README.md

---
language: bew
language_name: Betawi
language_family: austronesian_malay
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-austronesian_malay
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: 4.630
  - name: best_isotropy
    type: isotropy
    value: 0.7504
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-03
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Betawi** 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** | 3.806x | 3.81 | 0.1398% | 155,259 |
| **16k** | 4.118x | 4.13 | 0.1512% | 143,483 |
| **32k** | 4.386x | 4.39 | 0.1611% | 134,715 |
| **64k** | 4.630x 🏆 | 4.64 | 0.1700% | 127,635 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `D atawa hurup kecitnya d ya'entu hurup ke'ampat dalem hurup Latèn. Ruju'an Latèn`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁d ▁atawa ▁hurup ▁kecitnya ▁d ▁ya ' entu ▁hurup ▁ke ... (+11 more)` | 21 |
| 16k | `▁d ▁atawa ▁hurup ▁kecitnya ▁d ▁ya ' entu ▁hurup ▁ke ... (+11 more)` | 21 |
| 32k | `▁d ▁atawa ▁hurup ▁kecitnya ▁d ▁ya ' entu ▁hurup ▁ke ... (+10 more)` | 20 |
| 64k | `▁d ▁atawa ▁hurup ▁kecitnya ▁d ▁ya ' entu ▁hurup ▁ke ... (+10 more)` | 20 |

**Sample 2:** `Karawaci entu kecamatan nyang ada di Tanggerang Kota. Ni kecamatan ngejenggar am...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁kara wa ci ▁entu ▁kecamatan ▁nyang ▁ada ▁di ▁tanggerang ▁kota ... (+17 more)` | 27 |
| 16k | `▁kara wa ci ▁entu ▁kecamatan ▁nyang ▁ada ▁di ▁tanggerang ▁kota ... (+17 more)` | 27 |
| 32k | `▁kara wa ci ▁entu ▁kecamatan ▁nyang ▁ada ▁di ▁tanggerang ▁kota ... (+17 more)` | 27 |
| 64k | `▁karawaci ▁entu ▁kecamatan ▁nyang ▁ada ▁di ▁tanggerang ▁kota . ▁ni ... (+15 more)` | 25 |

**Sample 3:** `Limo entu kecamatan nyang ada di Dèpok Kota, Jawa Kulon, Indonésia. Ni kecamatan...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁li mo ▁entu ▁kecamatan ▁nyang ▁ada ▁di ▁dèpok ▁kota , ... (+21 more)` | 31 |
| 16k | `▁limo ▁entu ▁kecamatan ▁nyang ▁ada ▁di ▁dèpok ▁kota , ▁jawa ... (+20 more)` | 30 |
| 32k | `▁limo ▁entu ▁kecamatan ▁nyang ▁ada ▁di ▁dèpok ▁kota , ▁jawa ... (+20 more)` | 30 |
| 64k | `▁limo ▁entu ▁kecamatan ▁nyang ▁ada ▁di ▁dèpok ▁kota , ▁jawa ... (+20 more)` | 30 |


### Key Findings

- **Best Compression:** 64k achieves 4.630x compression
- **Lowest UNK Rate:** 8k with 0.1398% 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 | 2,340 | 11.19 | 7,241 | 31.0% | 59.9% |
| **2-gram** | Subword | 256 🏆 | 8.00 | 2,508 | 70.0% | 98.9% |
| **3-gram** | Word | 1,985 | 10.95 | 6,755 | 33.8% | 62.9% |
| **3-gram** | Subword | 1,910 | 10.90 | 16,523 | 30.0% | 74.7% |
| **4-gram** | Word | 3,084 | 11.59 | 9,753 | 29.7% | 56.5% |
| **4-gram** | Subword | 8,721 | 13.09 | 66,990 | 16.6% | 46.5% |
| **5-gram** | Word | 1,919 | 10.91 | 5,996 | 33.6% | 65.3% |
| **5-gram** | Subword | 22,647 | 14.47 | 131,412 | 12.5% | 33.4% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `arab gundul` | 3,312 |
| 2 | `hurup arab` | 3,190 |
| 3 | `ruju an` | 2,891 |
| 4 | `ada di` | 1,396 |
| 5 | `entu atu` | 1,364 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `hurup arab gundul` | 3,176 |
| 2 | `nyang ada di` | 741 |
| 3 | `ruju an di` | 723 |
| 4 | `nyang tinggal di` | 641 |
| 5 | `tinggal di mari` | 614 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `nyang tinggal di mari` | 609 |
| 2 | `orang nyang tinggal di` | 600 |
| 3 | `ruju an di indonésia` | 529 |
| 4 | `nyang ada di propinsi` | 509 |
| 5 | `km2 dengen kepadetan penduduknya` | 501 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `orang nyang tinggal di mari` | 584 |
| 2 | `nyang tinggal di mari ruju` | 442 |
| 3 | `tinggal di mari ruju an` | 442 |
| 4 | `di mari ruju an di` | 440 |
| 5 | `mari ruju an di indonésia` | 438 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a n` | 74,827 |
| 2 | `a _` | 60,507 |
| 3 | `n g` | 54,383 |
| 4 | `n _` | 46,937 |
| 5 | `_ a` | 35,570 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n y a` | 27,185 |
| 2 | `a n g` | 25,765 |
| 3 | `n g _` | 25,518 |
| 4 | `a n _` | 24,856 |
| 5 | `_ d i` | 20,857 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a n g _` | 17,737 |
| 2 | `n y a _` | 13,480 |
| 3 | `_ d i _` | 10,268 |
| 4 | `_ n y a` | 10,013 |
| 5 | `y a n g` | 9,660 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `y a n g _` | 9,531 |
| 2 | `_ n y a n` | 9,175 |
| 3 | `n y a n g` | 9,145 |
| 4 | `_ a m a _` | 5,520 |
| 5 | `e n t u _` | 5,202 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 256
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~33% of corpus
- **Recommendation:** 4-gram or 5-gram for best predictive performance

---
## 3. Markov Chain Evaluation

![Markov Entropy](visualizations/markov_entropy.png)

![Markov Contexts](visualizations/markov_contexts.png)

![Markov Branching](visualizations/markov_branching.png)

### Results

| Context | Variant | Avg Entropy | Perplexity | Branching Factor | Unique Contexts | Predictability |
|---------|---------|-------------|------------|------------------|-----------------|----------------|
| **1** | Word | 0.8296 | 1.777 | 4.87 | 41,205 | 17.0% |
| **1** | Subword | 0.7866 | 1.725 | 4.95 | 1,639 | 21.3% |
| **2** | Word | 0.2134 | 1.159 | 1.43 | 200,219 | 78.7% |
| **2** | Subword | 0.7991 | 1.740 | 4.40 | 8,105 | 20.1% |
| **3** | Word | 0.0565 | 1.040 | 1.10 | 285,266 | 94.3% |
| **3** | Subword | 0.7622 | 1.696 | 3.43 | 35,638 | 23.8% |
| **4** | Word | 0.0212 🏆 | 1.015 | 1.04 | 311,018 | 97.9% |
| **4** | Subword | 0.5570 | 1.471 | 2.34 | 122,163 | 44.3% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `di mari per sènsus tahon wayah ada singa laut malah bulu tepok tepok bulu dènemarken juga`
2. `nyang gocan berobah beneran gim kumpiuter hal ada 412 ama jadi dedengkot soldadu romèn hurup arap`
3. `ama kemajuan èkonomi kecil bakal dipisahin deri prasman tchad arab gundul ايسيت ièlah orang nyang ad...`

**Context Size 2:**

1. `arab gundul سورين entu tana rumput rata ada banyak bodoran tasawup nyang kenisbat ke dia punya anggu...`
2. `hurup arab gundul دمفا indonésia herpes nyang pires dampa ringkes hsv ièlah atu bangunan dasaran nya...`
3. `ruju an enclekan wikimédia jakarta`

**Context Size 3:**

1. `hurup arab gundul عصر atawa sembayang asar hurup arab gundul فراولين di kaèdah basa entu penglakon d...`
2. `nyang ada di propinsi jawa tenga ni kabupatèn punya sintrem guwernemèn ada di jailolo ni kabupatèn n...`
3. `ruju an di indonésia tenga kota`

**Context Size 4:**

1. `nyang tinggal di mari di indonésia tenga`
2. `orang nyang tinggal di mari ruju an di indonésia kulon kota`
3. `nyang ada di propinsi jawa tenga ni kabupatèn punya sintrem guwernemèn ada di pati ni kabupatèn ngej...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_t,_naèsa_(in_an`
2. `ah_ha_n_psc_sèn,`
3. `nalianyanngele-d`

**Context Size 2:**

1. `andanésin_nya_at.`
2. `a_dongan_1_jen._d`
3. `ng_bensia_or._ret`

**Context Size 3:**

1. `nya_ke_1:_6_ada_de`
2. `ang))_atu_kulon_de`
3. `ng_nya_punya,_kota`

**Context Size 4:**

1. `ang_damé_kalannya_b`
2. `nya_design:top;padd`
3. `_di_kota_lingking_k`


### Key Findings

- **Best Predictability:** Context-4 (word) with 97.9% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (122,163 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 | 18,200 |
| Total Tokens | 340,971 |
| Mean Frequency | 18.73 |
| Median Frequency | 4 |
| Frequency Std Dev | 164.32 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | di | 10,322 |
| 2 | nyang | 9,100 |
| 3 | ama | 5,533 |
| 4 | entu | 5,337 |
| 5 | ada | 4,148 |
| 6 | atawa | 3,973 |
| 7 | ni | 3,950 |
| 8 | punya | 3,836 |
| 9 | hurup | 3,638 |
| 10 | arab | 3,568 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | kirinya | 2 |
| 2 | ngeloncat | 2 |
| 3 | abi | 2 |
| 4 | gelanggang | 2 |
| 5 | writing | 2 |
| 6 | syaamil | 2 |
| 7 | fermentasi | 2 |
| 8 | oase | 2 |
| 9 | maimon | 2 |
| 10 | herawati | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0754 |
| R² (Goodness of Fit) | 0.994702 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 41.8% |
| Top 1,000 | 69.7% |
| Top 5,000 | 87.8% |
| Top 10,000 | 94.6% |

### Key Findings

- **Zipf Compliance:** R²=0.9947 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 41.8% of corpus
- **Long Tail:** 8,200 words needed for remaining 5.4% 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 | 0.7504 | 0.3662 | N/A | N/A |
| **mono_64d** | 64 | 0.4073 | 0.3304 | N/A | N/A |
| **mono_128d** | 128 | 0.0951 | 0.3259 | N/A | N/A |
| **aligned_32d** | 32 | 0.7504 🏆 | 0.3611 | 0.0280 | 0.1800 |
| **aligned_64d** | 64 | 0.4073 | 0.3298 | 0.0640 | 0.2540 |
| **aligned_128d** | 128 | 0.0951 | 0.3286 | 0.0840 | 0.2940 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.7504 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.3404. 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 | **0.957** | 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 |
|--------|----------|
| `-pe` | perinta, pernahkan, pengablagan |
| `-di` | dirangkèng, diplomat, dibelakonin |
| `-ke` | kepri, kerbala, kesannya |
| `-ng` | ngucap, ngelangsir, nglingkup |
| `-se` | secret, sejarah, sexual |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-n` | pernahkan, pengablagan, waringin |
| `-an` | pernahkan, pengablagan, tuan |
| `-a` | perinta, kakinya, udara |
| `-ya` | kakinya, bawaannya, kesannya |
| `-nya` | kakinya, bawaannya, kesannya |
| `-ng` | dirangkèng, peringgiorang, bambang |
| `-in` | waringin, lanjutin, ngusahain |

### 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 |
|------|----------|------------------|----------|
| `anya` | 1.55x | 72 contexts | tanya, nanya, anyar |
| `ngan` | 1.63x | 52 contexts | ongan, ringan, dengan |
| `angg` | 1.48x | 64 contexts | kanggo, bangga, mangga |
| `aran` | 1.38x | 71 contexts | maran, saran, garan |
| `enga` | 1.61x | 36 contexts | senga, nenga, tenga |
| `anny` | 1.68x | 27 contexts | annya, umannya, ujannya |
| `unya` | 1.65x | 27 contexts | punya, baunya, atunya |
| `rang` | 1.32x | 60 contexts | orang, prang, urang |
| `inya` | 1.49x | 36 contexts | sinyal, minyak, arinya |
| `atan` | 1.50x | 32 contexts | yatan, alatan, muatan |
| `ling` | 1.41x | 40 contexts | aling, èling, beling |
| `enge` | 1.48x | 25 contexts | pengen, tengen, denger |

### 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 |
|--------|--------|-----------|----------|
| `-pe` | `-n` | 250 words | pengrongrongan, penyatetan |
| `-pe` | `-an` | 238 words | pengrongrongan, penyatetan |
| `-di` | `-n` | 182 words | disebabin, dianyarin |
| `-ke` | `-n` | 180 words | kedoktoran, keaturan |
| `-di` | `-in` | 172 words | disebabin, dianyarin |
| `-ke` | `-an` | 167 words | kedoktoran, keaturan |
| `-ng` | `-n` | 145 words | ngirimin, ngatasin |
| `-ng` | `-in` | 140 words | ngirimin, ngatasin |
| `-se` | `-a` | 50 words | serba, seninya |
| `-pe` | `-a` | 47 words | pegihnja, perdananya |

### 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 |
|------|-----------------|------------|------|
| pengucapannya | **`pe-ng-ucap-an-nya`** | 9.0 | `ucap` |
| kesaktiannya | **`ke-sakti-an-nya`** | 7.5 | `sakti` |
| pengujungan | **`pe-ng-ujung-an`** | 7.5 | `ujung` |
| dibilangin | **`di-bila-ng-in`** | 7.5 | `bila` |
| pengrobahan | **`pe-ng-robah-an`** | 7.5 | `robah` |
| kedaulatannya | **`ke-daulat-an-nya`** | 7.5 | `daulat` |
| penggapaan | **`pe-ng-gapa-an`** | 7.5 | `gapa` |
| diterjemahinnya | **`di-terjemah-in-nya`** | 7.5 | `terjemah` |
| penggawéan | **`pe-ng-gawé-an`** | 7.5 | `gawé` |
| sampingannya | **`sampi-ng-an-nya`** | 7.5 | `sampi` |
| kebanyakannya | **`ke-banyak-an-nya`** | 7.5 | `banyak` |
| dilindungin | **`di-lindu-ng-in`** | 7.5 | `lindu` |
| dikeringin | **`di-ke-ring-in`** | 7.5 | `ring` |
| kebalikannya | **`ke-balik-an-nya`** | 7.5 | `balik` |
| dimaèninnya | **`di-maèn-in-nya`** | 7.5 | `maèn` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Betawi 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 (4.63x) |
| N-gram | **2-gram** | Lowest perplexity (256) |
| Markov | **Context-4** | Highest predictability (97.9%) |
| 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: 2026-01-03 18:42:18*