README.md

---
language: yo
language_name: Yoruba
language_family: atlantic_yoruba_igbo
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_yoruba_igbo
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: 3.758
  - name: best_isotropy
    type: isotropy
    value: 0.8242
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-11
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Yoruba** 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** | 3.147x | 3.15 | 0.2917% | 765,613 |
| **16k** | 3.396x | 3.40 | 0.3147% | 709,643 |
| **32k** | 3.597x | 3.60 | 0.3334% | 669,837 |
| **64k** | 3.758x 🏆 | 3.76 | 0.3482% | 641,232 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `jẹ́ plánẹ́tì kékeré ní ibi ìgbàjá ástẹ́rọ́ìdì. Itokasi ástẹ́rọ́ìdì`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁jẹ́ ▁plánẹ́tì ▁kékeré ▁ní ▁ibi ▁ìgbàjá ▁ástẹ́rọ́ìdì . ▁itokasi ▁ástẹ́rọ́ìdì` | 10 |
| 16k | `▁jẹ́ ▁plánẹ́tì ▁kékeré ▁ní ▁ibi ▁ìgbàjá ▁ástẹ́rọ́ìdì . ▁itokasi ▁ástẹ́rọ́ìdì` | 10 |
| 32k | `▁jẹ́ ▁plánẹ́tì ▁kékeré ▁ní ▁ibi ▁ìgbàjá ▁ástẹ́rọ́ìdì . ▁itokasi ▁ástẹ́rọ́ìdì` | 10 |
| 64k | `▁jẹ́ ▁plánẹ́tì ▁kékeré ▁ní ▁ibi ▁ìgbàjá ▁ástẹ́rọ́ìdì . ▁itokasi ▁ástẹ́rọ́ìdì` | 10 |

**Sample 2:** `je Aare orile-ede Haiti tele. Itokasi Ààrẹ ilẹ̀ Hàítì`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁je ▁aare ▁orile - ede ▁haiti ▁tele . ▁itokasi ▁ààrẹ ... (+2 more)` | 12 |
| 16k | `▁je ▁aare ▁orile - ede ▁haiti ▁tele . ▁itokasi ▁ààrẹ ... (+2 more)` | 12 |
| 32k | `▁je ▁aare ▁orile - ede ▁haiti ▁tele . ▁itokasi ▁ààrẹ ... (+2 more)` | 12 |
| 64k | `▁je ▁aare ▁orile - ede ▁haiti ▁tele . ▁itokasi ▁ààrẹ ... (+2 more)` | 12 |

**Sample 3:** `jẹ́ plánẹ́tì kékeré ní ibi ìgbàjá ástẹ́rọ́ìdì. Itokasi`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁jẹ́ ▁plánẹ́tì ▁kékeré ▁ní ▁ibi ▁ìgbàjá ▁ástẹ́rọ́ìdì . ▁itokasi` | 9 |
| 16k | `▁jẹ́ ▁plánẹ́tì ▁kékeré ▁ní ▁ibi ▁ìgbàjá ▁ástẹ́rọ́ìdì . ▁itokasi` | 9 |
| 32k | `▁jẹ́ ▁plánẹ́tì ▁kékeré ▁ní ▁ibi ▁ìgbàjá ▁ástẹ́rọ́ìdì . ▁itokasi` | 9 |
| 64k | `▁jẹ́ ▁plánẹ́tì ▁kékeré ▁ní ▁ibi ▁ìgbàjá ▁ástẹ́rọ́ìdì . ▁itokasi` | 9 |


### Key Findings

- **Best Compression:** 64k achieves 3.758x compression
- **Lowest UNK Rate:** 8k with 0.2917% 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 | 15,512 | 13.92 | 75,926 | 18.0% | 37.6% |
| **2-gram** | Subword | 467 🏆 | 8.87 | 6,012 | 53.2% | 97.2% |
| **3-gram** | Word | 29,860 | 14.87 | 120,521 | 14.8% | 28.4% |
| **3-gram** | Subword | 4,102 | 12.00 | 51,496 | 19.8% | 59.0% |
| **4-gram** | Word | 59,917 | 15.87 | 214,920 | 13.7% | 22.5% |
| **4-gram** | Subword | 22,011 | 14.43 | 265,494 | 12.0% | 33.3% |
| **5-gram** | Word | 40,150 | 15.29 | 156,085 | 16.5% | 24.8% |
| **5-gram** | Subword | 73,071 | 16.16 | 699,133 | 9.2% | 23.4% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `tí ó` | 19,475 |
| 2 | `ní ibi` | 14,923 |
| 3 | `kékeré ní` | 14,762 |
| 4 | `ibi ìgbàjá` | 14,739 |
| 5 | `ìgbàjá ástẹ́rọ́ìdì` | 14,725 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ní ibi ìgbàjá` | 14,739 |
| 2 | `kékeré ní ibi` | 14,738 |
| 3 | `ibi ìgbàjá ástẹ́rọ́ìdì` | 14,725 |
| 4 | `jẹ́ plánẹ́tì kékeré` | 14,688 |
| 5 | `plánẹ́tì kékeré ní` | 14,688 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `kékeré ní ibi ìgbàjá` | 14,738 |
| 2 | `ní ibi ìgbàjá ástẹ́rọ́ìdì` | 14,725 |
| 3 | `plánẹ́tì kékeré ní ibi` | 14,688 |
| 4 | `jẹ́ plánẹ́tì kékeré ní` | 14,688 |
| 5 | `ibi ìgbàjá ástẹ́rọ́ìdì itokasi` | 14,641 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `kékeré ní ibi ìgbàjá ástẹ́rọ́ìdì` | 14,724 |
| 2 | `plánẹ́tì kékeré ní ibi ìgbàjá` | 14,688 |
| 3 | `jẹ́ plánẹ́tì kékeré ní ibi` | 14,688 |
| 4 | `ní ibi ìgbàjá ástẹ́rọ́ìdì itokasi` | 14,641 |
| 5 | `ibi ìgbàjá ástẹ́rọ́ìdì itokasi ástẹ́rọ́ìdì` | 13,854 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n _` | 450,694 |
| 2 | `i _` | 405,534 |
| 3 | `_ a` | 300,083 |
| 4 | `_ n` | 283,323 |
| 5 | `_ t` | 247,960 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `t i _` | 153,979 |
| 2 | `_ n í` | 105,250 |
| 3 | `_ n i` | 102,296 |
| 4 | `w ọ n` | 90,977 |
| 5 | `ọ n _` | 90,343 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `w ọ n _` | 88,162 |
| 2 | `_ n í _` | 74,812 |
| 3 | `_ n i _` | 74,453 |
| 4 | `_ t i _` | 69,707 |
| 5 | `_ t í _` | 50,988 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `à w ọ n _` | 46,754 |
| 2 | `_ à w ọ n` | 46,122 |
| 3 | `a w ọ n _` | 30,885 |
| 4 | `_ a w ọ n` | 30,498 |
| 5 | `t ẹ́ r ọ́ ì` | 28,695 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 467
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~23% 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 | 0.8773 | 1.837 | 7.00 | 179,072 | 12.3% |
| **1** | Subword | 0.8392 | 1.789 | 6.66 | 2,526 | 16.1% |
| **2** | Word | 0.2998 | 1.231 | 1.81 | 1,250,964 | 70.0% |
| **2** | Subword | 0.8984 | 1.864 | 6.12 | 16,794 | 10.2% |
| **3** | Word | 0.1182 | 1.085 | 1.23 | 2,252,885 | 88.2% |
| **3** | Subword | 0.8307 | 1.779 | 4.43 | 102,698 | 16.9% |
| **4** | Word | 0.0490 🏆 | 1.035 | 1.08 | 2,755,002 | 95.1% |
| **4** | Subword | 0.6691 | 1.590 | 3.04 | 454,606 | 33.1% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `ni ojuiyipo re unje lilo ede nedalandi ó fara jọ ṣe ní òrìṣà ní ibi ìgbàjá`
2. `ní bẹ̀ mí a gbọ́ ni àwọn ẹni pé ayé to lower alpha capture and sun`
3. `ti àwọn ìròyìn òfegè tí ó lọ ti o tun a kìí ṣe ìwádìí tó wá`

**Context Size 2:**

1. `tí ó gbòòrò jùlọ ní orílẹ̀ èdè nàíjírìa ọjọ́ ìbí april 28 jẹ́ gbajúmọ̀ fún àwọ̀ dúdú`
2. `ní ibi ìgbàjá ástẹ́rọ́ìdì itokasi ástẹ́rọ́ìdì vec lista de zachia`
3. `kékeré ní ibi ìgbàjá ástẹ́rọ́ìdì itokasi ástẹ́rọ́ìdì vec lista de yebes`

**Context Size 3:**

1. `ní ibi ìgbàjá ástẹ́rọ́ìdì itokasi ástẹ́rọ́ìdì vec lista de adria`
2. `kékeré ní ibi ìgbàjá ástẹ́rọ́ìdì itokasi ástẹ́rọ́ìdì vec lista de aënna`
3. `ibi ìgbàjá ástẹ́rọ́ìdì itokasi ástẹ́rọ́ìdì vec lista de megaira`

**Context Size 4:**

1. `kékeré ní ibi ìgbàjá ástẹ́rọ́ìdì itokasi ástẹ́rọ́ìdì vec lista de zachia`
2. `ní ibi ìgbàjá ástẹ́rọ́ìdì itokasi ástẹ́rọ́ìdì vec lista de tolkien`
3. `plánẹ́tì kékeré ní ibi ìgbàjá ásítẹ́rọ́ìdì itokasi ástẹ́rọ́ìdì`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_ncan_denla_bíìd`
2. `i_-arẹ̀tuar_ààn_ì`
3. `n),_nín_aunerda_`

**Context Size 2:**

1. `n_ó_sìnlejì_àtò_ì`
2. `i_ìgballe_naind_t`
3. `_africanric_o_unt`

**Context Size 3:**

1. `ti_olùdarí_ìmọ̀_ráí`
2. `_ní_orilẹ_ni_fíìmù`
3. `_nipinle_kway_jẹ́_o`

**Context Size 4:**

1. `wọn_ìtàn_ìmọ̀-ẹ̀rọ_ti`
2. `_ní_èdè_egypt_leade`
3. `_ni_arábìnrin_wọ́n_g`


### 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 (454,606 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 | 79,381 |
| Total Tokens | 3,414,288 |
| Mean Frequency | 43.01 |
| Median Frequency | 4 |
| Frequency Std Dev | 725.10 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ní | 76,550 |
| 2 | ni | 76,509 |
| 3 | ti | 70,538 |
| 4 | tí | 52,513 |
| 5 | ó | 47,903 |
| 6 | àwọn | 46,664 |
| 7 | jẹ́ | 35,696 |
| 8 | o | 34,127 |
| 9 | awọn | 30,834 |
| 10 | ástẹ́rọ́ìdì | 28,681 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | shaik | 2 |
| 2 | ntombela | 2 |
| 3 | fayawọ | 2 |
| 4 | millarworld | 2 |
| 5 | ordinating | 2 |
| 6 | akọyọyọ | 2 |
| 7 | olùgbàlé | 2 |
| 8 | kẹẹẹ́dọ́gbọ̀n | 2 |
| 9 | ìbanilẹ́jẹ́ | 2 |
| 10 | obilor | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.1348 |
| R² (Goodness of Fit) | 0.995636 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 41.3% |
| Top 1,000 | 67.8% |
| Top 5,000 | 83.9% |
| Top 10,000 | 89.3% |

### Key Findings

- **Zipf Compliance:** R²=0.9956 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 41.3% of corpus
- **Long Tail:** 69,381 words needed for remaining 10.7% 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 | 0.8242 🏆 | 0.3333 | N/A | N/A |
| **mono_64d** | 64 | 0.8144 | 0.2438 | N/A | N/A |
| **mono_128d** | 128 | 0.7308 | 0.2103 | N/A | N/A |
| **aligned_32d** | 32 | 0.8242 | 0.3324 | 0.0980 | 0.4180 |
| **aligned_64d** | 64 | 0.8144 | 0.2547 | 0.1840 | 0.5340 |
| **aligned_128d** | 128 | 0.7308 | 0.2109 | 0.2460 | 0.6120 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.8242 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2642. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 24.6% 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 | **0.060** | 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 |
|--------|----------|
| `-a` | advocate, abáyọ, akọbi |
| `-s` | spainclay, spotlite, susanne |
| `-i` | itanka, ifiranšẹ, iléṣa |
| `-o` | onṣẹ, ologe, olagbegi |
| `-k` | kowloon, kobe, kulere |
| `-m` | mẹnuba, melaye, mathew |
| `-l` | láàrí, lẹ́ru, leili |
| `-b` | batman, basemera, bolanle |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-n` | ọlọ́fàgangan, batman, kowloon |
| `-e` | advocate, tope, helaine |
| `-s` | exegesis, dionýsios, aspergillus |
| `-a` | xinhua, mẹnuba, basemera |
| `-i` | níji, akọbi, akinjobi |
| `-o` | dioulasso, adugbo, woyo |
| `-d` | exiled, unsold, spelled |
| `-on` | kowloon, peterson, suggestion |

### 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 |
|------|----------|------------------|----------|
| `ment` | 2.58x | 41 contexts | moment, foment, mental |
| `tion` | 2.39x | 45 contexts | otiono, notion, action |
| `vers` | 2.40x | 41 contexts | verse, versa, ivers |
| `atio` | 2.30x | 36 contexts | ratio, patios, nation |
| `pínl` | 2.90x | 16 contexts | ìpínl, ìpínle, pínlẹ̀ |
| `nter` | 2.19x | 40 contexts | enter, inter, hunter |
| `mber` | 2.31x | 28 contexts | ember, amber, timber |
| `eria` | 2.17x | 34 contexts | neria, seria, iberia |
| `oríl` | 2.57x | 18 contexts | oríle, orílè, orílẹ |
| `iver` | 2.29x | 25 contexts | liver, ivers, river |
| `nìyà` | 2.47x | 19 contexts | nìyàn, ẹnìyàn, enìyàn |
| `ersi` | 2.71x | 13 contexts | persia, persian, persist |

### 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 |
|--------|--------|-----------|----------|
| `-a` | `-n` | 76 words | apáìwọ̀òrùn, amotekun |
| `-a` | `-e` | 63 words | affordable, ape |
| `-a` | `-a` | 54 words | aurora, ayuba |
| `-m` | `-n` | 53 words | mọ̀ọ̀yàn, mẹ́tin |
| `-o` | `-n` | 52 words | omicron, okon |
| `-k` | `-n` | 45 words | kpentomun, kìnnìún |
| `-o` | `-e` | 45 words | onirojinle, owańbe |
| `-s` | `-s` | 42 words | setaleyrodes, seas |
| `-a` | `-s` | 40 words | abbreviations, ages |
| `-o` | `-a` | 40 words | odambea, okúta |

### 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 |
|------|-----------------|------------|------|
| afamefuna | **`afamefu-n-a`** | 7.5 | `n` |
| telifisonu | **`telifis-on-u`** | 7.5 | `on` |
| wenceslaus | **`wencesl-a-us`** | 7.5 | `a` |
| recognise | **`recogni-s-e`** | 7.5 | `s` |
| housemate | **`housem-a-te`** | 7.5 | `a` |
| palæogene | **`palæoge-n-e`** | 7.5 | `n` |
| chimpanzees | **`chimpanz-e-es`** | 7.5 | `e` |
| berlusconi | **`berlusc-on-i`** | 7.5 | `on` |
| questioned | **`questi-on-ed`** | 7.5 | `on` |
| ailagbara | **`a-i-lagbara`** | 7.5 | `lagbara` |
| ibòmìíràn | **`i-b-òmìíràn`** | 6.0 | `òmìíràn` |
| abyssinian | **`abyssinia-n`** | 4.5 | `abyssinia` |
| ìfọwọ́sowọpọ̀ | **`ì-fọwọ́sowọpọ̀`** | 4.5 | `fọwọ́sowọpọ̀` |
| concerted | **`concert-ed`** | 4.5 | `concert` |
| interacts | **`interact-s`** | 4.5 | `interact` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Yoruba 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 (3.76x) |
| N-gram | **2-gram** | Lowest perplexity (467) |
| 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: 2026-01-11 05:59:56*