README.md

---
language: knc
language_name: Central Kanuri
language_family: african_saharan
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-african_saharan
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: 4.582
  - name: best_isotropy
    type: isotropy
    value: 0.7581
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

# Central Kanuri - Wikilangs Models
## Comprehensive Research Report & Full Ablation Study

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Central Kanuri** 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** | 3.865x | 3.87 | 0.1074% | 347,310 |
| **16k** | 4.175x | 4.18 | 0.1160% | 321,490 |
| **32k** | 4.383x | 4.39 | 0.1218% | 306,248 |
| **64k** | 4.582x 🏆 | 4.59 | 0.1273% | 292,925 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Kǝrmai kǝla lardǝbe dǝ shima kǝla lardǝwa gadebe lan amso-a kuru karewa so-a so-...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁kǝrmai ▁kǝla ▁lardǝbe ▁dǝ ▁shima ▁kǝla ▁lardǝwa ▁gadebe ▁lan ▁amso ... (+24 more)` | 34 |
| 16k | `▁kǝrmai ▁kǝla ▁lardǝbe ▁dǝ ▁shima ▁kǝla ▁lardǝwa ▁gadebe ▁lan ▁amso ... (+23 more)` | 33 |
| 32k | `▁kǝrmai ▁kǝla ▁lardǝbe ▁dǝ ▁shima ▁kǝla ▁lardǝwa ▁gadebe ▁lan ▁amso ... (+23 more)` | 33 |
| 64k | `▁kǝrmai ▁kǝla ▁lardǝbe ▁dǝ ▁shima ▁kǝla ▁lardǝwa ▁gadebe ▁lan ▁amso ... (+23 more)` | 33 |

**Sample 2:** `Johnny Cash John R Cash dǝ sha chasambo shi kayama cidi Amerika bǝ kuru kaya ruw...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁john ny ▁c ash ▁john ▁r ▁c ash ▁dǝ ▁sha ... (+19 more)` | 29 |
| 16k | `▁john ny ▁c ash ▁john ▁r ▁c ash ▁dǝ ▁sha ... (+19 more)` | 29 |
| 32k | `▁johnny ▁cash ▁john ▁r ▁cash ▁dǝ ▁sha ▁chasambo ▁shi ▁kayama ... (+16 more)` | 26 |
| 64k | `▁johnny ▁cash ▁john ▁r ▁cash ▁dǝ ▁sha ▁chasambo ▁shi ▁kayama ... (+16 more)` | 26 |

**Sample 3:** `Nasionalism dǝ shima raayi-a letǝgǝ-a do lardǝ-a lardǝ-a kalkalzǝyinma.`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁nas ional ism ▁dǝ ▁shima ▁raayi - a ▁letǝgǝ - ... (+12 more)` | 22 |
| 16k | `▁nas ional ism ▁dǝ ▁shima ▁raayi - a ▁letǝgǝ - ... (+12 more)` | 22 |
| 32k | `▁nas ional ism ▁dǝ ▁shima ▁raayi - a ▁letǝgǝ - ... (+11 more)` | 21 |
| 64k | `▁nasionalism ▁dǝ ▁shima ▁raayi - a ▁letǝgǝ - a ▁do ... (+8 more)` | 18 |


### Key Findings

- **Best Compression:** 64k achieves 4.582x compression
- **Lowest UNK Rate:** 8k with 0.1074% 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 | 4,001 | 11.97 | 9,862 | 23.1% | 50.6% |
| **2-gram** | Subword | 249 🏆 | 7.96 | 1,869 | 69.4% | 99.6% |
| **3-gram** | Word | 4,817 | 12.23 | 9,468 | 19.5% | 45.2% |
| **3-gram** | Subword | 1,863 | 10.86 | 14,091 | 29.8% | 74.9% |
| **4-gram** | Word | 8,323 | 13.02 | 14,018 | 13.9% | 34.8% |
| **4-gram** | Subword | 8,691 | 13.09 | 63,398 | 15.4% | 45.6% |
| **5-gram** | Word | 5,619 | 12.46 | 8,921 | 15.5% | 38.8% |
| **5-gram** | Subword | 24,681 | 14.59 | 137,201 | 10.0% | 30.6% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `saa lan` | 2,887 |
| 2 | `suro saa` | 2,636 |
| 3 | `bǝ lan` | 1,942 |
| 4 | `a kuru` | 1,725 |
| 5 | `ye lan` | 1,254 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `suro saa lan` | 836 |
| 2 | `suro saa yen` | 549 |
| 3 | `lan suro saa` | 420 |
| 4 | `duwun yar laarrin` | 401 |
| 5 | `saa duwun yar` | 373 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `saa duwun yar laarrin` | 356 |
| 2 | `bǝ lan suro saa` | 289 |
| 3 | `saa lan səta ro` | 279 |
| 4 | `lan səta ro saadənan` | 266 |
| 5 | `suro saa duwun yar` | 259 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `saa lan səta ro saadənan` | 250 |
| 2 | `suro saa duwun yar laarrin` | 246 |
| 3 | `lan suro saa duwun yar` | 226 |
| 4 | `bǝ lan suro saa duwun` | 215 |
| 5 | `lan sun nzu notuna ma` | 147 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a _` | 90,022 |
| 2 | `_ k` | 60,586 |
| 3 | `_ s` | 55,437 |
| 4 | `a n` | 52,900 |
| 5 | `e _` | 48,288 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `y e _` | 25,215 |
| 2 | `r o _` | 22,904 |
| 3 | `_ l a` | 22,653 |
| 4 | `l a n` | 19,693 |
| 5 | `_ k ə` | 17,814 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ k u r` | 14,890 |
| 2 | `_ l a n` | 13,928 |
| 3 | `_ s h i` | 12,446 |
| 4 | `l a n _` | 12,391 |
| 5 | `u r o _` | 10,490 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ l a n _` | 10,158 |
| 2 | `_ k u r u` | 9,916 |
| 3 | `k u r u _` | 9,220 |
| 4 | `_ s a a _` | 8,034 |
| 5 | `_ s u r o` | 7,792 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 249
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~31% 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 | 0.7789 | 1.716 | 4.96 | 48,641 | 22.1% |
| **1** | Subword | 1.1205 | 2.174 | 7.93 | 642 | 0.0% |
| **2** | Word | 0.2394 | 1.180 | 1.52 | 239,982 | 76.1% |
| **2** | Subword | 0.9246 | 1.898 | 5.50 | 5,088 | 7.5% |
| **3** | Word | 0.0706 | 1.050 | 1.11 | 362,994 | 92.9% |
| **3** | Subword | 0.8154 | 1.760 | 3.92 | 27,951 | 18.5% |
| **4** | Word | 0.0242 🏆 | 1.017 | 1.04 | 402,605 | 97.6% |
| **4** | Subword | 0.6137 | 1.530 | 2.57 | 109,408 | 38.6% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `a lardəwa asiabe sammaso stade mohammed goni shidə təlam romanbe kureye sun tzus military administra...`
2. `lan shiye bayanna kada ivan dychko a nabtə gomna malorossia ye askərra hutuyedə gozənadə tutsi kada`
3. `ye fimnzə səraanama zeland a halwa bəlin nankaro bakkada dunya bəlin gartəna shidoni ngawolan səta h...`

**Context Size 2:**

1. `saa lan washington college of the group of interparliamentary relations with the chevalier guard reg...`
2. `suro saa lan səta kəntawu marchye saa lan cotulowo kəntawu march saa acker yǝ bikke nəm sawa`
3. `bǝ lan suro nashawa league bǝ manchester city bǝ lan sun nzu notunaman sha chesambo yim fyakkin`

**Context Size 3:**

1. `suro saa lan shiga wakil majalis kuraye ro karrada loktu kərmai nigeria yǝ kən diyau medən gozə kowo...`
2. `suro saa yen loktu kura lardəye arturo umberto illia futu spanish lan bowotin kito r quechua kitu hu...`
3. `lan suro saa lan bərnidə wuratə saa woson kashi 11 5 səwandəna 9 futu razəwuye faraktənadən tubman y...`

**Context Size 4:**

1. `saa duwun yar laarrin findin laarrin bǝ lan kǝntawu razab bǝ lan suro saa duwun yar laarrin findin l...`
2. `bǝ lan suro saa duwun yar laarrin fitulurrin luko uwun bǝ lan sha katambo dekkel baktema cidi urugua...`
3. `saa lan səta ro saadənan demokradiyamen kura lardəye kartəro a saa 16 ro cidazəna kuru shima kamu ku...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_nandəbə_lainza_`
2. `awurso_kə_ku_dǝ.`
3. `niwobewo_sǝruwsə`

**Context Size 2:**

1. `a_ka_ko_aprey-zau`
2. `_kəriero;_shima_i`
3. `_surunyakkaradebe`

**Context Size 3:**

1. `ye_fro-a,_diodəna.`
2. `ro_kada_nəm_greef_`
3. `_lardero_suro_saa_`

**Context Size 4:**

1. `_kuru_nəmnzə-a_lan_`
2. `_lan,_bəladiya_lan_`
3. `_shima_lardə_bəlin_`


### Key Findings

- **Best Predictability:** Context-4 (word) with 97.6% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (109,408 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 | 19,840 |
| Total Tokens | 415,541 |
| Mean Frequency | 20.94 |
| Median Frequency | 3 |
| Frequency Std Dev | 234.13 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | a | 19,508 |
| 2 | lan | 13,848 |
| 3 | ye | 10,032 |
| 4 | kuru | 9,262 |
| 5 | saa | 8,070 |
| 6 | suro | 7,134 |
| 7 | də | 5,746 |
| 8 | bǝ | 4,251 |
| 9 | shima | 3,927 |
| 10 | ro | 3,601 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | rugbyye | 2 |
| 2 | brivero | 2 |
| 3 | chiefs | 2 |
| 4 | nuala | 2 |
| 5 | éireann | 2 |
| 6 | taghmon | 2 |
| 7 | seán | 2 |
| 8 | girton | 2 |
| 9 | ryandə | 2 |
| 10 | ucd | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.1101 |
| R² (Goodness of Fit) | 0.993201 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 42.1% |
| Top 1,000 | 70.5% |
| Top 5,000 | 88.4% |
| Top 10,000 | 94.5% |

### Key Findings

- **Zipf Compliance:** R²=0.9932 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 42.1% of corpus
- **Long Tail:** 9,840 words needed for remaining 5.5% 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 | 0.7581 | 0.3263 | N/A | N/A |
| **mono_64d** | 64 | 0.3103 | 0.3195 | N/A | N/A |
| **mono_128d** | 128 | 0.0510 | 0.3146 | N/A | N/A |
| **aligned_32d** | 32 | 0.7581 🏆 | 0.3414 | 0.0460 | 0.2480 |
| **aligned_64d** | 64 | 0.3103 | 0.3156 | 0.0720 | 0.3240 |
| **aligned_128d** | 128 | 0.0510 | 0.3140 | 0.0860 | 0.4020 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.7581 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.3219. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 8.6% 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 | **0.268** | 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 |
|--------|----------|
| `-a` | abrahamye, alao, awardsbe |
| `-s` | speaker, shawayen, saracenic |
| `-b` | beaumont, b3, bannazəna |
| `-k` | keryə, kəmbuzayin, kla |
| `-m` | manitobayen, maud, mukon |
| `-ma` | manitobayen, maud, magaji |
| `-d` | duro, dunyabewo, darajatin |
| `-c` | chaplin, crew, challenger |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-e` | zutəye, abrahamye, ukeje |
| `-n` | manitobayen, kəmbuzayin, chaplin |
| `-a` | kla, kəlanza, kaza |
| `-ə` | zamanbedə, keryə, kəzəkkə |
| `-be` | awardsbe, afghanistanbe, cathedralbe |
| `-o` | gowono, kadiwo, alao |
| `-ye` | zutəye, abrahamye, disembaye |
| `-də` | zamanbedə, kəradə, matədə |

### 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 |
|------|----------|------------------|----------|
| `zəna` | 1.88x | 142 contexts | azəna, zazəna, lazəna |
| `zana` | 1.90x | 71 contexts | nozana, rozana, rizana |
| `ardə` | 2.06x | 25 contexts | lardə, gardə, lardəa |
| `rmai` | 2.13x | 21 contexts | kǝrmai, kirmai, kərmai |
| `asha` | 1.85x | 33 contexts | jasha, nasha, sasha |
| `andi` | 1.60x | 43 contexts | sandi, fandi, nandi |
| `dəna` | 1.70x | 31 contexts | dənan, tədəna, gadəna |
| `ərma` | 2.00x | 17 contexts | kərma, kərmai, kərmaro |
| `lard` | 1.74x | 23 contexts | lardə, lardu, larde |
| `sand` | 1.73x | 22 contexts | sandi, sanda, sandǝ |
| `ambo` | 1.61x | 21 contexts | tambo, kambo, dambo |
| `nash` | 1.98x | 11 contexts | nasha, nashaa, nashan |

### 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 |
|--------|--------|-----------|----------|
| `-k` | `-a` | 164 words | karewa, kazadalawa |
| `-k` | `-e` | 142 words | kasattəbe, kungiyadəye |
| `-k` | `-n` | 128 words | koktənadən, kǝrǝn |
| `-k` | `-ə` | 120 words | kərmaitədə, kasuwudə |
| `-a` | `-e` | 109 words | augustusbe, alcockye |
| `-s` | `-n` | 96 words | smithsonian, sədin |
| `-k` | `-o` | 87 words | kərazənaro, karəngaro |
| `-s` | `-e` | 84 words | saharanye, samiye |
| `-b` | `-a` | 82 words | bannatəma, bega |
| `-b` | `-n` | 81 words | bernstein, baditənzən |

### 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 |
|------|-----------------|------------|------|
| tǝdǝnaben | **`tǝdǝna-be-n`** | 7.5 | `be` |
| februaryben | **`february-be-n`** | 7.5 | `be` |
| generally | **`general-l-y`** | 7.5 | `l` |
| daurabedə | **`daura-be-də`** | 7.5 | `be` |
| beakerbedə | **`beaker-be-də`** | 7.5 | `be` |
| shaizarbedə | **`shaizar-be-də`** | 7.5 | `be` |
| africaben | **`africa-be-n`** | 7.5 | `be` |
| professorbero | **`professor-be-ro`** | 7.5 | `be` |
| kamuwaben | **`kamuwa-be-n`** | 7.5 | `be` |
| faidatanadə | **`faidata-na-də`** | 7.5 | `na` |
| kərgənbedə | **`kərgən-be-də`** | 7.5 | `be` |
| gargammabe | **`gargam-ma-be`** | 7.5 | `ma` |
| rinderpestbeye | **`rinderpest-be-ye`** | 7.5 | `be` |
| faidatinmawo | **`faidatin-ma-wo`** | 7.5 | `ma` |
| turkeyben | **`turkey-be-n`** | 7.5 | `be` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Central Kanuri 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 (4.58x) |
| N-gram | **2-gram** | Lowest perplexity (249) |
| Markov | **Context-4** | Highest predictability (97.6%) |
| 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 08:01:45*