README.md

---
language: fon
language_name: Fon
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: 4.124
  - name: best_isotropy
    type: isotropy
    value: 0.6254
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-04
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Fon** 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** | 3.633x | 3.64 | 0.1627% | 178,834 |
| **16k** | 3.846x | 3.85 | 0.1723% | 168,913 |
| **32k** | 4.057x | 4.06 | 0.1817% | 160,142 |
| **64k** | 4.124x 🏆 | 4.13 | 0.1847% | 157,541 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Koffi Danger, ɔ́ nyí malànhwlɛ̀nvlɛ́tɔ́ Benɛɛ tɔn ɖé wɛ bɔ è jì i ɖò ɖò Gbɔ̀xikɔ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁koffi ▁dan ger , ▁ɔ́ ▁nyí ▁malànhwlɛ̀nvlɛ́ tɔ́ ▁benɛɛ ▁tɔn ... (+19 more)` | 29 |
| 16k | `▁koffi ▁danger , ▁ɔ́ ▁nyí ▁malànhwlɛ̀nvlɛ́ tɔ́ ▁benɛɛ ▁tɔn ▁ɖé ... (+18 more)` | 28 |
| 32k | `▁koffi ▁danger , ▁ɔ́ ▁nyí ▁malànhwlɛ̀nvlɛ́ tɔ́ ▁benɛɛ ▁tɔn ▁ɖé ... (+18 more)` | 28 |
| 64k | `▁koffi ▁danger , ▁ɔ́ ▁nyí ▁malànhwlɛ̀nvlɛ́ tɔ́ ▁benɛɛ ▁tɔn ▁ɖé ... (+18 more)` | 28 |

**Sample 2:** `Kuwanwangu nyi glekɔxwe ɖokpo nǔ tokpɔnlavi Kwaba tɔn nú tokpɔnla Natitingu tɔn ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ku wan wan gu ▁nyi ▁glekɔxwe ▁ɖokpo ▁nǔ ▁tokpɔnlavi ▁kwaba ... (+12 more)` | 22 |
| 16k | `▁kuwanwangu ▁nyi ▁glekɔxwe ▁ɖokpo ▁nǔ ▁tokpɔnlavi ▁kwaba ▁tɔn ▁nú ▁tokpɔnla ... (+9 more)` | 19 |
| 32k | `▁kuwanwangu ▁nyi ▁glekɔxwe ▁ɖokpo ▁nǔ ▁tokpɔnlavi ▁kwaba ▁tɔn ▁nú ▁tokpɔnla ... (+9 more)` | 19 |
| 64k | `▁kuwanwangu ▁nyi ▁glekɔxwe ▁ɖokpo ▁nǔ ▁tokpɔnlavi ▁kwaba ▁tɔn ▁nú ▁tokpɔnla ... (+9 more)` | 19 |

**Sample 3:** `Ablu ɔ hwenu e minyɔ̀ alo weziza han ɔ wɛ nɔ nyi mɔ̌. Xixa tɔn`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ab lu ▁ɔ ▁hwenu ▁e ▁min yɔ̀ ▁alo ▁weziza ▁han ... (+8 more)` | 18 |
| 16k | `▁ablu ▁ɔ ▁hwenu ▁e ▁minyɔ̀ ▁alo ▁weziza ▁han ▁ɔ ▁wɛ ... (+6 more)` | 16 |
| 32k | `▁ablu ▁ɔ ▁hwenu ▁e ▁minyɔ̀ ▁alo ▁weziza ▁han ▁ɔ ▁wɛ ... (+6 more)` | 16 |
| 64k | `▁ablu ▁ɔ ▁hwenu ▁e ▁minyɔ̀ ▁alo ▁weziza ▁han ▁ɔ ▁wɛ ... (+6 more)` | 16 |


### Key Findings

- **Best Compression:** 64k achieves 4.124x compression
- **Lowest UNK Rate:** 8k with 0.1627% 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 | 1,671 | 10.71 | 7,538 | 38.1% | 71.7% |
| **2-gram** | Subword | 265 🏆 | 8.05 | 2,254 | 68.9% | 98.7% |
| **3-gram** | Word | 2,808 | 11.46 | 12,455 | 33.4% | 62.3% |
| **3-gram** | Subword | 1,585 | 10.63 | 14,789 | 35.7% | 77.3% |
| **4-gram** | Word | 3,755 | 11.87 | 19,739 | 32.3% | 58.3% |
| **4-gram** | Subword | 5,749 | 12.49 | 55,463 | 22.8% | 55.5% |
| **5-gram** | Word | 2,983 | 11.54 | 15,474 | 34.1% | 61.1% |
| **5-gram** | Subword | 12,261 | 13.58 | 96,928 | 17.0% | 44.8% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `tɔn mɛ` | 7,028 |
| 2 | `mɛ ɖo` | 3,347 |
| 3 | `tɔn lɛ` | 2,790 |
| 4 | `mɛ e` | 2,133 |
| 5 | `dodo tɔn` | 1,886 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `tɔn mɛ ɖo` | 2,782 |
| 2 | `jì é ɖěè` | 1,274 |
| 3 | `ayi e jì` | 1,171 |
| 4 | `tɔn mɛ é` | 1,170 |
| 5 | `e jì é` | 1,168 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ayi e jì é` | 1,167 |
| 2 | `e jì é ɖěè` | 1,157 |
| 3 | `e ɖěè mɛ e` | 1,134 |
| 4 | `gbɛtɔ e ɖěè mɛ` | 1,133 |
| 5 | `tɔn mɛ ɖo benɛ` | 1,090 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ayi e jì é ɖěè` | 1,156 |
| 2 | `gbɛtɔ e ɖěè mɛ e` | 1,133 |
| 3 | `benɛ ayi e jì é` | 1,064 |
| 4 | `ɖo benɛ ayi e jì` | 1,060 |
| 5 | `mɛ ɖo benɛ ayi e` | 1,060 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n _` | 58,568 |
| 2 | `o _` | 46,161 |
| 3 | `_ t` | 45,106 |
| 4 | `ɔ n` | 41,894 |
| 5 | `_ ɖ` | 36,979 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ɔ n _` | 27,349 |
| 2 | `t ɔ n` | 25,832 |
| 3 | `_ t ɔ` | 24,140 |
| 4 | `_ ɖ o` | 19,620 |
| 5 | `ɖ o _` | 17,028 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ t ɔ n` | 23,518 |
| 2 | `t ɔ n _` | 22,408 |
| 3 | `_ ɖ o _` | 16,782 |
| 4 | `_ m ɛ _` | 10,812 |
| 5 | `k p ɔ n` | 8,817 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ t ɔ n _` | 20,896 |
| 2 | `_ t o k p` | 8,408 |
| 3 | `t o k p ɔ` | 8,400 |
| 4 | `o k p ɔ n` | 8,400 |
| 5 | `t ɔ n _ m` | 7,246 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 265
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~45% 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 | 0.7272 | 1.655 | 4.51 | 24,791 | 27.3% |
| **1** | Subword | 1.2806 | 2.429 | 14.66 | 265 | 0.0% |
| **2** | Word | 0.2756 | 1.210 | 1.70 | 111,357 | 72.4% |
| **2** | Subword | 1.1501 | 2.219 | 7.00 | 3,884 | 0.0% |
| **3** | Word | 0.1152 | 1.083 | 1.21 | 188,520 | 88.5% |
| **3** | Subword | 0.7806 | 1.718 | 3.61 | 27,160 | 21.9% |
| **4** | Word | 0.0471 🏆 | 1.033 | 1.08 | 227,466 | 95.3% |
| **4** | Subword | 0.5178 | 1.432 | 2.22 | 98,034 | 48.2% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `tɔn mɛ wli hwe ɔ huzu tokpɔnlavi agɔnkanmɛ tɔn bo ɖyɔ ɛ ylɔ ɛ ɖo yovogbè`
2. `ɖo tokpɔn alibori e nɔ̀ kpénukún tovixixa wǔ é kpo hɛnnu mɛ bo nɔ nyì do`
3. `e ɖo lé e é mɛ xwédo 1 lɛ nukɔnnɔtɔ hwɛxo tɔn ayi e yovo hwan`

**Context Size 2:**

1. `tɔn mɛ ɖò totaligbé gbadahweji benɛɛtò tɔn lɛ mi na mɔ xogbè to ɔ tɔn ɖo tantɔn`
2. `mɛ ɖo benɛ ayi e jì é ɖěè lěè akpɔkpɔ ɖé ɖe ɔ è sɔ ɛ ɖɛmɛnu`
3. `mɛ e lɛ́zun gletoxo do sɛ̀nxwĭ jí sin azan ayizin 6 xwejisùn léxwé tɔn mɛ toxoɖɔgbɛ tɔn`

**Context Size 3:**

1. `tɔn mɛ ɖo atacora e lɛ́ nyi gletoxo do sɛ̀nxwĭ jí sin azan ayizin 6 xwejisùn lé xwélé`
2. `jì é ɖěè zinvie ɖo tokpɔnlavi zinvié tɔn mɛ ɖo benɛɛto mɛ bo nyi sɔmi sɔmi ɖɛ̌mɛnu lɛ`
3. `ayi e jì é ɖěè tokpɔnlávì tayaku tɔn ɔ nyi tokpɔnlavi ɖokpo ɖo wò 10 ě ɖo tokpɔnla`

**Context Size 4:**

1. `ayi e jì é ɖěè dovogon ɖo tokpɔnlavi zogbodomey tɔn mɛ ɖo zou e lɛ́ nyí gletoxo ɖò sɛ̀nxwí`
2. `e jì é ɖěè bouhanrou ɖo tokpɔnlavi gomparou tɔn mɛ ɖo alibori e lɛ́ nyi gletoxo ɖo sɛ̀nxwĭ jí`
3. `e ɖěè mɛ e axɔsuxwe insae instad e nɔ̀n kpé nunkún tovixixa wǔ é lɛn xɔta 248 nǔ gbɛtɔ`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_be,._ɖo_ɖěè_e"_`
2. `nɔn_kuɖoudoku_to`
3. `o_é_mbe_gblɛn_ɖò`

**Context Size 2:**

1. `n_kan)_xɔtan_è_ɖo`
2. `o_tɔnla_akanɖie_ɖ`
3. `_tokpé_dodo_tɛntr`

**Context Size 3:**

1. `ɔn_atlant_dolore_t`
2. `tɔn_ɖó_azinkpo_ɔ,_`
3. `_tɔn_ɔ_tɔn_léxwé_d`

**Context Size 4:**

1. `_tɔn._ɖo_tokpɔn_atu`
2. `tɔn_lɛ_sin_azǎn_20ɔ`
3. `_ɖo_tokpɔnlavi_tɔn,`


### Key Findings

- **Best Predictability:** Context-4 (word) with 95.3% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (98,034 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 | 11,148 |
| Total Tokens | 363,048 |
| Mean Frequency | 32.57 |
| Median Frequency | 3 |
| Frequency Std Dev | 405.71 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | tɔn | 23,451 |
| 2 | ɖo | 16,822 |
| 3 | e | 15,001 |
| 4 | mɛ | 14,011 |
| 5 | é | 10,488 |
| 6 | ɔ | 10,251 |
| 7 | lɛ | 8,160 |
| 8 | nyi | 5,259 |
| 9 | nɔ | 5,214 |
| 10 | ɖò | 4,492 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | rust | 2 |
| 2 | gnu | 2 |
| 3 | programme | 2 |
| 4 | java | 2 |
| 5 | api | 2 |
| 6 | columns | 2 |
| 7 | break | 2 |
| 8 | inside | 2 |
| 9 | avoid | 2 |
| 10 | greek | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.1833 |
| R² (Goodness of Fit) | 0.993854 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 63.7% |
| Top 1,000 | 86.2% |
| Top 5,000 | 95.8% |
| Top 10,000 | 99.4% |

### Key Findings

- **Zipf Compliance:** R²=0.9939 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 63.7% of corpus
- **Long Tail:** 1,148 words needed for remaining 0.6% 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 | 0.6254 🏆 | 0.3950 | N/A | N/A |
| **mono_64d** | 64 | 0.3309 | 0.3691 | N/A | N/A |
| **mono_128d** | 128 | 0.0582 | 0.3829 | N/A | N/A |
| **aligned_32d** | 32 | 0.6254 | 0.3991 | 0.0100 | 0.1180 |
| **aligned_64d** | 64 | 0.3309 | 0.3687 | 0.0300 | 0.1420 |
| **aligned_128d** | 128 | 0.0582 | 0.3777 | 0.0520 | 0.2300 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.6254 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.3821. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 5.2% 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 | **5.000** | High morphological productivity | Reliable analysis |
| Idiomaticity Gap | **0.364** | 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 |
|--------|----------|

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-mɛ` | akwɛnyanumɛ, mimɛ, wùnmɛ |

### 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 |
|------|----------|------------------|----------|
| `okpo` | 1.55x | 21 contexts | xokpo, yokpo, lokpo |
| `ɖokp` | 1.57x | 16 contexts | ɖokpɔ, ɖokpò, ɖokpó |
| `ɔnyi` | 1.72x | 12 contexts | sɔnyi, lɔnyiji, ɖɔnyitɔ |
| `plɔn` | 1.72x | 12 contexts | kplɔn, kplɔnnǔ, kplɔnyi |
| `mɛnu` | 1.74x | 10 contexts | dɛmɛnu, wemɛnu, ɖɛmɛnu |
| `ntɔn` | 1.41x | 16 contexts | tantɔn, tǎntɔn, xɔntɔn |
| `ligb` | 1.67x | 9 contexts | aligbo, taligbé, taligbe |
| `pɔnl` | 1.58x | 10 contexts | kpɔnla, tokpɔnlá, tòkpɔnlà |
| `hwen` | 1.42x | 13 contexts | hwenù, hwenu, hwenú |
| `igbe` | 1.53x | 10 contexts | jigbe, yigbe, igbere |
| `ukun` | 1.53x | 9 contexts | wukun, nukun, bukunbé |
| `tokp` | 1.59x | 8 contexts | tokpn, tokpo, tokpa |

### 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 |
|------|-----------------|------------|------|
| liberiatòmɛ | **`liberiatò-mɛ`** | 4.5 | `liberiatò` |
| gabɔntomɛ | **`gabɔnto-mɛ`** | 4.5 | `gabɔnto` |
| jɔwunjɔjamɛ | **`jɔwunjɔja-mɛ`** | 4.5 | `jɔwunjɔja` |
| flanségbèmɛ | **`flanségbè-mɛ`** | 4.5 | `flanségbè` |
| kplekplemɛ | **`kplekple-mɛ`** | 4.5 | `kplekple` |
| flanségbémɛ | **`flanségbé-mɛ`** | 4.5 | `flanségbé` |
| senegaltòmɛ | **`senegaltò-mɛ`** | 4.5 | `senegaltò` |
| flansetomɛ | **`flanseto-mɛ`** | 4.5 | `flanseto` |
| kplékplémɛ | **`kplékplé-mɛ`** | 4.5 | `kplékplé` |
| avɔɖesinukunmɛ | **`avɔɖesinukun-mɛ`** | 1.5 | `avɔɖesinukun` |
| zogbodomɛ | **`zogbodo-mɛ`** | 1.5 | `zogbodo` |
| nùkplɔnmɛ | **`nùkplɔn-mɛ`** | 1.5 | `nùkplɔn` |
| kotoklomɛ | **`kotoklo-mɛ`** | 1.5 | `kotoklo` |
| adakplamɛ | **`adakpla-mɛ`** | 1.5 | `adakpla` |
| azɔnzunmɛ | **`azɔnzun-mɛ`** | 1.5 | `azɔnzun` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Fon 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 | **64k BPE** | Best compression (4.12x) |
| N-gram | **2-gram** | Lowest perplexity (265) |
| Markov | **Context-4** | Highest predictability (95.3%) |
| 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: 2026-01-04 14:47:03*