README.md

---
language: dty
language_name: Dotyali
language_family: indoaryan_central
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-indoaryan_central
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: 4.539
  - name: best_isotropy
    type: isotropy
    value: 0.9032
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-04
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Dotyali** 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** | 3.506x | 3.51 | 0.1249% | 181,747 |
| **16k** | 3.906x | 3.91 | 0.1391% | 163,156 |
| **32k** | 4.207x | 4.21 | 0.1499% | 151,469 |
| **64k** | 4.539x 🏆 | 4.55 | 0.1617% | 140,390 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `सुखविंदर सिंह भारतीय सांगीतिक क्षेत्रका पाश्व गायक हुन। सन्दर्भ गिदाराअन`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁सुख वि ंदर ▁सिंह ▁भारतीय ▁सांगीतिक ▁क्षेत्रका ▁पाश्व ▁गायक ▁हुन ... (+3 more)` | 13 |
| 16k | `▁सुख वि ंदर ▁सिंह ▁भारतीय ▁सांगीतिक ▁क्षेत्रका ▁पाश्व ▁गायक ▁हुन ... (+3 more)` | 13 |
| 32k | `▁सुख विंदर ▁सिंह ▁भारतीय ▁सांगीतिक ▁क्षेत्रका ▁पाश्व ▁गायक ▁हुन । ... (+2 more)` | 12 |
| 64k | `▁सुखविंदर ▁सिंह ▁भारतीय ▁सांगीतिक ▁क्षेत्रका ▁पाश्व ▁गायक ▁हुन । ▁सन्दर्भ ... (+1 more)` | 11 |

**Sample 2:** `सिंगौडी दैलेख जिल्लामी पडडे एक गाऊ विकास समिति हो । यी पनि हेर जिल्ला विकास समित...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁सि ंग ौ डी ▁दैलेख ▁जिल्लामी ▁पडडे ▁एक ▁गाऊ ▁विकास ... (+9 more)` | 19 |
| 16k | `▁सिंग ौ डी ▁दैलेख ▁जिल्लामी ▁पडडे ▁एक ▁गाऊ ▁विकास ▁समिति ... (+8 more)` | 18 |
| 32k | `▁सिंग ौडी ▁दैलेख ▁जिल्लामी ▁पडडे ▁एक ▁गाऊ ▁विकास ▁समिति ▁हो ... (+7 more)` | 17 |
| 64k | `▁सिंगौडी ▁दैलेख ▁जिल्लामी ▁पडडे ▁एक ▁गाऊ ▁विकास ▁समिति ▁हो ▁। ... (+6 more)` | 16 |

**Sample 3:** `बेनिन अफ्रिका महाद्वीपमाई रयाको एक देश हो। सन्दर्भ देशअन`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁बेन िन ▁अफ्रिका ▁महाद्वीपमाई ▁रयाको ▁एक ▁देश ▁हो । ▁सन्दर्भ ... (+1 more)` | 11 |
| 16k | `▁बेनिन ▁अफ्रिका ▁महाद्वीपमाई ▁रयाको ▁एक ▁देश ▁हो । ▁सन्दर्भ ▁देशअन` | 10 |
| 32k | `▁बेनिन ▁अफ्रिका ▁महाद्वीपमाई ▁रयाको ▁एक ▁देश ▁हो । ▁सन्दर्भ ▁देशअन` | 10 |
| 64k | `▁बेनिन ▁अफ्रिका ▁महाद्वीपमाई ▁रयाको ▁एक ▁देश ▁हो । ▁सन्दर्भ ▁देशअन` | 10 |


### Key Findings

- **Best Compression:** 64k achieves 4.539x compression
- **Lowest UNK Rate:** 8k with 0.1249% 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 | 5,114 | 12.32 | 8,849 | 15.4% | 44.5% |
| **2-gram** | Subword | 2,395 🏆 | 11.23 | 19,229 | 33.4% | 67.5% |
| **3-gram** | Word | 5,204 | 12.35 | 8,802 | 15.6% | 43.7% |
| **3-gram** | Subword | 18,338 | 14.16 | 76,407 | 10.5% | 33.0% |
| **4-gram** | Word | 9,926 | 13.28 | 16,181 | 11.8% | 33.3% |
| **4-gram** | Subword | 63,062 | 15.94 | 207,437 | 6.1% | 20.3% |
| **5-gram** | Word | 7,716 | 12.91 | 12,232 | 12.4% | 36.5% |
| **5-gram** | Subword | 95,990 | 16.55 | 239,024 | 4.9% | 15.8% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `सन्दर्भ सामग्रीअन` | 752 |
| 2 | `गाउँ विकास` | 631 |
| 3 | `वि सं` | 572 |
| 4 | `सन् मी` | 549 |
| 5 | `हो यो` | 514 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `सन्दर्भ सामग्रीअन भाइरा` | 305 |
| 2 | `सामग्रीअन भाइरा लिङ्कअन` | 282 |
| 3 | `विकास समिति हो` | 281 |
| 4 | `यो लै हेर` | 276 |
| 5 | `गाउँ विकास समिति` | 253 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `सन्दर्भ सामग्रीअन भाइरा लिङ्कअन` | 282 |
| 2 | `गाउँ विकास समिति हो` | 232 |
| 3 | `एक गाउँ विकास समिति` | 173 |
| 4 | `रयाको एक देश हो` | 150 |
| 5 | `सन्दर्भअन यिन लै हेरऽ` | 130 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `एक गाउँ विकास समिति हो` | 173 |
| 2 | `गाउँ विकास समितीन मध्येको एक` | 123 |
| 3 | `मध्येको एक गाउँ विकास समिति` | 123 |
| 4 | `समितीन मध्येको एक गाउँ विकास` | 123 |
| 5 | `विकास समितीन मध्येको एक गाउँ` | 123 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `को _` | 29,200 |
| 2 | `। _` | 25,775 |
| 3 | `न _` | 25,224 |
| 4 | `र _` | 22,897 |
| 5 | `_ स` | 20,865 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ । _` | 7,563 |
| 2 | `_ रे _` | 7,379 |
| 3 | `अ न _` | 5,308 |
| 4 | `ला ई _` | 4,856 |
| 5 | `_ उ न` | 4,051 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ स न्द र्भ` | 2,988 |
| 2 | `_ ए क _` | 2,776 |
| 3 | `_ ने पा ल` | 2,487 |
| 4 | `_ हो । _` | 2,146 |
| 5 | `स न्द र्भ _` | 2,025 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ स न्द र्भ _` | 2,024 |
| 2 | `। _ स न्द र्भ` | 1,726 |
| 3 | `_ च ल चि त्र` | 1,346 |
| 4 | `_ हो _ । _` | 1,310 |
| 5 | `_ उ न ले _` | 1,285 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 2,395
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~16% 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 | 0.6976 | 1.622 | 4.02 | 85,572 | 30.2% |
| **1** | Subword | 0.8621 | 1.818 | 10.06 | 6,314 | 13.8% |
| **2** | Word | 0.1550 | 1.113 | 1.27 | 343,062 | 84.5% |
| **2** | Subword | 0.5671 | 1.482 | 3.71 | 63,513 | 43.3% |
| **3** | Word | 0.0392 | 1.028 | 1.05 | 434,501 | 96.1% |
| **3** | Subword | 0.4781 | 1.393 | 2.53 | 235,438 | 52.2% |
| **4** | Word | 0.0141 🏆 | 1.010 | 1.02 | 456,418 | 98.6% |
| **4** | Subword | 0.2801 | 1.214 | 1.62 | 594,541 | 72.0% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `रे सचिवमी नियुक्त गरेको थियो वि स न पा हरूसाविक वडा छन् यै जिल्लामा बसोबास गर्दा`
2. `हो यी लै प्रजनन भारत सरकारले विवादित मौजाहरूको फिराद थियो हार्ट भल्भ हरु को केन्द्र बठे`
3. `छ यद्यपि यै जिल्लामी धान नाच तमरा रूचिका विषयन्मी लेख नयाँ दिल्लीमी नानाजी देशमुखलाई गोलवलकरले उत्तर`

**Context Size 2:**

1. `सन्दर्भ सामग्रीअन बाह्य कडीअन माइस्पेस आधिकारिक पृष्ठ रङ्गशालाको वातावरण फिफा विश्वकपका रङ्गशालाअन य...`
2. `गाउँ विकास समिति हो जनगणना अन्सारअ येइ ठउर को जनसङ्ख्या १६ ५८९ रह्याको थ्यो सन्दर्भ सामग्रीअन बाइल्ल...`
3. `वि सं राणा शमशेर जङ्गबहादुर राणा सत्चित शमशेर जङ्गबहादुर राणा नर शमशेर जङ्गबहादुर राणा बमबहादुर राणा...`

**Context Size 3:**

1. `सन्दर्भ सामग्रीअन भाइरा लिङ्कअन अभिनेताअन राजनीतिज्ञ`
2. `यो लै हेर घनप्रसाद शर्मा सन्दर्भ सामग्रीअन पिडित नागरिक`
3. `सामग्रीअन भाइरा लिङ्कअन कमंस कार्ल मार्क्स कार्ल मार्क्सको हो राष्ट्रधर्म चर्चित व्यक्तित्वअन`

**Context Size 4:**

1. `सन्दर्भ सामग्रीअन भाइरा लिङ्कअन यो लै हेर चलचित्र अभिनेत्रीअन मान्सु`
2. `गाउँ विकास समिति हो विकास समितिअन`
3. `एक गाउँ विकास समिति हो यी पन हेर्या जिल्ला विकास समितिअन`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_गण_यूक्त_भाइन्_'केसम्पर्क`
2. `रपागरेकीय_पनिकक्ष_रै_अनु`
3. `न_सयनले_स्रो,_विभिन्न_d_`

**Context Size 2:**

1. `को_पैल्ली_कि_लेप्चा_सम्बन्धित_अरे`
2. `।_यै_प्रस्ताव_भारतका_दीबहिनी`
3. `न_फुटबल"_आधुनिक_नसङ्ख्या_`

**Context Size 3:**

1. `_।_उप-अवधारणालाई_आज_स`
2. `_रे_सम्बत_साफ्टवेयर_लिग_च्याम्पि`
3. `अन_पिउने_विश्वामित्रो_जमघटका_`

**Context Size 4:**

1. `_सन्दर्भहरू_माइ_विषेश_दिन_नजि`
2. `_एक_अङ्ग_भङ्ग_गर्ने_अनुमति_दि`
3. `_नेपाल_रेड्डी_निर्देशक_हिन्दी_सिनेमा`


### Key Findings

- **Best Predictability:** Context-4 (word) with 98.6% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (594,541 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 | 32,797 |
| Total Tokens | 456,553 |
| Mean Frequency | 13.92 |
| Median Frequency | 3 |
| Frequency Std Dev | 85.63 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | रे | 7,392 |
| 2 | हो | 4,556 |
| 3 | छ | 3,784 |
| 4 | मी | 3,555 |
| 5 | एक | 2,814 |
| 6 | यो | 2,747 |
| 7 | को | 2,624 |
| 8 | र | 2,560 |
| 9 | सन्दर्भ | 2,229 |
| 10 | माइ | 2,088 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | पिक्सेल | 2 |
| 2 | प्रयोगकर्ताहरूद्वारा | 2 |
| 3 | महानिरीक्षकलाई | 2 |
| 4 | महानिरीक्षकअन | 2 |
| 5 | दुबधागो | 2 |
| 6 | हार्बिनको | 2 |
| 7 | अगुदा | 2 |
| 8 | शाङ्जिङ | 2 |
| 9 | प्रिफेक्चर | 2 |
| 10 | लिआङले | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 0.9878 |
| R² (Goodness of Fit) | 0.989849 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 23.7% |
| Top 1,000 | 52.9% |
| Top 5,000 | 76.7% |
| Top 10,000 | 85.9% |

### Key Findings

- **Zipf Compliance:** R²=0.9898 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 23.7% of corpus
- **Long Tail:** 22,797 words needed for remaining 14.1% 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 | 0.9032 🏆 | 0.3305 | N/A | N/A |
| **mono_64d** | 64 | 0.7587 | 0.2622 | N/A | N/A |
| **mono_128d** | 128 | 0.3039 | 0.2479 | N/A | N/A |
| **aligned_32d** | 32 | 0.9032 | 0.3256 | 0.0040 | 0.0640 |
| **aligned_64d** | 64 | 0.7587 | 0.2643 | 0.0060 | 0.0960 |
| **aligned_128d** | 128 | 0.3039 | 0.2488 | 0.0220 | 0.1640 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.9032 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2799. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 2.2% 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 | **1.309** | 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 |
|--------|----------|
| `-प्` | प्रयोगकर्ता, प्रवेशमी, प्रेसिडेन्ट |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-ा` | क्षेत्तीमा, नैपालमा, पणया |
| `-को` | ताराको, सिजनको, गोल्डकपको |
| `-का` | भनेका, आदिका, फाराक्का |
| `-ले` | ऋषिले, अहिले, देसाईले |
| `-मी` | लडाइँमी, प्रवेशमी, आरक्षमी |
| `-ाई` | जेफरसनलाई, पर्वलाई, कार्कीलाई |
| `-या` | पणया, लगाइया, आत्महत्या |

### 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.

*No significant bound stems detected.*


### 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 |
|--------|--------|-----------|----------|
| `-प्` | `-ा` | 27 words | प्रतिरक्षा, प्यासा |
| `-प्` | `-को` | 26 words | प्रजाको, प्राणीको |
| `-प्` | `-का` | 13 words | प्रियङ्का, प्रदर्शनका |
| `-प्` | `-मी` | 10 words | प्रकृतिमी, प्रहरीमी |
| `-प्` | `-ले` | 9 words | प्रकारले, प्रविधिले |
| `-प्` | `-ाई` | 9 words | प्रधानमन्त्रीलाई, प्रचलनमाई |

### 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 |
|------|-----------------|------------|------|
| संस्थानको | **`संस्थान-को`** | 4.5 | `संस्थान` |
| संस्कारमी | **`संस्कार-मी`** | 4.5 | `संस्कार` |
| सरस्वतीले | **`सरस्वती-ले`** | 4.5 | `सरस्वती` |
| आन्दोलनको | **`आन्दोलन-को`** | 4.5 | `आन्दोलन` |
| महिनाहरूको | **`महिनाहरू-को`** | 4.5 | `महिनाहरू` |
| त्रिपाठीको | **`त्रिपाठी-को`** | 4.5 | `त्रिपाठी` |
| पञ्चायतको | **`पञ्चायत-को`** | 4.5 | `पञ्चायत` |
| सुर्मासरोवरको | **`सुर्मासरोवर-को`** | 4.5 | `सुर्मासरोवर` |
| ब्राजिलले | **`ब्राजिल-ले`** | 4.5 | `ब्राजिल` |
| हार्बिनको | **`हार्बिन-को`** | 4.5 | `हार्बिन` |
| न्यायाधीशको | **`न्यायाधीश-को`** | 4.5 | `न्यायाधीश` |
| अध्यक्षका | **`अध्यक्ष-का`** | 4.5 | `अध्यक्ष` |
| सेमिफाइनलमी | **`सेमिफाइनल-मी`** | 4.5 | `सेमिफाइनल` |
| संस्कृतिका | **`संस्कृति-का`** | 4.5 | `संस्कृति` |
| सैनिकहरूको | **`सैनिकहरू-को`** | 4.5 | `सैनिकहरू` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Dotyali 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 (4.54x) |
| N-gram | **2-gram** | Lowest perplexity (2,395) |
| Markov | **Context-4** | Highest predictability (98.6%) |
| Embeddings | **100d** | Balanced semantic capture and isotropy |


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

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

### Tokenizer Metrics

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

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

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

### N-gram Model Metrics

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

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

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

### Markov Chain Metrics

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

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

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

### Vocabulary & Zipf's Law Metrics

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

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

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

### Word Embedding Metrics

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

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

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

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

### General Interpretation Guidelines

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


### Visualizations Index

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

---
## About This Project

### Data Source

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

### Project

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

### Maintainer

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

### Citation

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

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

### License

MIT License - Free for academic and commercial use.

### Links

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

*Report Date: 2026-01-04 02:49:05*