README.md

---
language: sn
language_name: Shona
language_family: bantu_eastern
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-bantu_eastern
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: 5.104
  - name: best_isotropy
    type: isotropy
    value: 0.8867
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Shona** 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** | 3.894x | 3.90 | 0.0419% | 276,797 |
| **16k** | 4.324x | 4.33 | 0.0465% | 249,296 |
| **32k** | 4.723x | 4.73 | 0.0508% | 228,204 |
| **64k** | 5.104x 🏆 | 5.11 | 0.0549% | 211,184 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Mukanyiwa (n. Dough). Kuvanga (v. Knead) kureva kukanya. Mitauro yeAfrika Lubwis...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁mu kan yiwa ▁( n . ▁do ugh ). ▁ku ... (+32 more)` | 42 |
| 16k | `▁mu kan yiwa ▁( n . ▁do ugh ). ▁ku ... (+31 more)` | 41 |
| 32k | `▁mu kan yiwa ▁( n . ▁dough ). ▁ku vanga ... (+27 more)` | 37 |
| 64k | `▁mukanyiwa ▁( n . ▁dough ). ▁kuvanga ▁( v . ... (+20 more)` | 30 |

**Sample 2:** `Dhuri munhu anofarira zvekurwa zvibhakerera. Mitauro yeAfrika Dinka vanoti dhuur...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁d hu ri ▁munhu ▁ano farira ▁zveku rwa ▁zvi bha ... (+27 more)` | 37 |
| 16k | `▁dhu ri ▁munhu ▁ano farira ▁zveku rwa ▁zvi bha ke ... (+23 more)` | 33 |
| 32k | `▁dhu ri ▁munhu ▁anofarira ▁zveku rwa ▁zvi bha kerera . ... (+19 more)` | 29 |
| 64k | `▁dhu ri ▁munhu ▁anofarira ▁zveku rwa ▁zvi bha kerera . ... (+18 more)` | 28 |

**Sample 3:** `Mukoko wenyuchi (beehive). Mukoko sezita (a family name).`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁muko ko ▁wen yuchi ▁( bee hi ve ). ▁muko ... (+7 more)` | 17 |
| 16k | `▁mukoko ▁wen yuchi ▁( bee hi ve ). ▁mukoko ▁sezita ... (+5 more)` | 15 |
| 32k | `▁mukoko ▁wenyuchi ▁( beehive ). ▁mukoko ▁sezita ▁( a ▁family ... (+2 more)` | 12 |
| 64k | `▁mukoko ▁wenyuchi ▁( beehive ). ▁mukoko ▁sezita ▁( a ▁family ... (+2 more)` | 12 |


### Key Findings

- **Best Compression:** 64k achieves 5.104x compression
- **Lowest UNK Rate:** 8k with 0.0419% 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 | 2,979 | 11.54 | 21,483 | 40.5% | 58.1% |
| **2-gram** | Subword | 240 🏆 | 7.91 | 3,490 | 69.2% | 99.6% |
| **3-gram** | Word | 2,908 | 11.51 | 32,746 | 46.6% | 57.5% |
| **3-gram** | Subword | 1,817 | 10.83 | 20,969 | 30.6% | 75.4% |
| **4-gram** | Word | 9,269 | 13.18 | 84,970 | 37.5% | 44.8% |
| **4-gram** | Subword | 9,071 | 13.15 | 105,458 | 17.8% | 46.9% |
| **5-gram** | Word | 7,269 | 12.83 | 73,171 | 40.2% | 47.4% |
| **5-gram** | Subword | 28,891 | 14.82 | 288,731 | 12.7% | 34.4% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `mamwe mazwi` | 7,603 |
| 2 | `ari pedyo` | 5,393 |
| 3 | `mazwi ari` | 5,117 |
| 4 | `mitauro yebantu` | 4,853 |
| 5 | `ari pasi` | 4,571 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `mazwi ari pedyo` | 5,088 |
| 2 | `pano panyorwa mazwi` | 4,522 |
| 3 | `zvachose nemazwi ari` | 4,521 |
| 4 | `nemazwi ari pasi` | 4,521 |
| 5 | `kusiyana zvachose nemazwi` | 4,521 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `zvachose nemazwi ari pasi` | 4,521 |
| 2 | `kusiyana zvachose nemazwi ari` | 4,521 |
| 3 | `zvigona kusiyana zvachose nemazwi` | 4,520 |
| 4 | `asi zvinoreva zita zvigona` | 4,519 |
| 5 | `zvinoreva zita zvigona kusiyana` | 4,519 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `kusiyana zvachose nemazwi ari pasi` | 4,521 |
| 2 | `zvigona kusiyana zvachose nemazwi ari` | 4,520 |
| 3 | `asi zvinoreva zita zvigona kusiyana` | 4,519 |
| 4 | `zvinoreva zita zvigona kusiyana zvachose` | 4,519 |
| 5 | `zita zvigona kusiyana zvachose nemazwi` | 4,519 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a _` | 324,415 |
| 2 | `a n` | 217,120 |
| 3 | `i _` | 172,321 |
| 4 | `_ m` | 155,472 |
| 5 | `v a` | 151,544 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ k u` | 96,240 |
| 2 | `_ v a` | 65,419 |
| 3 | `_ m a` | 59,080 |
| 4 | `t i _` | 56,775 |
| 5 | `c h i` | 55,374 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ v a n` | 40,523 |
| 2 | `r e v a` | 40,340 |
| 3 | `e v a _` | 39,808 |
| 4 | `n o t i` | 39,798 |
| 5 | `o t i _` | 39,573 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `r e v a _` | 39,628 |
| 2 | `n o t i _` | 39,196 |
| 3 | `_ v a n o` | 28,589 |
| 4 | `v a n o t` | 26,619 |
| 5 | `a n o t i` | 26,589 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 240
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~34% 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 | 0.7188 | 1.646 | 4.18 | 152,645 | 28.1% |
| **1** | Subword | 0.8114 | 1.755 | 5.35 | 2,484 | 18.9% |
| **2** | Word | 0.1584 | 1.116 | 1.34 | 634,035 | 84.2% |
| **2** | Subword | 0.6907 | 1.614 | 3.98 | 13,296 | 30.9% |
| **3** | Word | 0.0530 | 1.037 | 1.09 | 844,134 | 94.7% |
| **3** | Subword | 0.6592 | 1.579 | 3.47 | 52,844 | 34.1% |
| **4** | Word | 0.0249 🏆 | 1.017 | 1.04 | 908,458 | 97.5% |
| **4** | Subword | 0.6401 | 1.558 | 2.87 | 183,409 | 36.0% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `vanoti dukne v reward evanhu muzimbabwe vanhu vane zita zvigona kusiyana zvachose nemazwi ari pedyo ...`
2. `kureva mutezo wemhuka tsumo emubhaibheri ra hadhisoni inotora ruwa runonzi murombedzi sezita mitauro...`
3. `n cloth and twining plant fananidzai nokuti ichirikadzi a blow sokuti vanorarodutira doro nokugadzir...`

**Context Size 2:**

1. `mamwe mazwi mazulu vanoti tanda n middle mid centre kureva pakati gikyode vanoti karawoo n breakable...`
2. `ari pedyo nezita iri kune mimwe mitauro yeafrica asi zvinoreva zita zvigona kusiyana zvachose nemazw...`
3. `mazwi ari pedyo nezita iri kune mimwe mitauro yeafrica asi zvinoreva zita zvigona kusiyana zvachose ...`

**Context Size 3:**

1. `mazwi ari pedyo nezita iri kune mimwe mitauro yeafrika asi zvinoreva zita zvigona kusiyana zvachose ...`
2. `pano panyorwa mazwi ari pedyo nezita iri kune mimwe mitauro yechibantu asi zvinoreva zita zvigona ku...`
3. `kusiyana zvachose nemazwi ari pasi aya chinangwa ndechekutsvaga zvinoreva mazita zvikurusei kana maz...`

**Context Size 4:**

1. `kusiyana zvachose nemazwi ari pasi aya pane tarisiro yekuti mitauro yeafrika inotodzana mazwi xitson...`
2. `zvachose nemazwi ari pasi aya nyanja inoti tewa kana tiwa adj flat stoop down kureva mushevedzeri ma...`
3. `zvigona kusiyana zvachose nemazwi ari pasi aya pane tarisiro yekuti mitauro yeafrika inotodzana mazw...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_anamurod_ae_je_`
2. `ana_sev._mbany_-`
3. `irotulyo_shuho._`

**Context Size 2:**

1. `a_uye_mutersomo_y`
2. `ani;_dzi_ariremit`
3. `i_kance)._eva_84_`

**Context Size 3:**

1. `_kufambabwe_mitaur`
2. `_vano_pass.)_ostii`
3. `_makwaba_(to_(n._t`

**Context Size 4:**

1. `_vanonzi_chishona_k`
2. `reva_dze_nhasikana_`
3. `eva_munhu_mutsetse_`


### Key Findings

- **Best Predictability:** Context-4 (word) with 97.5% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (183,409 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 | 64,427 |
| Total Tokens | 1,165,961 |
| Mean Frequency | 18.10 |
| Median Frequency | 3 |
| Frequency Std Dev | 259.83 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | vanoti | 26,200 |
| 2 | kureva | 25,672 |
| 3 | n | 18,559 |
| 4 | kana | 17,759 |
| 5 | mazwi | 15,009 |
| 6 | mitauro | 13,604 |
| 7 | to | 12,577 |
| 8 | inoti | 12,418 |
| 9 | ari | 10,962 |
| 10 | iri | 10,923 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | miganho | 2 |
| 2 | nyasalendi | 2 |
| 3 | chebhiritani | 2 |
| 4 | chidzivirwana | 2 |
| 5 | mvumiwa | 2 |
| 6 | epuweti | 2 |
| 7 | rematabelelandi | 2 |
| 8 | pilibhit | 2 |
| 9 | ifuleyisitata | 2 |
| 10 | efreyistata | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 0.9530 |
| R² (Goodness of Fit) | 0.997578 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 38.1% |
| Top 1,000 | 57.5% |
| Top 5,000 | 74.2% |
| Top 10,000 | 81.6% |

### Key Findings

- **Zipf Compliance:** R²=0.9976 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 38.1% of corpus
- **Long Tail:** 54,427 words needed for remaining 18.4% 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 | 0.8867 | 0.2895 | N/A | N/A |
| **mono_64d** | 64 | 0.7943 | 0.2343 | N/A | N/A |
| **mono_128d** | 128 | 0.3518 | 0.2276 | N/A | N/A |
| **aligned_32d** | 32 | 0.8867 🏆 | 0.2975 | 0.0380 | 0.2580 |
| **aligned_64d** | 64 | 0.7943 | 0.2419 | 0.0800 | 0.3400 |
| **aligned_128d** | 128 | 0.3518 | 0.2227 | 0.1160 | 0.4200 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.8867 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2522. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 11.6% 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 | **0.377** | 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 |
|--------|----------|
| `-ku` | kung, kuvaya, kudenguma |
| `-mu` | munazarini, mudyo, musipa |
| `-ma` | maswiswi, masawu, mataya |
| `-m` | maswiswi, mwi, munazarini |
| `-n` | nerokuti, natalensis, numeral |
| `-s` | stuck, sikhatele, sefa |
| `-ne` | nerokuti, nekumakomo, nemuchero |
| `-k` | kihanda, kung, kwokunze |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-a` | akataura, ruzhowa, kihanda |
| `-e` | kwokunze, emergence, chemurume |
| `-i` | chekunzi, nerokuti, maswiswi |
| `-wa` | ruzhowa, gadzirwa, mhangwa |
| `-ra` | akataura, horamvura, yakabundira |
| `-o` | mudyo, kitendo, kwamashoko |
| `-ka` | tyinyuka, kika, chikwaka |
| `-s` | connections, natalensis, chums |

### 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 |
|------|----------|------------------|----------|
| `evan` | 2.31x | 46 contexts | evans, evana, pevane |
| `vanh` | 2.44x | 30 contexts | vanhu, evanhu, navanhu |
| `zvin` | 2.03x | 52 contexts | zvino, zvine, zvina |
| `chir` | 1.53x | 109 contexts | chira, chiri, chiro |
| `anhu` | 2.07x | 29 contexts | vanhu, kanhu, sanhu |
| `itau` | 2.02x | 30 contexts | chitau, mitauto, mitauro |
| `mamw` | 2.49x | 15 contexts | mamwe, mamwi, emamwe |
| `ikwa` | 1.88x | 38 contexts | sikwa, tsikwa, abikwa |
| `taur` | 1.68x | 51 contexts | taura, ataure, taurwa |
| `vach` | 1.98x | 23 contexts | vacho, tevach, zvacho |
| `nore` | 1.61x | 40 contexts | snore, ignore, snorer |
| `orev` | 1.95x | 20 contexts | roreva, yoreva, torevei |

### 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 |
|--------|--------|-----------|----------|
| `-ku` | `-a` | 576 words | kudzamura, kupandana |
| `-ch` | `-a` | 219 words | chasara, chinogochewa |
| `-n` | `-a` | 193 words | nunga, nekutaurirana |
| `-a` | `-a` | 178 words | anoyevedza, amateka |
| `-ku` | `-ra` | 150 words | kudzamura, kuruara |
| `-k` | `-a` | 131 words | kolesa, kudzamura |
| `-mu` | `-a` | 113 words | musasa, mutsuba |
| `-ma` | `-a` | 104 words | machayina, magombedzanwa |
| `-mu` | `-i` | 92 words | muvereki, mupinyi |
| `-n` | `-i` | 80 words | nemasevhisi, ndiamai |

### 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 |
|------|-----------------|------------|------|
| nekuvhara | **`nekuvh-a-ra`** | 7.5 | `a` |
| dindiziri | **`dindiz-i-ri`** | 7.5 | `i` |
| zveruvara | **`zveruv-a-ra`** | 7.5 | `a` |
| mafutanhara | **`mafutanh-a-ra`** | 7.5 | `a` |
| kufuruvara | **`kufuruv-a-ra`** | 7.5 | `a` |
| mbungamabari | **`mbungamab-a-ri`** | 7.5 | `a` |
| caribbean | **`caribbe-a-n`** | 7.5 | `a` |
| runyararo | **`runya-ra-ro`** | 7.5 | `ra` |
| yechidzimai | **`yechidzim-a-i`** | 7.5 | `a` |
| yechirwere | **`yechir-we-re`** | 7.5 | `we` |
| mumashure | **`mu-ma-shure`** | 7.5 | `shure` |
| musandarara | **`musandar-a-ra`** | 7.5 | `a` |
| pekuzvara | **`pekuzv-a-ra`** | 7.5 | `a` |
| kwaanogara | **`kwaanog-a-ra`** | 7.5 | `a` |
| ruambakare | **`ruamba-ka-re`** | 7.5 | `ka` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Shona 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 (5.10x) |
| N-gram | **2-gram** | Lowest perplexity (240) |
| Markov | **Context-4** | Highest predictability (97.5%) |
| 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 21:39:08*