README.md

---
language: atj
language_name: Atikamekw
language_family: american_algonquian
tags:
  - wikilangs
  - nlp
  - tokenizer
  - embeddings
  - n-gram
  - markov
  - wikipedia
  - feature-extraction
  - sentence-similarity
  - tokenization
  - n-grams
  - markov-chain
  - text-mining
  - fasttext
  - babelvec
  - vocabulous
  - vocabulary
  - monolingual
  - family-american_algonquian
license: mit
library_name: wikilangs
pipeline_tag: text-generation
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: 5.953
  - name: best_isotropy
    type: isotropy
    value: 0.1437
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-03
---

# Atikamekw - Wikilangs Models
## Comprehensive Research Report & Full Ablation Study

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Atikamekw** 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

![Performance Dashboard](visualizations/performance_dashboard.png)

### Analysis and Evaluation

- [1. Tokenizer Evaluation](#1-tokenizer-evaluation)
- [2. N-gram Model Evaluation](#2-n-gram-model-evaluation)
- [3. Markov Chain Evaluation](#3-markov-chain-evaluation)
- [4. Vocabulary Analysis](#4-vocabulary-analysis)
- [5. Word Embeddings Evaluation](#5-word-embeddings-evaluation)
- [6. Morphological Analysis (Experimental)](#6--morphological-analysis-experimental)
- [7. Summary & Recommendations](#7-summary--recommendations)
- [Metrics Glossary](#appendix-metrics-glossary--interpretation-guide)
- [Visualizations Index](#visualizations-index)

---
## 1. Tokenizer Evaluation

![Tokenizer Compression](visualizations/tokenizer_compression.png)

![Tokenizer Fertility](visualizations/tokenizer_fertility.png)

![Tokenizer OOV](visualizations/tokenizer_oov.png)

![Total Tokens](visualizations/tokenizer_total_tokens.png)

### Results

| Vocab Size | Compression | Avg Token Len | UNK Rate | Total Tokens |
|------------|-------------|---------------|----------|--------------|
| **8k** | 5.122x | 5.13 | 0.1886% | 91,751 |
| **16k** | 5.512x | 5.52 | 0.2029% | 85,261 |
| **32k** | 5.953x 🏆 | 5.97 | 0.2191% | 78,943 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Sainte-Anne-des-Monts oteno Kepek askik ici actew, Kanata. Irikik e tacinaniwok ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁sainte - anne - des - mont s ▁oteno ▁kepek ... (+16 more)` | 26 |
| 16k | `▁sainte - anne - des - monts ▁oteno ▁kepek ▁askik ... (+15 more)` | 25 |
| 32k | `▁sainte - anne - des - monts ▁oteno ▁kepek ▁askik ... (+15 more)` | 25 |

**Sample 2:** `Mulgrave oteno Nouvelle-Écosse aski ici actew, Kanata. Irikik e tacinaniwok 879 ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁m ul gra ve ▁oteno ▁nouvelle - écosse ▁aski ▁ici ... (+12 more)` | 22 |
| 16k | `▁mulgrave ▁oteno ▁nouvelle - écosse ▁aski ▁ici ▁actew , ▁kanata ... (+9 more)` | 19 |
| 32k | `▁mulgrave ▁oteno ▁nouvelle - écosse ▁aski ▁ici ▁actew , ▁kanata ... (+9 more)` | 19 |

**Sample 3:** `Gracefield oteno Kepek askik ici actew, Kanata. Irikik e tacinaniwok 2 462 matce...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁gra ce field ▁oteno ▁kepek ▁askik ▁ici ▁actew , ▁kanata ... (+11 more)` | 21 |
| 16k | `▁gra ce field ▁oteno ▁kepek ▁askik ▁ici ▁actew , ▁kanata ... (+11 more)` | 21 |
| 32k | `▁gracefield ▁oteno ▁kepek ▁askik ▁ici ▁actew , ▁kanata . ▁irikik ... (+9 more)` | 19 |


### Key Findings

- **Best Compression:** 32k achieves 5.953x compression
- **Lowest UNK Rate:** 8k with 0.1886% 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

![N-gram Perplexity](visualizations/ngram_perplexity.png)

![N-gram Unique](visualizations/ngram_unique.png)

![N-gram Coverage](visualizations/ngram_coverage.png)

### Results

| N-gram | Variant | Perplexity | Entropy | Unique N-grams | Top-100 Coverage | Top-1000 Coverage |
|--------|---------|------------|---------|----------------|------------------|-------------------|
| **2-gram** | Word | 755 | 9.56 | 2,021 | 44.7% | 84.2% |
| **2-gram** | Subword | 129 🏆 | 7.01 | 987 | 89.0% | 100.0% |
| **3-gram** | Word | 540 | 9.08 | 1,854 | 50.0% | 84.6% |
| **3-gram** | Subword | 759 | 9.57 | 5,467 | 41.9% | 92.6% |
| **4-gram** | Word | 584 | 9.19 | 2,555 | 50.3% | 75.4% |
| **4-gram** | Subword | 3,031 | 11.57 | 19,166 | 21.7% | 66.0% |
| **5-gram** | Word | 345 | 8.43 | 1,658 | 58.1% | 85.5% |
| **5-gram** | Subword | 7,892 | 12.95 | 37,893 | 14.8% | 46.5% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ici actew` | 888 |
| 2 | `actew kanata` | 771 |
| 3 | `manawan wemotaci` | 721 |
| 4 | `e ici` | 685 |
| 5 | `irikik e` | 672 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ici actew kanata` | 770 |
| 2 | `irikik e tacinaniwok` | 633 |
| 3 | `kanata irikik e` | 620 |
| 4 | `actew kanata irikik` | 620 |
| 5 | `askik ici actew` | 500 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `kanata irikik e tacinaniwok` | 620 |
| 2 | `actew kanata irikik e` | 620 |
| 3 | `ici actew kanata irikik` | 620 |
| 4 | `askik ici actew kanata` | 490 |
| 5 | `kepek askik ici actew` | 457 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ici actew kanata irikik e` | 620 |
| 2 | `actew kanata irikik e tacinaniwok` | 620 |
| 3 | `kepek askik ici actew kanata` | 455 |
| 4 | `askik ici actew kanata irikik` | 358 |
| 5 | `oteno kepek askik ici actew` | 326 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `c i` | 23,681 |
| 2 | `k a` | 23,540 |
| 3 | `_ k` | 23,289 |
| 4 | `t c` | 23,201 |
| 5 | `i k` | 21,032 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `t c i` | 11,312 |
| 2 | `_ k i` | 10,113 |
| 3 | `i t c` | 10,005 |
| 4 | `_ k a` | 9,180 |
| 5 | `c i _` | 8,655 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `i t c i` | 5,891 |
| 2 | `a n i w` | 5,154 |
| 3 | `_ k a _` | 4,777 |
| 4 | `n i w o` | 4,372 |
| 5 | `k a n i` | 4,233 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a n i w o` | 3,980 |
| 2 | `n i w o k` | 3,620 |
| 3 | `k a n i w` | 3,557 |
| 4 | `a k a n i` | 3,262 |
| 5 | `_ m a t c` | 2,919 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 129
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~47% of corpus
- **Recommendation:** 4-gram or 5-gram for best predictive performance

---
## 3. Markov Chain Evaluation

![Markov Entropy](visualizations/markov_entropy.png)

![Markov Contexts](visualizations/markov_contexts.png)

![Markov Branching](visualizations/markov_branching.png)

### Results

| Context | Variant | Avg Entropy | Perplexity | Branching Factor | Unique Contexts | Predictability |
|---------|---------|-------------|------------|------------------|-----------------|----------------|
| **1** | Word | 0.5828 | 1.498 | 3.55 | 19,248 | 41.7% |
| **1** | Subword | 1.5433 | 2.915 | 13.86 | 118 | 0.0% |
| **2** | Word | 0.1881 | 1.139 | 1.41 | 67,567 | 81.2% |
| **2** | Subword | 1.2598 | 2.395 | 6.30 | 1,635 | 0.0% |
| **3** | Word | 0.0530 | 1.037 | 1.09 | 93,703 | 94.7% |
| **3** | Subword | 0.7971 | 1.738 | 3.30 | 10,279 | 20.3% |
| **4** | Word | 0.0146 🏆 | 1.010 | 1.02 | 99,898 | 98.5% |
| **4** | Subword | 0.5503 | 1.464 | 2.26 | 33,860 | 45.0% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `e totcikatek arimatc aric kirowe warowik e iti matce tipaskonikik ka tato piponikarik awik e kitotc`
2. `ka takocinokopanen 22 otatakon pisimw nac mocak ki tesinikew kaie e tacinaniwok 352 395 matcectakani...`
3. `ki pe ocitakaniwoki mikiwama ki ponimatisirikopon marianne ki kicikateriw kitci matcihitisotc nehiro...`

**Context Size 2:**

1. `ici actew kanata irikik e tacinaniwok 53 939 matcectakaniwok`
2. `actew kanata irikik e tacinaniwok 10 051 matcectakaniwok`
3. `manawan wemotaci patak apitisiw anihe kirowe ka atiparik kecpin e orowinaniwok pitakamik e tacikaniw...`

**Context Size 3:**

1. `ici actew kanata irikik e tacinaniwok 20 161 e ici tipatcimomakak nicw takon anohwe nehiro oteno ket...`
2. `kanata irikik e tacinaniwok 10 051 matcectakaniwok`
3. `actew kanata irikik e tacinaniwok 2 216 matcectakaniwok`

**Context Size 4:**

1. `actew kanata irikik e tacinaniwok 7 347 matcectakaniwok`
2. `ici actew kanata irikik e tacinaniwok 7 282 matcectakaniwok`
3. `kanata irikik e tacinaniwok 973 matcectakaniwok`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `iwoka_di_naw_k_m`
2. `_m._ki_nanew._ka`
3. `atcotakie_ak,_ac`

**Context Size 2:**

1. `cina._tacimoodre_`
2. `kaniniwee_icitci_`
3. `_ki_ek_itcik._mot`

**Context Size 3:**

1. `tcik._matcectapwat`
2. `_ki_icitc_kitc_aga`
3. `itciwok._kaie_nta_`

**Context Size 4:**

1. `itcisowapinaniwiw_k`
2. `aniwonik_meka_ki_oc`
3. `_ka_tatopiponen_nip`


### Key Findings

- **Best Predictability:** Context-4 (word) with 98.5% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (33,860 contexts)
- **Recommendation:** Context-3 or Context-4 for text generation

---
## 4. Vocabulary Analysis

![Zipf's Law](visualizations/zipf_law.png)

![Top Words](visualizations/top20_words.png)

![Coverage Curve](visualizations/vocab_coverage.png)

### Statistics

| Metric | Value |
|--------|-------|
| Vocabulary Size | 6,458 |
| Total Tokens | 105,050 |
| Mean Frequency | 16.27 |
| Median Frequency | 3 |
| Frequency Std Dev | 131.25 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | e | 6,358 |
| 2 | ka | 4,817 |
| 3 | ki | 3,659 |
| 4 | ici | 2,655 |
| 5 | kitci | 1,874 |
| 6 | kaie | 1,655 |
| 7 | matcectakaniwok | 1,604 |
| 8 | micta | 1,222 |
| 9 | kirika | 1,111 |
| 10 | manawan | 972 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | nehirosi | 2 |
| 2 | cikomewokw | 2 |
| 3 | miitaw | 2 |
| 4 | droits | 2 |
| 5 | kiskinohamato | 2 |
| 6 | banque | 2 |
| 7 | mawotcicorianionik | 2 |
| 8 | fraser | 2 |
| 9 | otatisokaniwak | 2 |
| 10 | secwepemctsin | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0505 |
| R² (Goodness of Fit) | 0.987789 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 54.6% |
| Top 1,000 | 81.8% |
| Top 5,000 | 97.2% |
| Top 10,000 | 0.0% |

### Key Findings

- **Zipf Compliance:** R²=0.9878 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 54.6% of corpus
- **Long Tail:** -3,542 words needed for remaining 100.0% coverage

---
## 5. Word Embeddings Evaluation

![Embedding Isotropy](visualizations/embedding_isotropy.png)

![Similarity Matrix](visualizations/embedding_similarity.png)

![t-SNE Words](visualizations/tsne_words.png)

![t-SNE Sentences](visualizations/tsne_sentences.png)


### 5.1 Cross-Lingual Alignment

![Alignment Quality](visualizations/embedding_alignment_quality.png)

![Multilingual t-SNE](visualizations/embedding_tsne_multilingual.png)


### 5.2 Model Comparison

| Model | Dimension | Isotropy | Semantic Density | Alignment R@1 | Alignment R@10 |
|-------|-----------|----------|------------------|---------------|----------------|
| **mono_32d** | 32 | 0.1437 🏆 | 0.4915 | N/A | N/A |
| **mono_64d** | 64 | 0.0311 | 0.5012 | N/A | N/A |
| **mono_128d** | 128 | 0.0055 | 0.4973 | N/A | N/A |
| **aligned_32d** | 32 | 0.1437 | 0.4825 | 0.0091 | 0.1088 |
| **aligned_64d** | 64 | 0.0311 | 0.5079 | 0.0136 | 0.1066 |
| **aligned_128d** | 128 | 0.0055 | 0.4960 | 0.0317 | 0.1565 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.1437 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.4961. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 3.2% R@1 in cross-lingual retrieval.
- **Recommendation:** 128d aligned for best cross-lingual performance

---
## 6.  Morphological Analysis (Experimental)

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 | **4.183** | High morphological productivity | Reliable analysis |
| Idiomaticity Gap | **0.838** | High formulaic/idiomatic 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 |
|--------|----------|
| `-ki` | kitciki, kimosapitc, kinowapitamokw |
| `-mi` | mireritamiriwa, mitciso, mirokiw |
| `-ma` | maninikatew, matcectakaniwok, mars |
| `-ot` | ototokon, otenocic, otenawa |
| `-ni` | nitowakik, nikomesak, nitawikiritci |
| `-ic` | icikapowiw, icinikatikik, icinkatew |
| `-wi` | wirino, witamotcik, wirtip |
| `-ta` | takociretc, tacikeriwa, taritci |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-k` | titopiponikak, kanawapitcikatek, nitowakik |
| `-w` | pakonehohakiniwiw, kinowapitamokw, nipiriw |
| `-c` | kimosapitc, ponihatc, pamatisitc |
| `-n` | ototokon, owen, foundation |
| `-ik` | nitowakik, witamotcik, totowakaniwitcik |
| `-tc` | kimosapitc, ponihatc, pamatisitc |
| `-ok` | itakiniwok, ntokihitisohok, nakapewonok |
| `-iw` | pakonehohakiniwiw, nipiriw, mowakiniwiw |

### 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 |
|------|----------|------------------|----------|
| `tako` | 1.33x | 29 contexts | takok, takon, takoke |
| `taka` | 1.42x | 22 contexts | pataka, otakai, otakaci |
| `mitc` | 1.35x | 22 contexts | mitci, mitca, mitcim |
| `erit` | 1.54x | 14 contexts | wewerita, oreritam, iteritci |
| `apit` | 1.44x | 17 contexts | apita, tapit, apitc |
| `aniw` | 1.36x | 19 contexts | aniwe, kaniwok, nikaniw |
| `iwok` | 1.42x | 16 contexts | apiwok, irniwok, askiwok |
| `niwo` | 1.50x | 13 contexts | irniwok, koniwok, kaniwok |
| `kana` | 1.36x | 15 contexts | kanapé, kanada, oskana |
| `irow` | 1.51x | 11 contexts | kirowe, kewirow, wirowaw |
| `itak` | 1.35x | 15 contexts | witak, titak, kitaki |
| `kate` | 1.32x | 16 contexts | katek, makate, kateri |

### 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.

| Prefix | Suffix | Frequency | Examples |
|--------|--------|-----------|----------|
| `-ki` | `-k` | 127 words | kiceriniwok, kinokepitcikanik |
| `-mi` | `-k` | 89 words | mirwacinik, mictikok |
| `-ma` | `-k` | 89 words | matakanik, matcikonak |
| `-ki` | `-w` | 68 words | kicteritakoniw, kiskinohamakew |
| `-mi` | `-w` | 65 words | mitcetaw, micaw |
| `-ni` | `-k` | 60 words | nikickowatcik, nikapewnok |
| `-ot` | `-k` | 57 words | ototewok, otcikowik |
| `-ki` | `-ik` | 56 words | kinokepitcikanik, kickapiskarik |
| `-ki` | `-c` | 51 words | kinikositc, kictapeitc |
| `-ta` | `-k` | 49 words | tarasak, tacikaniwonik |

### 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 |
|------|-----------------|------------|------|
| otaskitcik | **`ot-aski-tc-ik`** | 7.5 | `aski` |
| wikiconvention | **`wi-ki-convention`** | 6.0 | `convention` |
| nehirowisitcik | **`nehirowisi-tc-ik`** | 6.0 | `nehirowisi` |
| kiskerimakaniwiw | **`ki-skerimak-an-iw-iw`** | 6.0 | `skerimak` |
| takapikenikaniw | **`ta-kapiken-ik-an-iw`** | 6.0 | `kapiken` |
| wicamakaniwiw | **`wi-camak-an-iw-iw`** | 6.0 | `camak` |
| nikickotatotcik | **`ni-ki-ckotato-tc-ik`** | 6.0 | `ckotato` |
| kackihotcik | **`kackiho-tc-ik`** | 6.0 | `kackiho` |
| tipatcimotcik | **`tipatcimo-tc-ik`** | 6.0 | `tipatcimo` |
| takociretcik | **`ta-kocire-tc-ik`** | 4.5 | `kocire` |
| apatcihakaniwiw | **`apatcihak-an-iw-iw`** | 4.5 | `apatcihak` |
| takocinitcik | **`ta-kocini-tc-ik`** | 4.5 | `kocini` |
| kicowekaniw | **`ki-cowek-an-iw`** | 4.5 | `cowek` |
| emitcikocimotc | **`emitcikocimo-tc`** | 4.5 | `emitcikocimo` |
| apitcihakaniwiw | **`apitcihak-an-iw-iw`** | 4.5 | `apitcihak` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Atikamekw shows high morphological productivity. The subword models are significantly more efficient than word models, suggesting a rich system of affixation or compounding.

> **Note on Idiomaticity:** The high Idiomaticity Gap suggests a large number of frequent multi-word expressions or formulaic sequences that are statistically distinct from their component parts.

---
## 7. Summary & Recommendations

![Performance Dashboard](visualizations/performance_dashboard.png)

### Production Recommendations

| Component | Recommended | Rationale |
|-----------|-------------|-----------|
| Tokenizer | **32k BPE** | Best compression (5.95x) |
| N-gram | **2-gram** | Lowest perplexity (129) |
| Markov | **Context-4** | Highest predictability (98.5%) |
| 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](https://huggingface.co/datasets/omarkamali/wikipedia-monthly) - a monthly snapshot of Wikipedia articles across 300+ languages.

### Project

A project by **[Wikilangs](https://wikilangs.org)** - Open-source NLP models for every Wikipedia language.

### Maintainer

[Omar Kamali](https://omarkamali.com) - [Omneity Labs](https://omneitylabs.com)

### Citation

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

```bibtex
@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](https://wikilangs.org)
- 🤗 Models: [huggingface.co/wikilangs](https://huggingface.co/wikilangs)
- 📊 Data: [wikipedia-monthly](https://huggingface.co/datasets/omarkamali/wikipedia-monthly)
- 👤 Author: [Omar Kamali](https://huggingface.co/omarkamali)
- 🤝 Sponsor: [Featherless AI](https://featherless.ai)
---
*Generated by Wikilangs Models Pipeline*

*Report Date: 2026-01-03 17:35:34*