNE Words	2D word embedding visualization
t-SNE Sentences	2D sentence embedding visualization
Position Encoding	Encoding method comparison
Model Sizes	Storage requirements
Performance Dashboard	Comprehensive performance overview

---

About This Project

Data Source

Models trained on wikipedia-monthly - a monthly snapshot of Wikipedia articles across 300+ languages.

Project

A project by Wikilangs - Open-source NLP models for every Wikipedia language.

Maintainer

Omar Kamali - Omneity Labs

Citation

If you use these models in your research, please cite:

@misc{wikilangs2025,
  author = {Kamali, Omar},
  title = {Wikilangs: Open NLP Models for Wikipedia Languages},
  year = {2025},
  doi = {10.5281/zenodo.18073153},
  publisher = {Zenodo},
  url = {https://huggingface.co/wikilangs}
  institution = {Omneity Labs}
}

License

MIT License - Free for academic and commercial use.

Links

• 🌐 Website: wikilangs.org
• 🤗 Models: huggingface.co/wikilangs
• 📊 Data: wikipedia-monthly
• 👤 Author: Omar Kamali
• 🤝 Sponsor: Featherless AI

---

Generated by Wikilangs Models Pipeline

Report Date: 2026-01-03 08:48:16