README.md

---
language: tk
language_name: Turkmen
language_family: turkic_oghuz
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-turkic_oghuz
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: 4.949
  - name: best_isotropy
    type: isotropy
    value: 0.8902
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-11
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Turkmen** 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** | 3.867x | 3.87 | 0.1563% | 394,866 |
| **16k** | 4.295x | 4.30 | 0.1736% | 355,501 |
| **32k** | 4.665x | 4.67 | 0.1885% | 327,292 |
| **64k** | 4.949x 🏆 | 4.95 | 0.2000% | 308,505 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Wakalar Sebitler boýunça Tema boýunça <noinclude> Dünýä inenler Aradan çykanlar`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁wakalar ▁sebitler ▁boýunça ▁tema ▁boýunça ▁< noinclude > ▁dünýä ▁inenler ... (+2 more)` | 12 |
| 16k | `▁wakalar ▁sebitler ▁boýunça ▁tema ▁boýunça ▁< noinclude > ▁dünýä ▁inenler ... (+2 more)` | 12 |
| 32k | `▁wakalar ▁sebitler ▁boýunça ▁tema ▁boýunça ▁< noinclude > ▁dünýä ▁inenler ... (+2 more)` | 12 |
| 64k | `▁wakalar ▁sebitler ▁boýunça ▁tema ▁boýunça ▁< noinclude > ▁dünýä ▁inenler ... (+2 more)` | 12 |

**Sample 2:** `Wakalar Sebitler boýunça Tema boýunça <noinclude> Dünýä inenler Aradan çykanlar`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁wakalar ▁sebitler ▁boýunça ▁tema ▁boýunça ▁< noinclude > ▁dünýä ▁inenler ... (+2 more)` | 12 |
| 16k | `▁wakalar ▁sebitler ▁boýunça ▁tema ▁boýunça ▁< noinclude > ▁dünýä ▁inenler ... (+2 more)` | 12 |
| 32k | `▁wakalar ▁sebitler ▁boýunça ▁tema ▁boýunça ▁< noinclude > ▁dünýä ▁inenler ... (+2 more)` | 12 |
| 64k | `▁wakalar ▁sebitler ▁boýunça ▁tema ▁boýunça ▁< noinclude > ▁dünýä ▁inenler ... (+2 more)` | 12 |

**Sample 3:** `Seýdi etraby — Lebap welayatynyň bir etrabydyr. etraplary welaýaty welaýatyndaky...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁seý di ▁etraby ▁— ▁lebap ▁welayat ynyň ▁bir ▁etraby dyr ... (+5 more)` | 15 |
| 16k | `▁seýdi ▁etraby ▁— ▁lebap ▁welayat ynyň ▁bir ▁etraby dyr . ... (+4 more)` | 14 |
| 32k | `▁seýdi ▁etraby ▁— ▁lebap ▁welayatynyň ▁bir ▁etrabydyr . ▁etraplary ▁welaýaty ... (+2 more)` | 12 |
| 64k | `▁seýdi ▁etraby ▁— ▁lebap ▁welayatynyň ▁bir ▁etrabydyr . ▁etraplary ▁welaýaty ... (+2 more)` | 12 |


### Key Findings

- **Best Compression:** 64k achieves 4.949x compression
- **Lowest UNK Rate:** 8k with 0.1563% 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 | 11,088 | 13.44 | 23,947 | 14.6% | 32.8% |
| **2-gram** | Subword | 355 🏆 | 8.47 | 4,493 | 61.5% | 98.3% |
| **3-gram** | Word | 7,047 | 12.78 | 19,707 | 21.5% | 35.2% |
| **3-gram** | Subword | 2,934 | 11.52 | 34,530 | 22.8% | 66.5% |
| **4-gram** | Word | 20,732 | 14.34 | 46,279 | 14.6% | 21.3% |
| **4-gram** | Subword | 14,717 | 13.85 | 159,071 | 11.4% | 36.9% |
| **5-gram** | Word | 15,656 | 13.93 | 36,681 | 16.0% | 22.7% |
| **5-gram** | Subword | 46,546 | 15.51 | 363,230 | 6.8% | 23.5% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ýa da` | 2,786 |
| 2 | `aradan çykanlar` | 2,220 |
| 3 | `tema boýunça` | 2,220 |
| 4 | `dünýä inenler` | 2,217 |
| 5 | `sebitler boýunça` | 2,216 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `wakalar sebitler boýunça` | 2,208 |
| 2 | `boýunça tema boýunça` | 2,201 |
| 3 | `sebitler boýunça tema` | 2,201 |
| 4 | `dünýä inenler aradan` | 2,174 |
| 5 | `inenler aradan çykanlar` | 2,174 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `sebitler boýunça tema boýunça` | 2,201 |
| 2 | `wakalar sebitler boýunça tema` | 2,196 |
| 3 | `dünýä inenler aradan çykanlar` | 2,174 |
| 4 | `tema boýunça noinclude dünýä` | 2,119 |
| 5 | `boýunça noinclude dünýä inenler` | 2,119 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `wakalar sebitler boýunça tema boýunça` | 2,196 |
| 2 | `tema boýunça noinclude dünýä inenler` | 2,119 |
| 3 | `sebitler boýunça tema boýunça noinclude` | 2,112 |
| 4 | `boýunça tema boýunça noinclude dünýä` | 2,112 |
| 5 | `noinclude dünýä inenler aradan çykanlar` | 2,085 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a r` | 188,493 |
| 2 | `l a` | 152,165 |
| 3 | `a n` | 151,310 |
| 4 | `_ b` | 146,537 |
| 5 | `a _` | 138,776 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `l a r` | 82,667 |
| 2 | `a r y` | 58,594 |
| 3 | `y ň _` | 57,971 |
| 4 | `a n _` | 55,883 |
| 5 | `r . _` | 53,874 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `l a r y` | 41,638 |
| 2 | `n y ň _` | 30,386 |
| 3 | `_ w e _` | 29,297 |
| 4 | `y n d a` | 26,755 |
| 5 | `l e r i` | 26,718 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ b i l e` | 16,563 |
| 2 | `i l e n _` | 16,493 |
| 3 | `y n d a _` | 16,259 |
| 4 | `y n y ň _` | 15,844 |
| 5 | `b i l e n` | 14,698 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 355
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~23% 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 | 0.8425 | 1.793 | 5.31 | 167,857 | 15.8% |
| **1** | Subword | 1.0332 | 2.047 | 8.72 | 1,227 | 0.0% |
| **2** | Word | 0.1779 | 1.131 | 1.35 | 888,328 | 82.2% |
| **2** | Subword | 1.0291 | 2.041 | 6.32 | 10,675 | 0.0% |
| **3** | Word | 0.0393 | 1.028 | 1.06 | 1,193,586 | 96.1% |
| **3** | Subword | 0.8531 | 1.806 | 4.15 | 67,431 | 14.7% |
| **4** | Word | 0.0110 🏆 | 1.008 | 1.01 | 1,255,469 | 98.9% |
| **4** | Subword | 0.6220 | 1.539 | 2.69 | 279,783 | 37.8% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `we hemişe eline düşüpdir aýallaryñ häkimlik edýär bangkokdaky ýurduň 12 15 eretriýadan hem de ýylyň ...`
2. `bilen icc bütindünýä güni kyýamat gününi alada üns berilýär asteroidler ýaly düzüp ol birwagtlar zaý...`
3. `hem satuwa çykaryldy awstro wengriýa bilen kagyz ýüzündeligine galdy ž gulart käbir bölekleriniň geç...`

**Context Size 2:**

1. `ýa da mikaýyl bin seljuk bin dükak ýylda mälik şa üçin jelaly kalendaryny hijri kalendaryny mysal hö...`
2. `tema boýunça noinclude dünýä inenler aradan çykanlar kategoriýa`
3. `dünýä inenler aradan çykanlar salgylanmalar`

**Context Size 3:**

1. `wakalar sebitler boýunça tema boýunça noinclude dünýä inenler aradan çykanlar 31`
2. `boýunça tema boýunça noinclude dünýä inenler aradan çykanlar 104`
3. `sebitler boýunça tema boýunça noinclude dünýä inenler aradan çykanlar 29`

**Context Size 4:**

1. `sebitler boýunça tema boýunça noinclude dünýä inenler aradan çykanlar 26`
2. `wakalar sebitler boýunça tema boýunça noinclude dünýä inenler aradan çykanlar baýramçylyklar`
3. `boýunça noinclude dünýä inenler aradan çykanlar towşan esenowa hydyr derýaýew kerim gurbannepesow`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_botaýärkmp),_öz`
2. `a_gumgitdar_nyle`
3. `ebury_der._ýasah`

**Context Size 2:**

1. `ar._oduşli_düşdir`
2. `lar.ilbaşdyry,_ob`
3. `an_emlündama_(ýar`

**Context Size 3:**

1. `laryň_daşly_şübhes`
2. `ary_12-150-nji_mil`
3. `yň_aýatynyň_keşler`

**Context Size 4:**

1. `lary_deňde_gölli,_o`
2. `nyň_bolandygynda_ru`
3. `_we_goşuny,_hassa_t`


### Key Findings

- **Best Predictability:** Context-4 (word) with 98.9% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (279,783 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 | 70,850 |
| Total Tokens | 1,266,247 |
| Mean Frequency | 17.87 |
| Median Frequency | 4 |
| Frequency Std Dev | 172.27 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | we | 29,419 |
| 2 | bilen | 14,593 |
| 3 | hem | 9,723 |
| 4 | bu | 9,296 |
| 5 | bir | 7,148 |
| 6 | üçin | 7,116 |
| 7 | da | 6,676 |
| 8 | boýunça | 6,346 |
| 9 | ol | 6,099 |
| 10 | ýylda | 5,569 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | halaçda | 2 |
| 2 | byradarlygynyň | 2 |
| 3 | halaja | 2 |
| 4 | bakynyň | 2 |
| 5 | esaslandyrylanlar | 2 |
| 6 | ailəsi | 2 |
| 7 | yörükler | 2 |
| 8 | ýarymgoragçysy | 2 |
| 9 | jizak | 2 |
| 10 | kolhozçi | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 0.9487 |
| R² (Goodness of Fit) | 0.992202 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 22.4% |
| Top 1,000 | 47.7% |
| Top 5,000 | 70.0% |
| Top 10,000 | 79.1% |

### Key Findings

- **Zipf Compliance:** R²=0.9922 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 22.4% of corpus
- **Long Tail:** 60,850 words needed for remaining 20.9% 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 | 0.8902 | 0.2916 | N/A | N/A |
| **mono_64d** | 64 | 0.8799 | 0.2188 | N/A | N/A |
| **mono_128d** | 128 | 0.6945 | 0.1696 | N/A | N/A |
| **aligned_32d** | 32 | 0.8902 🏆 | 0.2952 | 0.0120 | 0.1680 |
| **aligned_64d** | 64 | 0.8799 | 0.2224 | 0.0560 | 0.2240 |
| **aligned_128d** | 128 | 0.6945 | 0.1700 | 0.0840 | 0.3140 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.8902 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2279. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 8.4% 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 | **-0.035** | Low formulaic 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` | awyny, andrýu, alta |
| `-s` | saklanýandyr, saýylmadyk, stories |
| `-g` | gyşy, guzlar, gallipoli |
| `-b` | beloklaryny, basílio, basylan |
| `-m` | meýi, maersk, mortier |
| `-k` | kekene, klisfeniň, kesil |
| `-d` | diskriminasiýa, deňlemek, dakylýar |
| `-t` | theodore, territoriýasyndaky, taýynlapdyr |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-ň` | operasiýalaryň, klisfeniň, aşgabadyň |
| `-r` | saklanýandyr, guzlar, mortier |
| `-y` | beloklaryny, gyşy, awyny |
| `-a` | diskriminasiýa, alta, gatyşmagynda |
| `-yň` | operasiýalaryň, aşgabadyň, wahýyň |
| `-n` | humaýun, basylan, araçäkleşýän |
| `-i` | meýi, redmi, erişleri |
| `-an` | basylan, gan, barylýan |

### 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 |
|------|----------|------------------|----------|
| `kmen` | 3.11x | 26 contexts | rkmen, sökmen, çekmen |
| `anla` | 1.82x | 155 contexts | sanlar, panlar, hanlar |
| `asyn` | 1.76x | 181 contexts | ýasyn, masyn, gasyn |
| `erin` | 1.91x | 103 contexts | lerin, erine, yerin |
| `rkme` | 3.11x | 14 contexts | rkmen, türkmer, turkmen |
| `tlar` | 1.70x | 133 contexts | atlar, otlar, otlara |
| `rler` | 1.83x | 86 contexts | ärler, ÿrler, ýerler |
| `nlar` | 1.84x | 79 contexts | onlar, gunlar, hunlar |
| `erle` | 1.63x | 96 contexts | ýerler, ỳerler, gerlen |
| `ylar` | 1.63x | 72 contexts | lylar, kylar, sylar |
| `rlar` | 1.60x | 76 contexts | arlar, durlar, ýarlar |
| `klar` | 1.67x | 63 contexts | uklar, klark, oklar |

### 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 |
|--------|--------|-----------|----------|
| `-g` | `-y` | 123 words | gaşy, gatnawy |
| `-g` | `-r` | 121 words | gaçypdyrlar, girilýär |
| `-g` | `-a` | 96 words | gidrogeologiýa, graflyklara |
| `-b` | `-r` | 92 words | bir, bazaar |
| `-g` | `-n` | 89 words | gaýtarylan, gelmeýän |
| `-g` | `-i` | 88 words | geçmegi, güýçli |
| `-s` | `-ň` | 87 words | sahypalaryň, süýümleriniň |
| `-s` | `-y` | 80 words | sostawyny, satmagy |
| `-g` | `-ň` | 76 words | goýumdarlarynyň, guramaklygyň |
| `-b` | `-y` | 75 words | bozulmagy, bidgatçy |

### 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 |
|------|-----------------|------------|------|
| slawýanlarda | **`slawýanl-ar-da`** | 7.5 | `ar` |
| görkezipdir | **`görkezip-di-r`** | 7.5 | `di` |
| oktýabrdan | **`oktýabr-da-n`** | 7.5 | `da` |
| balyklaryñ | **`balykl-ar-yñ`** | 7.5 | `ar` |
| sazandalary | **`sazandal-ar-y`** | 7.5 | `ar` |
| bolanlary | **`bolanl-ar-y`** | 7.5 | `ar` |
| garşydaşlary | **`garşydaşl-ar-y`** | 7.5 | `ar` |
| halykynyň | **`halyky-n-yň`** | 7.5 | `n` |
| manjurlaryň | **`manjurl-ar-yň`** | 7.5 | `ar` |
| mukdarlary | **`mukdarl-ar-y`** | 7.5 | `ar` |
| ybadatlarda | **`ybadatl-ar-da`** | 7.5 | `ar` |
| guşaklyklary | **`guşaklykl-ar-y`** | 7.5 | `ar` |
| amallaryň | **`amall-ar-yň`** | 7.5 | `ar` |
| ugurlarda | **`ugurl-ar-da`** | 7.5 | `ar` |
| ýakynlarda | **`ýakynl-ar-da`** | 7.5 | `ar` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Turkmen shows high morphological productivity. The subword models are significantly more efficient than word models, suggesting a rich system of affixation or compounding.

---
## 7. Summary & Recommendations

![Performance Dashboard](visualizations/performance_dashboard.png)

### Production Recommendations

| Component | Recommended | Rationale |
|-----------|-------------|-----------|
| Tokenizer | **64k BPE** | Best compression (4.95x) |
| N-gram | **2-gram** | Lowest perplexity (355) |
| Markov | **Context-4** | Highest predictability (98.9%) |
| 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: 2026-01-11 01:05:04*