README.md

---
language: guw
language_name: Gun
language_family: atlantic_kwa
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-atlantic_kwa
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: 4.344
  - name: best_isotropy
    type: isotropy
    value: 0.6893
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Gun** 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** | 3.777x | 3.78 | 0.8806% | 316,930 |
| **16k** | 4.030x | 4.03 | 0.9396% | 297,045 |
| **32k** | 4.225x | 4.23 | 0.9851% | 283,312 |
| **64k** | 4.344x 🏆 | 4.35 | 1.0127% | 275,607 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Aliko Dangote GCON (he yin jiji to azán 10tọ Lidosun yin ajọwatọ daho dé wẹ eyin...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ali ko ▁dan go te ▁g con ▁( he ▁yin ... (+26 more)` | 36 |
| 16k | `▁ali ko ▁dan go te ▁g con ▁( he ▁yin ... (+26 more)` | 36 |
| 32k | `▁aliko ▁dangote ▁gcon ▁( he ▁yin ▁jiji ▁to ▁azán ▁ ... (+22 more)` | 32 |
| 64k | `▁aliko ▁dangote ▁gcon ▁( he ▁yin ▁jiji ▁to ▁azán ▁ ... (+22 more)` | 32 |

**Sample 2:** `Fausat Adebola Ibikunle yin Gandutọ na Lizọnyizọn tito na Ayimatẹn Kaduna Tọn (M...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁fa us at ▁ade bola ▁ibi kunle ▁yin ▁gandutọ ▁na ... (+20 more)` | 30 |
| 16k | `▁fausat ▁ade bola ▁ibikunle ▁yin ▁gandutọ ▁na ▁lizọnyizọn ▁tito ▁na ... (+14 more)` | 24 |
| 32k | `▁fausat ▁adebola ▁ibikunle ▁yin ▁gandutọ ▁na ▁lizọnyizọn ▁tito ▁na ▁ayimatẹn ... (+13 more)` | 23 |
| 64k | `▁fausat ▁adebola ▁ibikunle ▁yin ▁gandutọ ▁na ▁lizọnyizọn ▁tito ▁na ▁ayimatẹn ... (+12 more)` | 22 |

**Sample 3:** `Mexico yin otò de to whèzẹtẹnwaji America tọn.he má do ayimatẹn voovo 32 ji`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁mexico ▁yin ▁otò ▁de ▁to ▁whèzẹtẹn waji ▁america ▁tọn . ... (+9 more)` | 19 |
| 16k | `▁mexico ▁yin ▁otò ▁de ▁to ▁whèzẹtẹnwaji ▁america ▁tọn . he ... (+8 more)` | 18 |
| 32k | `▁mexico ▁yin ▁otò ▁de ▁to ▁whèzẹtẹnwaji ▁america ▁tọn . he ... (+8 more)` | 18 |
| 64k | `▁mexico ▁yin ▁otò ▁de ▁to ▁whèzẹtẹnwaji ▁america ▁tọn . he ... (+8 more)` | 18 |


### Key Findings

- **Best Compression:** 64k achieves 4.344x compression
- **Lowest UNK Rate:** 8k with 0.8806% 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 | 3,238 | 11.66 | 9,467 | 26.6% | 57.5% |
| **2-gram** | Subword | 287 🏆 | 8.17 | 2,304 | 65.6% | 98.7% |
| **3-gram** | Word | 6,817 | 12.73 | 13,761 | 16.3% | 41.1% |
| **3-gram** | Subword | 2,147 | 11.07 | 16,102 | 29.3% | 71.3% |
| **4-gram** | Word | 13,775 | 13.75 | 22,441 | 11.1% | 27.8% |
| **4-gram** | Subword | 9,790 | 13.26 | 67,472 | 15.9% | 44.3% |
| **5-gram** | Word | 8,850 | 13.11 | 13,471 | 12.7% | 30.6% |
| **5-gram** | Subword | 25,712 | 14.65 | 135,866 | 10.9% | 31.6% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `tọn mẹ` | 2,333 |
| 2 | `tọn to` | 1,877 |
| 3 | `to owhe` | 1,528 |
| 4 | `tọn lẹ` | 1,460 |
| 5 | `he yin` | 1,165 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `yin jiji to` | 883 |
| 2 | `lẹ gbẹzan tọn` | 635 |
| 3 | `tọn mẹ to` | 527 |
| 4 | `alọdlẹndonu lẹ gbẹzan` | 518 |
| 5 | `he nọ yin` | 482 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `alọdlẹndonu lẹ gbẹzan tọn` | 518 |
| 2 | `he ye ji to` | 328 |
| 3 | `ji to owhe lẹ` | 325 |
| 4 | `ye ji to owhe` | 325 |
| 5 | `tọn he ye ji` | 268 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `he ye ji to owhe` | 325 |
| 2 | `ye ji to owhe lẹ` | 325 |
| 3 | `gbẹzan tọn he ye ji` | 268 |
| 4 | `tọn he ye ji to` | 268 |
| 5 | `lẹ gbẹzan tọn he ye` | 193 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n _` | 74,173 |
| 2 | `_ t` | 57,983 |
| 3 | `o _` | 53,938 |
| 4 | `e _` | 47,550 |
| 5 | `ọ n` | 34,910 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ọ n _` | 24,580 |
| 2 | `_ t o` | 24,003 |
| 3 | `t ọ n` | 22,974 |
| 4 | `t o _` | 22,793 |
| 5 | `_ t ọ` | 18,136 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ t o _` | 21,322 |
| 2 | `t ọ n _` | 19,056 |
| 3 | `_ t ọ n` | 17,899 |
| 4 | `_ y i n` | 10,442 |
| 5 | `y i n _` | 10,150 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ t ọ n _` | 14,591 |
| 2 | `_ y i n _` | 9,182 |
| 3 | `n _ t o _` | 5,040 |
| 4 | `_ t o _ a` | 4,718 |
| 5 | `e t ọ n _` | 4,023 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 287
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~32% 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 | 0.8470 | 1.799 | 5.09 | 31,940 | 15.3% |
| **1** | Subword | 1.3188 | 2.495 | 11.91 | 349 | 0.0% |
| **2** | Word | 0.2988 | 1.230 | 1.72 | 162,357 | 70.1% |
| **2** | Subword | 1.1476 | 2.215 | 6.83 | 4,157 | 0.0% |
| **3** | Word | 0.1238 | 1.090 | 1.22 | 279,288 | 87.6% |
| **3** | Subword | 0.8350 | 1.784 | 3.86 | 28,401 | 16.5% |
| **4** | Word | 0.0494 🏆 | 1.035 | 1.07 | 339,321 | 95.1% |
| **4** | Subword | 0.5898 | 1.505 | 2.45 | 109,522 | 41.0% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `to otogbo naijilia tọn ga ògán ná leke po akwashiki to abẹokuta to otannugbo kano e`
2. `tọn mẹ zọ̀nlinzinzin lọ bọ e hẹn azọn whẹdida tọn mẹ wa jogbe ɖɔ silent cal`
3. `yin didọ gándego e sọ pọn todohukanji kọnugbe hogbe po diọ yinkọ he e kú to`

**Context Size 2:**

1. `tọn mẹ to owhe kandewiatọ̀n yinyin mẹ alọdlẹndonu lẹ gbẹzan tọn he ko sọawuhia to aihundida cantata`
2. `tọn to abeokuta dopolọ finẹ wẹ zẹ́ndótọzọ́nwatẹn ladi kwali tọn ladi kwali tọn ladi kwali mẹhe sin`
3. `to owhe lẹ kú to whenẹnu freedom park lopo awọnlin tọn sọta dahomey first franco dahomean war`

**Context Size 3:**

1. `yin jiji to oto kutaisi nọvisunnu etọn we revaz gamkrelidze ewọ lọsu yin kanlinmọ de he yin hinhẹn`
2. `tọn mẹ to ayimatẹn kano tọn podọ to nu taidi owhe enẹlẹ e zindonukọn nado wazọn taidi ayinamẹtọ`
3. `alọdlẹndonu lẹ gbẹzan tọn lẹ he ye ji to owhe lẹ kú to owhe lẹ lẹ lẹ to`

**Context Size 4:**

1. `alọdlẹndonu lẹ gbẹzan tọn he ye ji to owhe lẹ kú to owhe lẹ lẹ lẹ to naijilia lẹ`
2. `he ye ji to owhe lẹ benẹnu gbẹzan tọn`
3. `ye ji to owhe lẹ lẹ lẹ to naijilia he ye ji to owhe lẹ kú to owhe lẹ`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_e_ay_topl_e_aba`
2. `ntọntohe_maovi_a`
3. `an_whunkalazunto`

**Context Size 2:**

1. `n_awe_yionu,_gbọn`
2. `_to_ogbàn_lẹzane_`
3. `o_wharcy_sia_yinu`

**Context Size 3:**

1. `ọn_mussive_sọn_alọ`
2. `_to_gbankan_e_nọ_m`
3. `to_arau_zogbe_kuku`

**Context Size 4:**

1. `_to_ogbe_de_avọ̀ta_l`
2. `tọn_azan_kpóɖɔ_to_n`
3. `_tọn_mẹ_e_jẹ_yọnnu_`


### Key Findings

- **Best Predictability:** Context-4 (word) with 95.1% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (109,522 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 | 15,734 |
| Total Tokens | 380,906 |
| Mean Frequency | 24.21 |
| Median Frequency | 4 |
| Frequency Std Dev | 294.06 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | to | 21,455 |
| 2 | tọn | 17,851 |
| 3 | lẹ | 9,999 |
| 4 | yin | 9,460 |
| 5 | e | 7,419 |
| 6 | he | 7,045 |
| 7 | po | 6,884 |
| 8 | mẹ | 6,420 |
| 9 | na | 4,037 |
| 10 | nọ | 3,975 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | zimbabwe | 2 |
| 2 | zambezi | 2 |
| 3 | okavango | 2 |
| 4 | nyagbé | 2 |
| 5 | malgache | 2 |
| 6 | enseignement | 2 |
| 7 | supérieur | 2 |
| 8 | labo | 2 |
| 9 | gadomè | 2 |
| 10 | linguistique | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.1244 |
| R² (Goodness of Fit) | 0.995927 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 51.3% |
| Top 1,000 | 76.5% |
| Top 5,000 | 91.6% |
| Top 10,000 | 96.9% |

### Key Findings

- **Zipf Compliance:** R²=0.9959 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 51.3% of corpus
- **Long Tail:** 5,734 words needed for remaining 3.1% 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 | 0.6893 | 0.3774 | N/A | N/A |
| **mono_64d** | 64 | 0.2689 | 0.3713 | N/A | N/A |
| **mono_128d** | 128 | 0.0512 | 0.3704 | N/A | N/A |
| **aligned_32d** | 32 | 0.6893 🏆 | 0.3922 | 0.0440 | 0.1880 |
| **aligned_64d** | 64 | 0.2689 | 0.3680 | 0.0480 | 0.2180 |
| **aligned_128d** | 128 | 0.0512 | 0.3656 | 0.0560 | 0.2900 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.6893 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.3742. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 5.6% 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 | **5.000** | High morphological productivity | Reliable analysis |
| Idiomaticity Gap | **-0.030** | 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 |
|--------|----------|
| `-tọ` | wehiatọ, wàtọ, banọhotọ |
| `-an` | gban, avɔsinsan, pan |

### 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` | 1.71x | 16 contexts | action, nation, auction |
| `ọnnu` | 1.63x | 15 contexts | fọnnu, yọnnu, dọnnu |
| `nukọ` | 1.62x | 15 contexts | nukọn, nukọ̀n, jẹnukọn |
| `ukun` | 1.59x | 14 contexts | wukun, nukun, kukuna |
| `nuku` | 1.63x | 13 contexts | anuku, nukun, jinukun |
| `ukọn` | 1.60x | 13 contexts | nukọn, jẹnukọn, nukọnna |
| `yọnẹ` | 1.69x | 11 contexts | yọnẹn, oyọnẹn, nuyọnẹn |
| `hund` | 1.52x | 14 contexts | aihunda, hundote, ahundopo |
| `ọnẹn` | 1.78x | 9 contexts | yọnẹn, oyọnẹn, nuyọnẹn |
| `nlin` | 1.47x | 15 contexts | online, kanlin, linlin |
| `henu` | 1.79x | 8 contexts | whenu, whenue, vuwhenu |
| `gand` | 1.47x | 12 contexts | gando, gandó, gandọ |

### 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 |
|------|-----------------|------------|------|
| lẹdogbedevomẹtọ | **`lẹdogbedevomẹ-tọ`** | 4.5 | `lẹdogbedevomẹ` |
| dahomeyan | **`dahomey-an`** | 4.5 | `dahomey` |
| ṣiantọ̀ntọ | **`ṣiantọ̀n-tọ`** | 4.5 | `ṣiantọ̀n` |
| nuplọnmẹtọ | **`nuplọnmẹ-tọ`** | 4.5 | `nuplọnmẹ` |
| gbanewheawetọ | **`gbanewheawe-tọ`** | 4.5 | `gbanewheawe` |
| azọ́nwatọ | **`azọ́nwa-tọ`** | 4.5 | `azọ́nwa` |
| nukunpedonugotọ | **`nukunpedonugo-tọ`** | 4.5 | `nukunpedonugo` |
| weplọnmẹtọ | **`weplọnmẹ-tọ`** | 4.5 | `weplọnmẹ` |
| alọgọnamẹtọ | **`alọgọnamẹ-tọ`** | 4.5 | `alọgọnamẹ` |
| togbogantọ | **`togbog-an-tọ`** | 3.0 | `togbog` |
| linlinwekantọ | **`linlinwek-an-tọ`** | 3.0 | `linlinwek` |
| whenuhokàntọ | **`whenuhokàn-tọ`** | 1.5 | `whenuhokàn` |
| avọ́sinsan | **`avọ́sins-an`** | 1.5 | `avọ́sins` |
| koewhèdopotọ | **`koewhèdopo-tọ`** | 1.5 | `koewhèdopo` |
| walɔyizan | **`walɔyiz-an`** | 1.5 | `walɔyiz` |

### 6.6 Linguistic Interpretation

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

---
## 7. Summary & Recommendations

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

### Production Recommendations

| Component | Recommended | Rationale |
|-----------|-------------|-----------|
| Tokenizer | **64k BPE** | Best compression (4.34x) |
| N-gram | **2-gram** | Lowest perplexity (287) |
| Markov | **Context-4** | Highest predictability (95.1%) |
| 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: 2026-01-10 00:40:48*