README.md

---
language: ace
language_name: Acehnese
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: 4.925
  - name: best_isotropy
    type: isotropy
    value: 0.4644
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-03
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Acehnese** 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** | 4.118x | 4.13 | 0.2676% | 125,584 |
| **16k** | 4.487x | 4.50 | 0.2916% | 115,243 |
| **32k** | 4.726x | 4.74 | 0.3071% | 109,414 |
| **64k** | 4.925x 🏆 | 4.93 | 0.3200% | 104,998 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Jonathan Alberto "John" Leguizamo – ) nakeuh sidroe aktor asay Amirika Syarikat.`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁jonathan ▁albert o ▁" john " ▁leg ui zam o ... (+9 more)` | 19 |
| 16k | `▁jonathan ▁albert o ▁" john " ▁leg ui zam o ... (+9 more)` | 19 |
| 32k | `▁jonathan ▁alberto ▁" john " ▁leg uizamo ▁– ▁) ▁nakeuh ... (+6 more)` | 16 |
| 64k | `▁jonathan ▁alberto ▁" john " ▁leguizamo ▁– ▁) ▁nakeuh ▁sidroe ... (+5 more)` | 15 |

**Sample 2:** `Spencer Breslin nakeuh sidroe aktor asay Amirika Utara.`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁sp en cer ▁br es lin ▁nakeuh ▁sidroe ▁aktor ▁asay ... (+3 more)` | 13 |
| 16k | `▁sp en cer ▁br es lin ▁nakeuh ▁sidroe ▁aktor ▁asay ... (+3 more)` | 13 |
| 32k | `▁spencer ▁br es lin ▁nakeuh ▁sidroe ▁aktor ▁asay ▁amirika ▁utara ... (+1 more)` | 11 |
| 64k | `▁spencer ▁breslin ▁nakeuh ▁sidroe ▁aktor ▁asay ▁amirika ▁utara .` | 9 |

**Sample 3:** `Pasi Mali nakeuh saboh gampông nyang na lam keucamatan Woyla Barat, Kabupaten Ac...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁pasi ▁mali ▁nakeuh ▁saboh ▁gampông ▁nyang ▁na ▁lam ▁keucamatan ▁woyla ... (+11 more)` | 21 |
| 16k | `▁pasi ▁mali ▁nakeuh ▁saboh ▁gampông ▁nyang ▁na ▁lam ▁keucamatan ▁woyla ... (+11 more)` | 21 |
| 32k | `▁pasi ▁mali ▁nakeuh ▁saboh ▁gampông ▁nyang ▁na ▁lam ▁keucamatan ▁woyla ... (+11 more)` | 21 |
| 64k | `▁pasi ▁mali ▁nakeuh ▁saboh ▁gampông ▁nyang ▁na ▁lam ▁keucamatan ▁woyla ... (+11 more)` | 21 |


### Key Findings

- **Best Compression:** 64k achieves 4.925x compression
- **Lowest UNK Rate:** 8k with 0.2676% 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 | 640 | 9.32 | 7,037 | 62.5% | 83.3% |
| **2-gram** | Subword | 224 🏆 | 7.81 | 2,200 | 71.8% | 99.5% |
| **3-gram** | Word | 582 | 9.19 | 8,345 | 65.3% | 85.4% |
| **3-gram** | Subword | 1,199 | 10.23 | 14,644 | 37.8% | 84.8% |
| **4-gram** | Word | 678 | 9.41 | 12,913 | 64.4% | 83.6% |
| **4-gram** | Subword | 3,579 | 11.81 | 59,564 | 26.1% | 67.4% |
| **5-gram** | Word | 585 | 9.19 | 10,187 | 66.3% | 85.3% |
| **5-gram** | Subword | 6,530 | 12.67 | 114,683 | 21.4% | 60.4% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `bak laman` | 7,389 |
| 2 | `gunong nyoe` | 7,388 |
| 3 | `nyoe bak` | 5,543 |
| 4 | `nakeuh saboh` | 5,048 |
| 5 | `di acèh` | 4,747 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `gunong nyoe bak` | 5,541 |
| 2 | `nyoe bak laman` | 3,694 |
| 3 | `lumbôi gampông nyoe` | 3,567 |
| 4 | `acèh lumbôi gampông` | 3,564 |
| 5 | `nyoe lam data` | 3,499 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `gunong nyoe bak laman` | 3,694 |
| 2 | `acèh lumbôi gampông nyoe` | 3,564 |
| 3 | `lam data peumeurèntah nakeuh` | 3,499 |
| 4 | `nyoe lam data peumeurèntah` | 3,499 |
| 5 | `gampông nyoe lam data` | 3,499 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `nyoe lam data peumeurèntah nakeuh` | 3,499 |
| 2 | `gampông nyoe lam data peumeurèntah` | 3,499 |
| 3 | `lumbôi gampông nyoe lam data` | 3,498 |
| 4 | `acèh lumbôi gampông nyoe lam` | 3,495 |
| 5 | `lam data peumeurèntah nakeuh nè` | 3,489 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `e u` | 118,044 |
| 2 | `_ n` | 79,550 |
| 3 | `a n` | 69,741 |
| 4 | `h _` | 68,205 |
| 5 | `n g` | 67,768 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n g _` | 44,547 |
| 2 | `_ n a` | 31,665 |
| 3 | `_ b a` | 30,517 |
| 4 | `k e u` | 30,367 |
| 5 | `_ n y` | 26,591 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `e u h _` | 23,358 |
| 2 | `b a k _` | 23,289 |
| 3 | `_ d i _` | 21,170 |
| 4 | `k e u h` | 21,124 |
| 5 | `a k e u` | 20,698 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `k e u h _` | 21,003 |
| 2 | `n a k e u` | 20,623 |
| 3 | `a k e u h` | 20,621 |
| 4 | `_ n a k e` | 20,596 |
| 5 | `_ b a k _` | 18,136 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 224
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~60% 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 | 0.7505 | 1.682 | 4.34 | 36,359 | 25.0% |
| **1** | Subword | 0.8631 | 1.819 | 5.38 | 1,270 | 13.7% |
| **2** | Word | 0.2142 | 1.160 | 1.44 | 156,380 | 78.6% |
| **2** | Subword | 0.7734 | 1.709 | 4.50 | 6,829 | 22.7% |
| **3** | Word | 0.0653 | 1.046 | 1.11 | 222,450 | 93.5% |
| **3** | Subword | 0.7578 | 1.691 | 3.55 | 30,660 | 24.2% |
| **4** | Word | 0.0241 🏆 | 1.017 | 1.04 | 244,189 | 97.6% |
| **4** | Subword | 0.5683 | 1.483 | 2.36 | 108,651 | 43.2% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `di ateuh keude neulop ii dari mèssana strabô ngön sichuan jinoë sukèë calameae aseuli 苗族 haraih`
2. `nakeuh saboh gampông nyoe bak wikidata data peumeurèntah nakeuh saboh spèsiès nibak volume 82 nibak ...`
3. `bak laman sunrisesunset com di jeupun lé shogakkukan nè seuneubeuet bak laman sunrisesunset com di s...`

**Context Size 2:**

1. `bak laman nasa data matauroe teubiet teunom di da irah bak laman geonames data gunong nyoe bak`
2. `gunong nyoe bak laman nasa data matauroe teubiet teunom di da irah ajyad 500 ngon 700 meté`
3. `nyoe bak wikidata data cuaca daerah gunong nyoe bak wikidata data cuaca daerah gunong nyoe bak laman`

**Context Size 3:**

1. `gunong nyoe bak wikidata data cuaca daerah gunong nyoe bak laman nasa data matauroe teubiet teunom d...`
2. `nyoe bak laman geonames data gunong nyoe bak laman geonames data gunong nyoe bak wikidata data cuaca...`
3. `lumbôi gampông nyoe lam data peumeurèntah nakeuh nè di pidie pidie`

**Context Size 4:**

1. `gunong nyoe bak laman nasa data matauroe teubiet teunom di da irah bak laman sunrisesunset com di ac...`
2. `acèh lumbôi gampông nyoe lam data peumeurèntah nakeuh nè di acèh rayek acèh rayek`
3. `gampông nyoe lam data peumeurèntah nakeuh nè di acèh seulatan raja acèh seulatan`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_peulopôt_onohoo`
2. `acoeuh_dd_teumph`
3. `nta'ôn_1,_ba),_b`

**Context Size 2:**

1. `eurènteuh_nè_deuh`
2. `_nakeuneuropinak_`
3. `an_acilife_39_nya`

**Context Size 3:**

1. `ng_di_daerah_cuaca`
2. `_najôh,_sha_peunaw`
3. `_bagoë_di_kabupatè`

**Context Size 4:**

1. `euh_babah_la'èn_nya`
2. `bak_jijak_ulee_stud`
3. `_di_muhammouaneuh'e`


### Key Findings

- **Best Predictability:** Context-4 (word) with 97.6% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (108,651 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 | 15,619 |
| Total Tokens | 516,593 |
| Mean Frequency | 33.07 |
| Median Frequency | 3 |
| Frequency Std Dev | 414.79 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | di | 21,222 |
| 2 | nakeuh | 20,611 |
| 3 | bak | 18,176 |
| 4 | acèh | 17,532 |
| 5 | nyoe | 13,191 |
| 6 | data | 11,090 |
| 7 | gunong | 10,023 |
| 8 | nyang | 9,056 |
| 9 | gampông | 8,794 |
| 10 | lam | 7,951 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | influence | 2 |
| 2 | across | 2 |
| 3 | represent | 2 |
| 4 | raising | 2 |
| 5 | ceremony | 2 |
| 6 | flown | 2 |
| 7 | reconstructions | 2 |
| 8 | bendera | 2 |
| 9 | bekas | 2 |
| 10 | jawatimu | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.1698 |
| R² (Goodness of Fit) | 0.995531 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 63.1% |
| Top 1,000 | 84.1% |
| Top 5,000 | 94.2% |
| Top 10,000 | 97.8% |

### Key Findings

- **Zipf Compliance:** R²=0.9955 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 63.1% of corpus
- **Long Tail:** 5,619 words needed for remaining 2.2% 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 | 0.4644 | 0.4250 | N/A | N/A |
| **mono_64d** | 64 | 0.1432 | 0.4182 | N/A | N/A |
| **mono_128d** | 128 | 0.0251 | 0.4207 | N/A | N/A |
| **aligned_32d** | 32 | 0.4644 🏆 | 0.4392 | 0.0240 | 0.1600 |
| **aligned_64d** | 64 | 0.1432 | 0.4223 | 0.0340 | 0.2120 |
| **aligned_128d** | 128 | 0.0251 | 0.4223 | 0.0540 | 0.2900 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.4644 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.4246. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 5.4% 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 | **0.411** | 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 |
|--------|----------|
| `-ge` | geudapeuta, geuseutöt, geutanyoe |
| `-me` | meuubah, meuasai, meupawôt |
| `-geu` | geudapeuta, geuseutöt, geutanyoe |
| `-meu` | meuubah, meuasai, meupawôt |
| `-pe` | perdagangan, peunténg, peuradaban |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-ng` | lambéng, peunténg, gadông |
| `-an` | perdagangan, azerbaijan, pikeran |
| `-ah` | pamarèntah, meuubah, bhah |

### 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 |
|------|----------|------------------|----------|
| `eung` | 1.43x | 64 contexts | reung, jeung, meung |
| `uneu` | 1.75x | 28 contexts | uneun, runeu, meuneu |
| `euna` | 1.43x | 60 contexts | keuna, beuna, peuna |
| `euen` | 1.53x | 38 contexts | leuen, eueng, meuen |
| `ubeu` | 1.48x | 22 contexts | ubeut, neubeu, keubeu |
| `umeu` | 1.43x | 23 contexts | jumeu, geumeu, jeumeu |
| `meur` | 1.61x | 15 contexts | meurô, meuri, meurak |
| `beue` | 1.55x | 16 contexts | beuet, rabeue, abeuek |
| `teun` | 1.34x | 25 contexts | uteun, ateung, teuntè |
| `neub` | 1.61x | 14 contexts | neuba, neubeu, neubôh |
| `eune` | 1.65x | 12 contexts | meuneu, seuneu, jeuneh |
| `anga` | 1.33x | 23 contexts | langa, manga, panga |

### 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 |
|--------|--------|-----------|----------|
| `-ge` | `-ng` | 64 words | geumeujuang, geulumpang |
| `-pe` | `-an` | 54 words | permulaan, peumeréntahan |
| `-me` | `-ng` | 27 words | meuteureubang, meugang |
| `-pe` | `-ng` | 27 words | peuseunang, peujuang |
| `-me` | `-ah` | 21 words | meriah, meutuwah |
| `-ge` | `-ah` | 20 words | geuminah, geujajah |
| `-pe` | `-ah` | 15 words | pemerintah, peumeuréntah |
| `-me` | `-an` | 14 words | mediterranian, meurakan |
| `-ge` | `-an` | 6 words | geuritan, geulawan |

### 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 |
|------|-----------------|------------|------|
| geulumbang | **`geu-lumba-ng`** | 6.0 | `lumba` |
| geutanyong | **`geu-tanyo-ng`** | 6.0 | `tanyo` |
| geumeupakat | **`geu-meu-pakat`** | 6.0 | `pakat` |
| geulanggang | **`geu-langga-ng`** | 6.0 | `langga` |
| gelombang | **`ge-lomba-ng`** | 6.0 | `lomba` |
| meupangkat | **`meu-pangkat`** | 4.5 | `pangkat` |
| meuhubôngan | **`meu-hubô-ng-an`** | 4.5 | `hubô` |
| meujangeun | **`meu-jangeun`** | 4.5 | `jangeun` |
| meuneunguy | **`meu-neunguy`** | 4.5 | `neunguy` |
| meusayeuëp | **`meu-sayeuëp`** | 4.5 | `sayeuëp` |
| meupapeuen | **`meu-papeuen`** | 4.5 | `papeuen` |
| geupeuleumah | **`geu-pe-uleum-ah`** | 4.5 | `uleum` |
| meubintéh | **`meu-bintéh`** | 4.5 | `bintéh` |
| meupoliték | **`meu-politék`** | 4.5 | `politék` |
| meuteukeubi | **`meu-teukeubi`** | 4.5 | `teukeubi` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Acehnese 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 (4.93x) |
| N-gram | **2-gram** | Lowest perplexity (224) |
| Markov | **Context-4** | Highest predictability (97.6%) |
| 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: 2026-01-03 16:16:20*