README.md

---
language: nup
language_name: Nupe-Nupe-Tako
language_family: atlantic_other
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_other
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: 4.182
  - name: best_isotropy
    type: isotropy
    value: 0.0436
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

# Nupe-Nupe-Tako - Wikilangs Models
## Comprehensive Research Report & Full Ablation Study

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Nupe-Nupe-Tako** 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** | 3.745x | 3.75 | 0.1160% | 125,813 |
| **16k** | 4.044x | 4.05 | 0.1253% | 116,510 |
| **32k** | 4.182x 🏆 | 4.19 | 0.1296% | 112,656 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Enna bolu zhi nyan Nasarawa wunyi enna na ge na dan ezhi nin Lafiya'o, Nasarawa....`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁enna ▁bolu ▁zhi ▁nyan ▁nasarawa ▁wunyi ▁enna ▁na ▁ge ▁na ... (+21 more)` | 31 |
| 16k | `▁enna ▁bolu ▁zhi ▁nyan ▁nasarawa ▁wunyi ▁enna ▁na ▁ge ▁na ... (+21 more)` | 31 |
| 32k | `▁enna ▁bolu ▁zhi ▁nyan ▁nasarawa ▁wunyi ▁enna ▁na ▁ge ▁na ... (+19 more)` | 29 |

**Sample 2:** `Bàbò (Lagenaria siceraria)Blench, Roger. Nupe plants and trees: their names and ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁b à b ò ▁( l agen aria ▁s ic ... (+30 more)` | 40 |
| 16k | `▁bàbò ▁( lagenaria ▁sicer aria ) blench , ▁roger . ... (+20 more)` | 30 |
| 32k | `▁bàbò ▁( lagenaria ▁siceraria ) blench , ▁roger . ▁nupe ... (+17 more)` | 27 |

**Sample 3:** `Aisha Muharrar (12 wunga amawuo), wungayi eyankachi yan America Television wunma...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁aisha ▁mu har r ar ▁( 1 2 ▁wunga ▁ama ... (+21 more)` | 31 |
| 16k | `▁aisha ▁mu harrar ▁( 1 2 ▁wunga ▁amawuo ), ▁wungayi ... (+16 more)` | 26 |
| 32k | `▁aisha ▁muharrar ▁( 1 2 ▁wunga ▁amawuo ), ▁wungayi ▁eyankachi ... (+14 more)` | 24 |


### Key Findings

- **Best Compression:** 32k achieves 4.182x compression
- **Lowest UNK Rate:** 8k with 0.1160% 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 | 941 | 9.88 | 1,983 | 37.8% | 81.5% |
| **2-gram** | Subword | 227 🏆 | 7.83 | 1,160 | 69.5% | 99.8% |
| **3-gram** | Word | 1,254 | 10.29 | 2,206 | 30.4% | 72.8% |
| **3-gram** | Subword | 1,537 | 10.59 | 7,263 | 32.0% | 77.7% |
| **4-gram** | Word | 2,126 | 11.05 | 3,106 | 21.3% | 56.3% |
| **4-gram** | Subword | 6,047 | 12.56 | 26,183 | 19.1% | 50.5% |
| **5-gram** | Word | 1,529 | 10.58 | 1,902 | 20.6% | 65.7% |
| **5-gram** | Subword | 12,552 | 13.62 | 42,618 | 14.0% | 38.2% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `wun yi` | 703 |
| 2 | `o nan` | 596 |
| 3 | `ah be` | 579 |
| 4 | `yi o` | 526 |
| 5 | `nan wun` | 439 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `wun yi o` | 454 |
| 2 | `ah man u` | 238 |
| 3 | `yi o nan` | 218 |
| 4 | `nan ah kpeye` | 137 |
| 5 | `ah kpeye be` | 126 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `wun yi o nan` | 187 |
| 2 | `nan ah kpeye be` | 113 |
| 3 | `from the original on` | 100 |
| 4 | `nan wun yi o` | 81 |
| 5 | `wun yi o wun` | 74 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `archived from the original on` | 60 |
| 2 | `kin america wun yi o` | 44 |
| 3 | `wun yi o nan e` | 42 |
| 4 | `nyan kin america wun yi` | 39 |
| 5 | `wun yi o nan de` | 31 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a n` | 16,676 |
| 2 | `n _` | 16,511 |
| 3 | `a _` | 11,948 |
| 4 | `e _` | 9,985 |
| 5 | `_ n` | 9,524 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a n _` | 8,945 |
| 2 | `_ n a` | 4,610 |
| 3 | `n a n` | 4,016 |
| 4 | `u n _` | 3,299 |
| 5 | `y a n` | 3,272 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ n a n` | 3,560 |
| 2 | `_ w u n` | 3,054 |
| 3 | `y a n _` | 2,972 |
| 4 | `n y a n` | 2,846 |
| 5 | `_ n y a` | 2,812 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ n y a n` | 2,652 |
| 2 | `n y a n _` | 2,610 |
| 3 | `_ w u n _` | 1,957 |
| 4 | `_ n a n _` | 1,855 |
| 5 | `_ k i n _` | 980 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 227
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~38% 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 | 0.7131 | 1.639 | 3.99 | 12,109 | 28.7% |
| **1** | Subword | 1.1738 | 2.256 | 7.94 | 375 | 0.0% |
| **2** | Word | 0.2337 | 1.176 | 1.48 | 47,930 | 76.6% |
| **2** | Subword | 1.0147 | 2.021 | 5.23 | 2,976 | 0.0% |
| **3** | Word | 0.0783 | 1.056 | 1.12 | 70,052 | 92.2% |
| **3** | Subword | 0.7842 | 1.722 | 3.28 | 15,575 | 21.6% |
| **4** | Word | 0.0281 🏆 | 1.020 | 1.04 | 77,857 | 97.2% |
| **4** | Subword | 0.5165 | 1.430 | 2.10 | 51,106 | 48.3% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `nan enan wuncin de chikan toh finishing santatun theft auto gta enan siyasa ah de nan`
2. `be playdata e ce yegboro santatun nyan payin wun yi pentagon etishi chi tun eya fiti`
3. `nyan tswanyin chi ya toh yizhele be nyana gan nan ewun dan mini yetu wun de`

**Context Size 2:**

1. `wun yi o egi enan bolu wuncin de yesan yizhe kaman wun yi o gap inc ga`
2. `o nan de egwa du ya be lila keba nyan eni r b afropop pop ah be`
3. `ah be donald wilson wun wugwa wun man yebo gan nan yi kpako ebo dindan nyan bolu`

**Context Size 3:**

1. `wun yi o chi de kukukeba be eko yilozun e66 eko oud metha be d73 eko 2nd za`
2. `ah man u august 26 edzo yesan chi stuntman ah be cowboy nan ah la dan prorodeo hall`
3. `yi o nan e che bolu ta zuma o na ya kin retrieved 9 april santatun`

**Context Size 4:**

1. `wun yi o nan de tswitswa gwata kampany motorola mobility zuk mobile ah be medio gwala lenovo ela apr...`
2. `nan ah kpeye be doka madureira koma doka nan egi kin brazil nan yi coach toh bolu chechi nyan`
3. `from the original on 29 august retrieved 3 september 2baba ga yi eza chaba nan gi riatwa mtv ema`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_dorn_(a_eand_n_`
2. `a_e_nyann_nsa_e_`
3. `n_wspr_betunatst`

**Context Size 2:**

1. `angeraticoundan_1`
2. `n_ellemi_eko_ment`
3. `a_shot_nangi_larf`

**Context Size 3:**

1. `an_de_li_gan_janu'`
2. `_nan_zhe_fool_on_n`
3. `nan._millege_u.s_k`

**Context Size 4:**

1. `_nan_tswafo_gwegi_v`
2. `_wun_marchived_18_a`
3. `yan_payin_wun_yilaz`


### Key Findings

- **Best Predictability:** Context-4 (word) with 97.2% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (51,106 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 | 4,787 |
| Total Tokens | 80,735 |
| Mean Frequency | 16.87 |
| Median Frequency | 3 |
| Frequency Std Dev | 107.35 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | nan | 3,508 |
| 2 | be | 2,579 |
| 3 | nyan | 2,500 |
| 4 | o | 2,417 |
| 5 | wun | 2,108 |
| 6 | yi | 1,722 |
| 7 | ah | 1,483 |
| 8 | de | 1,371 |
| 9 | chi | 1,047 |
| 10 | kin | 995 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | alderny | 2 |
| 2 | jersey | 2 |
| 3 | halmstad | 2 |
| 4 | basshunter | 2 |
| 5 | gunini | 2 |
| 6 | cox | 2 |
| 7 | wikitorial | 2 |
| 8 | rangaunu | 2 |
| 9 | kaiwaka | 2 |
| 10 | application | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0809 |
| R² (Goodness of Fit) | 0.989658 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 55.6% |
| Top 1,000 | 84.5% |
| Top 5,000 | 0.0% |
| Top 10,000 | 0.0% |

### Key Findings

- **Zipf Compliance:** R²=0.9897 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 55.6% of corpus
- **Long Tail:** -5,213 words needed for remaining 100.0% 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 | 0.0436 🏆 | 0.6527 | N/A | N/A |
| **mono_64d** | 64 | 0.0084 | 0.6738 | N/A | N/A |
| **mono_128d** | 128 | 0.0017 | 0.6732 | N/A | N/A |
| **aligned_32d** | 32 | 0.0436 | 0.6316 | 0.0040 | 0.0520 |
| **aligned_64d** | 64 | 0.0084 | 0.6533 | 0.0100 | 0.0480 |
| **aligned_128d** | 128 | 0.0017 | 0.6773 | 0.0040 | 0.0460 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.0436 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.6603. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 1.0% 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 | **0.719** | 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 |
|--------|----------|
| `-s` | sati, southern, stage |
| `-a` | australian, alaska, adara |
| `-b` | bodo, bididi, behind |
| `-m` | my, minority, miss |
| `-e` | ezagbakozhi, etin, egwagan |
| `-g` | gwala, gap, ganwagi |
| `-k` | kpeuye, kamina, kala |
| `-c` | continent, climate, cambridge |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-n` | australian, etin, dukun |
| `-a` | gwala, alaska, tarawa |
| `-i` | ezagbakozhi, ganwagi, dasuki |
| `-e` | kpeuye, climate, kpeye |
| `-s` | this, miss, macleans |
| `-r` | register, factor, myanmar |
| `-an` | australian, urban, egwagan |
| `-o` | ronaldinho, bodo, kano |

### 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 |
|------|----------|------------------|----------|
| `angi` | 1.30x | 15 contexts | dangi, nangi, sangi |

### 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 |
|--------|--------|-----------|----------|
| `-e` | `-i` | 29 words | ezagbakozhi, emi |
| `-e` | `-n` | 29 words | etin, egwagan |
| `-a` | `-a` | 22 words | alaska, adara |
| `-c` | `-n` | 21 words | canadian, children |
| `-a` | `-s` | 21 words | assets, athletes |
| `-k` | `-a` | 20 words | kamina, kala |
| `-m` | `-i` | 19 words | mardini, makarini |
| `-c` | `-s` | 19 words | chillies, christmas |
| `-s` | `-s` | 19 words | ships, s |
| `-m` | `-a` | 18 words | mehsana, mokwa |

### 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 |
|------|-----------------|------------|------|
| kabalagala | **`kabalag-al-a`** | 7.5 | `al` |
| gbagbangi | **`g-ba-gbangi`** | 7.5 | `gbangi` |
| augustine | **`august-in-e`** | 7.5 | `in` |
| chinwanchi | **`ch-in-wanchi`** | 7.5 | `wanchi` |
| musulunci | **`musulu-n-ci`** | 7.5 | `n` |
| universiade | **`universia-d-e`** | 7.5 | `d` |
| kamindondo | **`ka-mi-ndondo`** | 6.0 | `ndondo` |
| enyanichi | **`enyan-ic-hi`** | 6.0 | `enyan` |
| brazilian | **`brazil-i-an`** | 6.0 | `brazil` |
| ezhiminsun | **`ezhimi-ns-un`** | 6.0 | `ezhimi` |
| journalist | **`journal-i-st`** | 6.0 | `journal` |
| engineering | **`engineer-i-ng`** | 6.0 | `engineer` |
| nationale | **`national-e`** | 4.5 | `national` |
| amalouchio | **`a-ma-louchio`** | 4.5 | `louchio` |
| commissioner | **`commission-er`** | 4.5 | `commission` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Nupe-Nupe-Tako 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 | **32k BPE** | Best compression (4.18x) |
| N-gram | **2-gram** | Lowest perplexity (227) |
| Markov | **Context-4** | Highest predictability (97.2%) |
| 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 16:17:39*