README.md

---
language: kbp
language_name: Kabiyè
language_family: atlantic_gur
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_gur
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: 4.466
  - name: best_isotropy
    type: isotropy
    value: 0.8100
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Kabiyè** 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** | 3.774x | 3.78 | 0.1841% | 414,495 |
| **16k** | 4.034x | 4.04 | 0.1968% | 387,731 |
| **32k** | 4.245x | 4.25 | 0.2071% | 368,493 |
| **64k** | 4.466x 🏆 | 4.47 | 0.2179% | 350,205 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Kimeɣa wiye kɛ kɩyakʋ kagbanzɩ ñɩŋa kpɩtaʋ taa. Kɩkɛ Sarakawaɣ wiye ɛsɩntaa nɛ M...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ki me ɣa ▁wiye ▁kɛ ▁kɩyakʋ ▁kagbanzɩ ▁ñɩŋa ▁kpɩtaʋ ▁taa ... (+15 more)` | 25 |
| 16k | `▁kimeɣa ▁wiye ▁kɛ ▁kɩyakʋ ▁kagbanzɩ ▁ñɩŋa ▁kpɩtaʋ ▁taa . ▁kɩkɛ ... (+11 more)` | 21 |
| 32k | `▁kimeɣa ▁wiye ▁kɛ ▁kɩyakʋ ▁kagbanzɩ ▁ñɩŋa ▁kpɩtaʋ ▁taa . ▁kɩkɛ ... (+11 more)` | 21 |
| 64k | `▁kimeɣa ▁wiye ▁kɛ ▁kɩyakʋ ▁kagbanzɩ ▁ñɩŋa ▁kpɩtaʋ ▁taa . ▁kɩkɛ ... (+11 more)` | 21 |

**Sample 2:** `Aloma fenaɣ kɛ fenaɣ hiu ñɩŋa pɩnaɣ taa. Kɛwɛ Salaŋ fenaɣ ɛsɩntaa nɛ Kamɩŋ fenaɣ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁aloma ▁fenaɣ ▁kɛ ▁fenaɣ ▁hiu ▁ñɩŋa ▁pɩnaɣ ▁taa . ▁kɛwɛ ... (+20 more)` | 30 |
| 16k | `▁aloma ▁fenaɣ ▁kɛ ▁fenaɣ ▁hiu ▁ñɩŋa ▁pɩnaɣ ▁taa . ▁kɛwɛ ... (+19 more)` | 29 |
| 32k | `▁aloma ▁fenaɣ ▁kɛ ▁fenaɣ ▁hiu ▁ñɩŋa ▁pɩnaɣ ▁taa . ▁kɛwɛ ... (+18 more)` | 28 |
| 64k | `▁aloma ▁fenaɣ ▁kɛ ▁fenaɣ ▁hiu ▁ñɩŋa ▁pɩnaɣ ▁taa . ▁kɛwɛ ... (+18 more)` | 28 |

**Sample 3:** `Kpɛlɩ kpɛlɩkɩtʋ kɛ kedeŋa lɛɣtʋ ndʋ tɩñɩnɩɣ se tɩtɩlɩ mbʋ pɩkɛ tɛtɛɛ ñɩm nɛ ɛzɩm...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁kpɛlɩ ▁kpɛlɩ kɩ tʋ ▁kɛ ▁kedeŋa ▁lɛɣtʋ ▁ndʋ ▁tɩ ñɩ ... (+19 more)` | 29 |
| 16k | `▁kpɛlɩ ▁kpɛlɩkɩtʋ ▁kɛ ▁kedeŋa ▁lɛɣtʋ ▁ndʋ ▁tɩ ñɩnɩɣ ▁se ▁tɩ ... (+16 more)` | 26 |
| 32k | `▁kpɛlɩ ▁kpɛlɩkɩtʋ ▁kɛ ▁kedeŋa ▁lɛɣtʋ ▁ndʋ ▁tɩñɩnɩɣ ▁se ▁tɩ tɩlɩ ... (+15 more)` | 25 |
| 64k | `▁kpɛlɩ ▁kpɛlɩkɩtʋ ▁kɛ ▁kedeŋa ▁lɛɣtʋ ▁ndʋ ▁tɩñɩnɩɣ ▁se ▁tɩtɩlɩ ▁mbʋ ... (+14 more)` | 24 |


### Key Findings

- **Best Compression:** 64k achieves 4.466x compression
- **Lowest UNK Rate:** 8k with 0.1841% 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 | 4,663 | 12.19 | 12,056 | 19.7% | 51.7% |
| **2-gram** | Subword | 264 🏆 | 8.05 | 2,105 | 67.2% | 99.4% |
| **3-gram** | Word | 7,434 | 12.86 | 14,539 | 12.1% | 42.0% |
| **3-gram** | Subword | 1,733 | 10.76 | 15,395 | 31.4% | 76.5% |
| **4-gram** | Word | 10,847 | 13.40 | 20,789 | 13.1% | 35.7% |
| **4-gram** | Subword | 7,524 | 12.88 | 63,955 | 16.9% | 48.7% |
| **5-gram** | Word | 5,747 | 12.49 | 12,317 | 19.5% | 45.4% |
| **5-gram** | Subword | 21,059 | 14.36 | 129,546 | 10.9% | 33.2% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `taa lɛ` | 2,665 |
| 2 | `ɛjaɖɛ taa` | 1,955 |
| 3 | `taa nɛ` | 1,862 |
| 4 | `payaɣ se` | 1,402 |
| 5 | `ndɩ ndɩ` | 1,291 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ɛjaɖɛ ɖɩnɛ ɖɩ` | 472 |
| 2 | `mbʊ pʊyɔɔ yɔ` | 344 |
| 3 | `nɖɩ ɖɩ taa` | 308 |
| 4 | `ŋga ka taa` | 292 |
| 5 | `ndʊ tɩ taa` | 286 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ɛjaɖɛ ɖɩnɛ ɖɩ taa` | 259 |
| 2 | `ɛjaɖɛ nɖɩ ɖɩ taa` | 156 |
| 3 | `pɩnaɣ ŋga ka taa` | 144 |
| 4 | `ɖɩnɛ ɖɩ taa lɛ` | 139 |
| 5 | `ŋga ka taa kɛ` | 135 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ɛjaɖɛ ɖɩnɛ ɖɩ taa lɛ` | 137 |
| 2 | `pɩnaɣ ŋga ka taa kɛ` | 118 |
| 3 | `fenaɣ ɖomaɣ fenaɣ agoza fenaɣ` | 117 |
| 4 | `lakɩŋ fenaɣ ɖomaɣ fenaɣ agoza` | 117 |
| 5 | `fenaɣ kamɩŋ fenaɣ saŋayɩŋ fenaɣ` | 116 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a _` | 83,584 |
| 2 | `ɛ _` | 81,574 |
| 3 | `_ p` | 59,307 |
| 4 | `a a` | 55,348 |
| 5 | `_ k` | 55,328 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a a _` | 36,136 |
| 2 | `n ɛ _` | 30,041 |
| 3 | `_ n ɛ` | 27,234 |
| 4 | `t a a` | 25,484 |
| 5 | `_ t a` | 23,580 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ n ɛ _` | 26,590 |
| 2 | `_ t a a` | 19,890 |
| 3 | `t a a _` | 18,248 |
| 4 | `n a ɣ _` | 9,933 |
| 5 | `_ s e _` | 9,465 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ t a a _` | 14,437 |
| 2 | `_ n ɛ _ p` | 7,151 |
| 3 | `a _ n ɛ _` | 5,925 |
| 4 | `ɛ j a ɖ ɛ` | 5,595 |
| 5 | `ɩ n a ɣ _` | 5,587 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 264
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~33% 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 | 0.7357 | 1.665 | 5.08 | 43,639 | 26.4% |
| **1** | Subword | 1.1604 | 2.235 | 8.95 | 577 | 0.0% |
| **2** | Word | 0.2778 | 1.212 | 1.70 | 221,221 | 72.2% |
| **2** | Subword | 1.0063 | 2.009 | 5.82 | 5,164 | 0.0% |
| **3** | Word | 0.0968 | 1.069 | 1.17 | 374,524 | 90.3% |
| **3** | Subword | 0.8237 | 1.770 | 3.76 | 30,035 | 17.6% |
| **4** | Word | 0.0352 🏆 | 1.025 | 1.05 | 437,756 | 96.5% |
| **4** | Subword | 0.5901 | 1.505 | 2.44 | 112,917 | 41.0% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `nɛ powoki pruksɛɛlɩ tɛtʊ taa ana pɩlɩna pʊtʊ nɔyʊ cɔlɔ mbʊ papazɩ tʋ nɔɔyʋ eekeŋna ɩ`
2. `taa sɩnɩma tʊma sakɩyɛ sakɩyɛ ayaba wɛɛ ana yɔ takayaɣ kiɖeɖeɣa taa yɔ pɩ tɛ paɣtʋ`
3. `yɔ kɛ tomisi tɛtʊ ciidiɣna lɩm wɛɛ nɛ ɛ taabalʊ caacibeɣa taa ɛyʋ ɛlaba pɩnzɩ naadozo`

**Context Size 2:**

1. `taa lɛ apple lɛɣtʋ kɩfatʋ yaa sɔnɔ mba nabɛyɩ kɔyɔ hɩlaɣ nɛ sakɩyɛ taa category lɛɣtʋ`
2. `ɛjaɖɛ taa pɛlɔ ɖoɖoo agatha christie nɛ jules verne pɛɖɛna ɩ sibérie narym tɛtʊ taa théodule ribot`
3. `taa nɛ sonarwa tɛtʋ taa ajɛya 42 taa tɛtʊ cikpetʊ natʊyʊ nɛ etazuunii ɛjaɖɛ ɖɩnɛ ɖɩ halanzɩ`

**Context Size 3:**

1. `ɛjaɖɛ ɖɩnɛ ɖɩ ɛjaɖɛ nɛ ajɛɛ lɛɛna kpeekpe pasɩna ɖama kamasɩ piresiili ɛjaɖɛ kɛwɛ yomiye taa nɛ awɛɛ`
2. `mbʊ pʊyɔɔ yɔ kɩhaɣa ɖoŋ ɖɩkpaɣ ɛzɩ pɩnaɣ alɩwaatʊ antoine césar becquerel suzuu mbʊ karɩbɔnɩ kaakɛ k...`
3. `nɖɩ ɖɩ taa palʋla ɖajaa sɔsɔ miguel de cervantes saavedra ɛnɛ ɛ hɩɖɛ kʋyɩ siŋŋ pɩlɩɩna ɛmaɣzɩm takay...`

**Context Size 4:**

1. `ɛjaɖɛ ɖɩnɛ ɖɩ taa lɛ paana ɛyaa ɛzɩ miliyɔɔnaa 6 931 071 yɔ nɛ yee pakalɩʊ ɛyaa kɛ kilomɛtanaa`
2. `ɛjaɖɛ nɖɩ ɖɩ taa pɩzɩɣ nɛ pɛlɛdɩɣ ɖama taa tadɩyɛ nɔmɔʊ taa pʊ tʊʊ tobi taa se ɖama hɛkɩŋ`
3. `pɩnaɣ ŋga ka taa ɖɔɖɔ lɛ cpp ŋgbɛyɛ paɣzɩ nesi ɖʋʋ nɛ ɖɩpaɣzɩ maʋ paɣtʋ kɩfatʋ paɖʋ paɣtʋ ndʋ`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_pa_nɛ_ltɩ_mbe_k`
2. `aaabisɔ_tʊ._peŋ_`
3. `ɛ_pakalɩ-hadɔɔ_t`

**Context Size 2:**

1. `a_yɔ_yɔ_pena_wɛ_v`
2. `ɛ_fekpeetiidiyele`
3. `_patepaa_sɩ_apɩna`

**Context Size 3:**

1. `aa_tɩ-yɔɔ_kɛ_ɛwɛ_n`
2. `nɛ_pɩtalɩnaa_sii_ɛ`
3. `_nɛ_pɔyɔ._tɛtʋ_way`

**Context Size 4:**

1. `_nɛ_wɩsɩ_(célering_`
2. `_taa._londre_sukuli`
3. `taa_tɛtʋ_wandamm_ka`


### Key Findings

- **Best Predictability:** Context-4 (word) with 96.5% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (112,917 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 | 17,479 |
| Total Tokens | 477,906 |
| Mean Frequency | 27.34 |
| Median Frequency | 4 |
| Frequency Std Dev | 345.24 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | nɛ | 26,735 |
| 2 | taa | 23,518 |
| 3 | yɔ | 15,303 |
| 4 | se | 9,792 |
| 5 | lɛ | 8,015 |
| 6 | kɛ | 6,975 |
| 7 | ɛjaɖɛ | 5,550 |
| 8 | yɔɔ | 5,505 |
| 9 | pɩnaɣ | 5,287 |
| 10 | ɛ | 4,794 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | admira | 2 |
| 2 | mário | 2 |
| 3 | fernandes | 2 |
| 4 | graça | 2 |
| 5 | housna | 2 |
| 6 | corte | 2 |
| 7 | suprema | 2 |
| 8 | cassazione | 2 |
| 9 | kpɛkpɛ | 2 |
| 10 | feltrinelli | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.1819 |
| R² (Goodness of Fit) | 0.995226 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 49.3% |
| Top 1,000 | 77.9% |
| Top 5,000 | 91.8% |
| Top 10,000 | 96.6% |

### Key Findings

- **Zipf Compliance:** R²=0.9952 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 49.3% of corpus
- **Long Tail:** 7,479 words needed for remaining 3.4% 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 | 0.8100 🏆 | 0.3163 | N/A | N/A |
| **mono_64d** | 64 | 0.4344 | 0.2959 | N/A | N/A |
| **mono_128d** | 128 | 0.0748 | 0.2853 | N/A | N/A |
| **aligned_32d** | 32 | 0.8100 | 0.3232 | 0.0260 | 0.1360 |
| **aligned_64d** | 64 | 0.4344 | 0.2914 | 0.0180 | 0.1780 |
| **aligned_128d** | 128 | 0.0748 | 0.2971 | 0.0500 | 0.2020 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.8100 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.3015. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 5.0% 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 | **0.371** | 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 |
|--------|----------|
| `-k` | kɩwɩlaɣ, kpoŋgbolo, kʊɖʊʊ |
| `-p` | pɩɖɔma, pɩnsɩ, pahɩʊ |
| `-pa` | pahɩʊ, patʊlɩɣ, paayɔda |
| `-s` | sʊzʊʊ, sklodowska, super |
| `-a` | apama, agbaa, ajɛɛ |
| `-t` | tuurkii, tobiyasi, toofɛŋna |
| `-m` | margrethe, malɩtɩ, mabɩyaa |
| `-ka` | kalʊbɩna, kata, kan̄azɩɣ |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-a` | halʊpɩɣa, pɩɖɔma, apama |
| `-ɩ` | pɩnsɩ, pɔritigalɩ, arabɩ |
| `-i` | tuurkii, ruusi, gueorgui |
| `-e` | margrethe, pɩerre, fefere |
| `-na` | kalʊbɩna, pɩtʊʊzɩna, toofɛŋna |
| `-aa` | agbaa, pɩpaɣlaa, kpaaa |
| `-ʊ` | sʊzʊʊ, pahɩʊ, pɛkpɛlɛkʊ |
| `-ɣ` | kɩwɩlaɣ, ɛmaɣmaɣ, kodudaɣ |

### 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 |
|------|----------|------------------|----------|
| `pɛnd` | 1.80x | 67 contexts | kpɛndʊ, kpɛndʋ, kpɛndɩ |
| `kpɛn` | 1.78x | 58 contexts | kpɛndʊ, kpɛndʋ, kpɛnaʋ |
| `yɔɔd` | 1.70x | 66 contexts | yɔɔdʊ, yɔɔda, yɔɔdɩ |
| `maɣz` | 1.61x | 46 contexts | maɣzʊ, maɣzm, maɣzɩ |
| `ɛlɛk` | 1.97x | 21 contexts | kpɛlɛkʋ, kpɛlɛkʊ, kpɛlɛkɩ |
| `ɩlɩn` | 1.76x | 26 contexts | ɩlɩna, pɩlɩnɛ, wɩlɩna |
| `aɣzɩ` | 1.38x | 57 contexts | maɣzɩ, paɣzɩ, ñaɣzɩɣ |
| `kpɛl` | 1.88x | 18 contexts | kpɛlɛ, kpɛlɩ, kpɛlɛkʋ |
| `mɩyɛ` | 1.87x | 16 contexts | kamɩyɛ, nɩmɩyɛ, camɩyɛ |
| `ɩŋga` | 1.48x | 26 contexts | ñɩŋga, tɩŋga, cɩŋga |
| `kuli` | 1.66x | 17 contexts | kulii, ŋkuli, ekuli |
| `ɩnaɣ` | 1.62x | 17 contexts | mɩnaɣ, kɩnaɣ, tɩnaɣ |

### 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 |
|--------|--------|-----------|----------|
| `-p` | `-a` | 226 words | pɩta, pɩkɛdʊna |
| `-k` | `-a` | 174 words | katamsɩna, kʊya |
| `-p` | `-na` | 149 words | pɩkɛdʊna, pɩtɛkɛna |
| `-p` | `-ɣ` | 118 words | pɔlɔwaɣ, pamaɣwaɣ |
| `-k` | `-ɣ` | 107 words | keɖeyaɣ, kakɩlɩɣ |
| `-k` | `-ʊ` | 101 words | kɩɖalʊʊ, kpɛʊ |
| `-p` | `-ɩ` | 95 words | pasɩŋgɩ, pɩtatɩɩ |
| `-k` | `-ɩ` | 90 words | kanɩɩ, kadanzɩ |
| `-a` | `-a` | 61 words | anasayɩnaa, aŋgolaa |
| `-p` | `-ʊ` | 60 words | papɩsʊʊ, pamaɣzʊ |

### 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 |
|------|-----------------|------------|------|
| naakomnaa | **`naakom-na-a`** | 7.5 | `na` |
| kɩnatɩnaa | **`kɩ-na-tɩnaa`** | 7.5 | `tɩnaa` |
| afrikansi | **`afrika-n-si`** | 7.5 | `n` |
| fideyonaa | **`fideyo-na-a`** | 7.5 | `na` |
| raadiyoonaa | **`raadiyoo-na-a`** | 7.5 | `na` |
| miiliyarɩ | **`miiliy-a-rɩ`** | 7.5 | `a` |
| kondolokonaa | **`kondoloko-na-a`** | 7.5 | `na` |
| fɔɔfɔɔnaa | **`fɔɔfɔɔ-na-a`** | 7.5 | `na` |
| lanhɛzɩyɛ | **`la-n-hɛzɩyɛ`** | 7.5 | `hɛzɩyɛ` |
| kɩkpɛndasɩ | **`kɩkpɛnd-a-sɩ`** | 7.5 | `a` |
| ɖamasɩnaʋ | **`ɖamasɩ-na-ʋ`** | 7.5 | `na` |
| kɛgbɛdasɩ | **`kɛgbɛd-a-sɩ`** | 7.5 | `a` |
| pakʋyʋʋna | **`pa-kʋyʋʋ-na`** | 6.0 | `kʋyʋʋ` |
| wilhelmine | **`wilhelm-i-ne`** | 6.0 | `wilhelm` |
| pefezuuna | **`pe-fezuu-na`** | 6.0 | `fezuu` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Kabiyè 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 (4.47x) |
| N-gram | **2-gram** | Lowest perplexity (264) |
| Markov | **Context-4** | Highest predictability (96.5%) |
| Embeddings | **100d** | Balanced semantic capture and isotropy |


---
## Appendix: Metrics Glossary & Interpretation Guide

This section provides definitions, intuitions, and guidance for interpreting the metrics used throughout this report.

### Tokenizer Metrics

**Compression Ratio**
> *Definition:* The ratio of characters to tokens (chars/token). Measures how efficiently the tokenizer represents text.
>
> *Intuition:* Higher compression means fewer tokens needed to represent the same text, reducing sequence lengths for downstream models. A 3x compression means ~3 characters per token on average.
>
> *What to seek:* Higher is generally better for efficiency, but extremely high compression may indicate overly aggressive merging that loses morphological information.

**Average Token Length (Fertility)**
> *Definition:* Mean number of characters per token produced by the tokenizer.
>
> *Intuition:* Reflects the granularity of tokenization. Longer tokens capture more context but may struggle with rare words; shorter tokens are more flexible but increase sequence length.
>
> *What to seek:* Balance between 2-5 characters for most languages. Arabic/morphologically-rich languages may benefit from slightly longer tokens.

**Unknown Token Rate (OOV Rate)**
> *Definition:* Percentage of tokens that map to the unknown/UNK token, indicating words the tokenizer cannot represent.
>
> *Intuition:* Lower OOV means better vocabulary coverage. High OOV indicates the tokenizer encounters many unseen character sequences.
>
> *What to seek:* Below 1% is excellent; below 5% is acceptable. BPE tokenizers typically achieve very low OOV due to subword fallback.

### N-gram Model Metrics

**Perplexity**
> *Definition:* Measures how "surprised" the model is by test data. Mathematically: 2^(cross-entropy). Lower values indicate better prediction.
>
> *Intuition:* If perplexity is 100, the model is as uncertain as if choosing uniformly among 100 options at each step. A perplexity of 10 means effectively choosing among 10 equally likely options.
>
> *What to seek:* Lower is better. Perplexity decreases with larger n-grams (more context). Values vary widely by language and corpus size.

**Entropy**
> *Definition:* Average information content (in bits) needed to encode the next token given the context. Related to perplexity: perplexity = 2^entropy.
>
> *Intuition:* High entropy means high uncertainty/randomness; low entropy means predictable patterns. Natural language typically has entropy between 1-4 bits per character.
>
> *What to seek:* Lower entropy indicates more predictable text patterns. Entropy should decrease as n-gram size increases.

**Coverage (Top-K)**
> *Definition:* Percentage of corpus occurrences explained by the top K most frequent n-grams.
>
> *Intuition:* High coverage with few patterns indicates repetitive/formulaic text; low coverage suggests diverse vocabulary usage.
>
> *What to seek:* Depends on use case. For language modeling, moderate coverage (40-60% with top-1000) is typical for natural text.

### Markov Chain Metrics

**Average Entropy**
> *Definition:* Mean entropy across all contexts, measuring average uncertainty in next-word prediction.
>
> *Intuition:* Lower entropy means the model is more confident about what comes next. Context-1 has high entropy (many possible next words); Context-4 has low entropy (few likely continuations).
>
> *What to seek:* Decreasing entropy with larger context sizes. Very low entropy (<0.1) indicates highly deterministic transitions.

**Branching Factor**
> *Definition:* Average number of unique next tokens observed for each context.
>
> *Intuition:* High branching = many possible continuations (flexible but uncertain); low branching = few options (predictable but potentially repetitive).
>
> *What to seek:* Branching factor should decrease with context size. Values near 1.0 indicate nearly deterministic chains.

**Predictability**
> *Definition:* Derived metric: (1 - normalized_entropy) × 100%. Indicates how deterministic the model's predictions are.
>
> *Intuition:* 100% predictability means the next word is always certain; 0% means completely random. Real text falls between these extremes.
>
> *What to seek:* Higher predictability for text generation quality, but too high (>98%) may produce repetitive output.

### Vocabulary & Zipf's Law Metrics

**Zipf's Coefficient**
> *Definition:* The slope of the log-log plot of word frequency vs. rank. Zipf's law predicts this should be approximately -1.
>
> *Intuition:* A coefficient near -1 indicates the corpus follows natural language patterns where a few words are very common and most words are rare.
>
> *What to seek:* Values between -0.8 and -1.2 indicate healthy natural language distribution. Deviations may suggest domain-specific or artificial text.

**R² (Coefficient of Determination)**
> *Definition:* Measures how well the linear fit explains the frequency-rank relationship. Ranges from 0 to 1.
>
> *Intuition:* R² near 1.0 means the data closely follows Zipf's law; lower values indicate deviation from expected word frequency patterns.
>
> *What to seek:* R² > 0.95 is excellent; > 0.99 indicates near-perfect Zipf adherence typical of large natural corpora.

**Vocabulary Coverage**
> *Definition:* Cumulative percentage of corpus tokens accounted for by the top N words.
>
> *Intuition:* Shows how concentrated word usage is. If top-100 words cover 50% of text, the corpus relies heavily on common words.
>
> *What to seek:* Top-100 covering 30-50% is typical. Higher coverage indicates more repetitive text; lower suggests richer vocabulary.

### Word Embedding Metrics

**Isotropy**
> *Definition:* Measures how uniformly distributed vectors are in the embedding space. Computed as the ratio of minimum to maximum singular values.
>
> *Intuition:* High isotropy (near 1.0) means vectors spread evenly in all directions; low isotropy means vectors cluster in certain directions, reducing expressiveness.
>
> *What to seek:* Higher isotropy generally indicates better-quality embeddings. Values > 0.1 are reasonable; > 0.3 is good. Lower-dimensional embeddings tend to have higher isotropy.

**Average Norm**
> *Definition:* Mean magnitude (L2 norm) of word vectors in the embedding space.
>
> *Intuition:* Indicates the typical "length" of vectors. Consistent norms suggest stable training; high variance may indicate some words are undertrained.
>
> *What to seek:* Relatively consistent norms across models. The absolute value matters less than consistency (low std deviation).

**Cosine Similarity**
> *Definition:* Measures angular similarity between vectors, ranging from -1 (opposite) to 1 (identical direction).
>
> *Intuition:* Words with similar meanings should have high cosine similarity. This is the standard metric for semantic relatedness in embeddings.
>
> *What to seek:* Semantically related words should score > 0.5; unrelated words should be near 0. Synonyms often score > 0.7.

**t-SNE Visualization**
> *Definition:* t-Distributed Stochastic Neighbor Embedding - a dimensionality reduction technique that preserves local structure for visualization.
>
> *Intuition:* Clusters in t-SNE plots indicate groups of semantically related words. Spread indicates vocabulary diversity; tight clusters suggest semantic coherence.
>
> *What to seek:* Meaningful clusters (e.g., numbers together, verbs together). Avoid over-interpreting distances - t-SNE preserves local, not global, structure.

### General Interpretation Guidelines

1. **Compare within model families:** Metrics are most meaningful when comparing models of the same type (e.g., 8k vs 64k tokenizer).
2. **Consider trade-offs:** Better performance on one metric often comes at the cost of another (e.g., compression vs. OOV rate).
3. **Context matters:** Optimal values depend on downstream tasks. Text generation may prioritize different metrics than classification.
4. **Corpus influence:** All metrics are influenced by corpus characteristics. Wikipedia text differs from social media or literature.
5. **Language-specific patterns:** Morphologically rich languages (like Arabic) may show different optimal ranges than analytic languages.


### Visualizations Index

| Visualization | Description |
|---------------|-------------|
| Tokenizer Compression | Compression ratios by vocabulary size |
| Tokenizer Fertility | Average token length by vocabulary |
| Tokenizer OOV | Unknown token rates |
| Tokenizer Total Tokens | Total tokens by vocabulary |
| N-gram Perplexity | Perplexity by n-gram size |
| N-gram Entropy | Entropy by n-gram size |
| N-gram Coverage | Top pattern coverage |
| N-gram Unique | Unique n-gram counts |
| Markov Entropy | Entropy by context size |
| Markov Branching | Branching factor by context |
| Markov Contexts | Unique context counts |
| Zipf's Law | Frequency-rank distribution with fit |
| Vocab Frequency | Word frequency distribution |
| Top 20 Words | Most frequent words |
| Vocab Coverage | Cumulative coverage curve |
| Embedding Isotropy | Vector space uniformity |
| Embedding Norms | Vector magnitude distribution |
| Embedding Similarity | Word similarity heatmap |
| Nearest Neighbors | Similar words for key terms |
| t-SNE Words | 2D word embedding visualization |
| t-SNE Sentences | 2D sentence embedding visualization |
| Position Encoding | Encoding method comparison |
| Model Sizes | Storage requirements |
| Performance Dashboard | Comprehensive performance overview |

---
## About This Project

### Data Source

Models trained on [wikipedia-monthly](https://huggingface.co/datasets/omarkamali/wikipedia-monthly) - a monthly snapshot of Wikipedia articles across 300+ languages.

### Project

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

### Maintainer

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

### Citation

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

```bibtex
@misc{wikilangs2025,
  author = {Kamali, Omar},
  title = {Wikilangs: Open NLP Models for Wikipedia Languages},
  year = {2025},
  doi = {10.5281/zenodo.18073153},
  publisher = {Zenodo},
  url = {https://huggingface.co/wikilangs}
  institution = {Omneity Labs}
}
```

### License

MIT License - Free for academic and commercial use.

### Links

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

*Report Date: 2026-01-10 07:22:53*