README.md

---
language: iba
language_name: Iban
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
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: 5.202
  - name: best_isotropy
    type: isotropy
    value: 0.8124
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Iban** 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** | 4.581x | 4.58 | 0.1303% | 239,370 |
| **16k** | 4.888x | 4.89 | 0.1391% | 224,316 |
| **32k** | 5.091x | 5.09 | 0.1449% | 215,386 |
| **64k** | 5.202x 🏆 | 5.21 | 0.1480% | 210,759 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Gawai, Sebuah kampung di Chitwan, Nepal . Gawai Dayak, pengerami ninting taun ti...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁gawai , ▁sebuah ▁kampung ▁di ▁ch it wan , ▁nepal ... (+16 more)` | 26 |
| 16k | `▁gawai , ▁sebuah ▁kampung ▁di ▁chit wan , ▁nepal ▁. ... (+15 more)` | 25 |
| 32k | `▁gawai , ▁sebuah ▁kampung ▁di ▁chitwan , ▁nepal ▁. ▁gawai ... (+14 more)` | 24 |
| 64k | `▁gawai , ▁sebuah ▁kampung ▁di ▁chitwan , ▁nepal ▁. ▁gawai ... (+14 more)` | 24 |

**Sample 2:** `Bangkok tauka nama iya dalam jaku Thai, Krung Thep Maha Nakhon nya indu nengeri ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁bangkok ▁tauka ▁nama ▁iya ▁dalam ▁jaku ▁thai , ▁k rung ... (+17 more)` | 27 |
| 16k | `▁bangkok ▁tauka ▁nama ▁iya ▁dalam ▁jaku ▁thai , ▁k rung ... (+17 more)` | 27 |
| 32k | `▁bangkok ▁tauka ▁nama ▁iya ▁dalam ▁jaku ▁thai , ▁krung ▁thep ... (+15 more)` | 25 |
| 64k | `▁bangkok ▁tauka ▁nama ▁iya ▁dalam ▁jaku ▁thai , ▁krung ▁thep ... (+15 more)` | 25 |

**Sample 3:** `Lemari iya nya kabinet bediri ti tinggi tauka sederhana endur nyimpan gari tauka...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁lem ari ▁iya ▁nya ▁kabinet ▁bediri ▁ti ▁tinggi ▁tauka ▁sed ... (+13 more)` | 23 |
| 16k | `▁lemari ▁iya ▁nya ▁kabinet ▁bediri ▁ti ▁tinggi ▁tauka ▁sederhana ▁endur ... (+8 more)` | 18 |
| 32k | `▁lemari ▁iya ▁nya ▁kabinet ▁bediri ▁ti ▁tinggi ▁tauka ▁sederhana ▁endur ... (+8 more)` | 18 |
| 64k | `▁lemari ▁iya ▁nya ▁kabinet ▁bediri ▁ti ▁tinggi ▁tauka ▁sederhana ▁endur ... (+8 more)` | 18 |


### Key Findings

- **Best Compression:** 64k achieves 5.202x compression
- **Lowest UNK Rate:** 8k with 0.1303% 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 | 6,394 | 12.64 | 13,442 | 15.3% | 43.4% |
| **2-gram** | Subword | 194 🏆 | 7.60 | 1,944 | 77.0% | 99.7% |
| **3-gram** | Word | 9,236 | 13.17 | 13,930 | 9.9% | 32.2% |
| **3-gram** | Subword | 1,402 | 10.45 | 13,716 | 34.0% | 81.6% |
| **4-gram** | Word | 12,791 | 13.64 | 15,883 | 6.7% | 22.4% |
| **4-gram** | Subword | 6,509 | 12.67 | 60,183 | 17.8% | 51.0% |
| **5-gram** | Word | 5,997 | 12.55 | 7,027 | 8.9% | 30.2% |
| **5-gram** | Subword | 18,422 | 14.17 | 130,688 | 12.0% | 34.9% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `iya nya` | 2,053 |
| 2 | `dalam taun` | 1,897 |
| 3 | `pelilih menua` | 882 |
| 4 | `kereban sanding` | 782 |
| 5 | `kandang menua` | 689 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `dikelala enggau nama` | 415 |
| 2 | `garis entara menua` | 246 |
| 3 | `dalam taun iya` | 197 |
| 4 | `nyadi sebagi ari` | 179 |
| 5 | `web ke bukai` | 165 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `laman web ke bukai` | 158 |
| 2 | `kereban sanding laman web` | 78 |
| 3 | `mega dikelala enggau nama` | 74 |
| 4 | `sanding laman web ke` | 73 |
| 5 | `ti dikelala enggau nama` | 64 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `kereban sanding laman web ke` | 73 |
| 2 | `sanding laman web ke bukai` | 72 |
| 3 | `penyanding laman web ke bukai` | 45 |
| 4 | `bekunsi garis entara menua enggau` | 45 |
| 5 | `negeri sarawak kunsil negeri sarawak` | 44 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a _` | 110,486 |
| 2 | `n g` | 83,490 |
| 3 | `i _` | 77,339 |
| 4 | `e n` | 67,953 |
| 5 | `a n` | 64,094 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `e n g` | 32,899 |
| 2 | `_ p e` | 27,770 |
| 3 | `y a _` | 21,779 |
| 4 | `_ d i` | 21,511 |
| 5 | `n y a` | 21,129 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n g g a` | 16,842 |
| 2 | `_ n y a` | 16,502 |
| 3 | `_ e n g` | 16,010 |
| 4 | `e n g g` | 15,955 |
| 5 | `g a u _` | 15,431 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `e n g g a` | 15,887 |
| 2 | `n g g a u` | 15,423 |
| 3 | `_ e n g g` | 15,391 |
| 4 | `g g a u _` | 15,348 |
| 5 | `_ i y a _` | 9,735 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 194
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~35% 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 | 0.9852 | 1.980 | 6.85 | 34,574 | 1.5% |
| **1** | Subword | 0.7946 | 1.735 | 5.41 | 1,153 | 20.5% |
| **2** | Word | 0.3188 | 1.247 | 1.75 | 236,220 | 68.1% |
| **2** | Subword | 0.8091 | 1.752 | 4.73 | 6,234 | 19.1% |
| **3** | Word | 0.0977 | 1.070 | 1.16 | 410,924 | 90.2% |
| **3** | Subword | 0.7951 | 1.735 | 3.72 | 29,465 | 20.5% |
| **4** | Word | 0.0275 🏆 | 1.019 | 1.04 | 473,414 | 97.3% |
| **4** | Subword | 0.5984 | 1.514 | 2.54 | 109,660 | 40.2% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `enggau danau victoria lalu mangku pengawa iya ulih dikena ngumbai diri nyadi tuai republik india sel...`
2. `iya ari taun 212 iku lebuh 3 711 pampang eksekutif opis pelajar ba waterford sebagi ari`
3. `ba sarawak chunto pengawa sida penroses beratika sekat bansa bidayuh enggau tuai ba pendam ruti nya`

**Context Size 2:**

1. `iya nya sebengkah menuamultiple sources ba asia tenggara kereban sanding laman web ke bukai baka lil...`
2. `dalam taun lalu diaku enggau rasmi nya strok lalu ditangkan enggau pemeri sida lalu dimartir kena vi...`
3. `pelilih menua segamat muar enggau tangkak ba johor karipap dikelala enggau nama il santo sante bemac...`

**Context Size 3:**

1. `dikelala enggau nama highland fold scottish fold longhair longhair fold and coupari pansik udah mada...`
2. `garis entara menua thailand puangthong rungswasdisab thailands response to the threat of climate cha...`
3. `dalam taun iya peturun rose fortune siku peranak virginia ke nyadi polis indu keterubah di malaysia ...`

**Context Size 4:**

1. `laman web ke bukai aum besai gerempung bansa bansa beserakup dalam taun iya nerima anugerah indu pem...`
2. `kereban sanding laman web ke bukai lirik lagu tu ba lirik lagu iban chord gitar lagu tu enggau lagu`
3. `mega dikelala enggau nama tumpuk pendiau sitak pengawa bepilih enggau bagi mit mukim iya nyadi tuai ...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_r._sem_pag_sa'l`
2. `a_tany)1_e,_nga_`
3. `nya_bembermplung`

**Context Size 2:**

1. `a_sidur_bang,_ti_`
2. `ngul_ngka_megoret`
3. `i_iyadagayuh_peng`

**Context Size 3:**

1. `enggerika_nama_dik`
2. `_penya_sebeda_karn`
3. `ya_bic_dite_sebaju`

**Context Size 4:**

1. `nggau_dalam_taun_tu`
2. `_nyadika_limau)_dik`
3. `_english_ruhnu._haa`


### Key Findings

- **Best Predictability:** Context-4 (word) with 97.3% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (109,660 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 | 16,192 |
| Total Tokens | 490,947 |
| Mean Frequency | 30.32 |
| Median Frequency | 4 |
| Frequency Std Dev | 265.56 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | enggau | 15,341 |
| 2 | iya | 10,907 |
| 3 | ba | 10,320 |
| 4 | ti | 9,965 |
| 5 | nya | 9,469 |
| 6 | ke | 8,465 |
| 7 | ari | 7,379 |
| 8 | dalam | 5,806 |
| 9 | nyadi | 5,795 |
| 10 | taun | 5,418 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | verbum | 2 |
| 2 | tychy | 2 |
| 3 | miniaturowej | 2 |
| 4 | sztuki | 2 |
| 5 | profesjonalnej | 2 |
| 6 | wideo | 2 |
| 7 | nietypowe | 2 |
| 8 | sztalugi | 2 |
| 9 | zapałek | 2 |
| 10 | tuareg | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.2366 |
| R² (Goodness of Fit) | 0.987474 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 43.2% |
| Top 1,000 | 75.2% |
| Top 5,000 | 92.5% |
| Top 10,000 | 97.2% |

### Key Findings

- **Zipf Compliance:** R²=0.9875 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 43.2% of corpus
- **Long Tail:** 6,192 words needed for remaining 2.8% 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 | 0.8124 | 0.3506 | N/A | N/A |
| **mono_64d** | 64 | 0.4625 | 0.3269 | N/A | N/A |
| **mono_128d** | 128 | 0.0966 | 0.3153 | N/A | N/A |
| **aligned_32d** | 32 | 0.8124 🏆 | 0.3472 | 0.0680 | 0.3200 |
| **aligned_64d** | 64 | 0.4625 | 0.3265 | 0.0760 | 0.3900 |
| **aligned_128d** | 128 | 0.0966 | 0.3184 | 0.0900 | 0.3580 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.8124 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.3308. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 9.0% R@1 in cross-lingual retrieval.
- **Recommendation:** 128d aligned for best cross-lingual performance

---
## 6.  Morphological Analysis (Experimental)

This section presents an automated morphological analysis derived from the statistical divergence between word-level and subword-level models. By analyzing where subword predictability spikes and where word-level coverage fails, we can infer linguistic structures without supervised data.

### 6.1 Productivity & Complexity

| Metric | Value | Interpretation | Recommendation |
|--------|-------|----------------|----------------|
| Productivity Index | **5.000** | High morphological productivity | Reliable analysis |
| Idiomaticity Gap | **-0.134** | 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 |
|--------|----------|
| `-s` | sisal, siaran, sebilion |
| `-di` | diarkib, digambarka, dipendam |
| `-be` | bebilion, beting, besaing |
| `-a` | acutis, annie, alice |
| `-b` | bebilion, beting, barito |
| `-p` | perfectus, pansut, pengirau |
| `-m` | music, mutuska, materials |
| `-pe` | perfectus, pengirau, pengari |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-n` | telekomunikasyen, lateran, siaran |
| `-a` | mutuska, digambarka, ikea |
| `-s` | perfectus, acutis, materials |
| `-i` | nyapai, pengari, diganti |
| `-ng` | beting, besaing, petang |
| `-g` | beting, besaing, petang |
| `-an` | lateran, siaran, labuan |
| `-e` | annie, code, divide |

### 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 |
|------|----------|------------------|----------|
| `ngka` | 1.53x | 69 contexts | engka, angka, bangka |
| `enga` | 1.41x | 60 contexts | lenga, lengan, dengan |
| `ngga` | 1.49x | 39 contexts | rongga, anggap, enggay |
| `dang` | 1.58x | 30 contexts | udang, kadang, undang |
| `enya` | 1.50x | 35 contexts | menya, kenya, lenyau |
| `syen` | 1.79x | 19 contexts | fesyen, mosyen, aksyen |
| `engk` | 1.50x | 27 contexts | engka, engku, tengku |
| `nger` | 1.64x | 19 contexts | ngeri, ranger, ngerak |
| `ndan` | 1.60x | 20 contexts | undan, undang, pandan |
| `enge` | 1.71x | 16 contexts | mengeri, nengeri, avenged |
| `peny` | 1.70x | 16 contexts | penyu, penyah, penyai |
| `pema` | 1.44x | 27 contexts | pemar, pemai, pemali |

### 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 |
|--------|--------|-----------|----------|
| `-di` | `-a` | 111 words | dikuingka, diformalka |
| `-p` | `-n` | 105 words | penulin, patron |
| `-di` | `-ka` | 93 words | dikuingka, diformalka |
| `-k` | `-n` | 84 words | kondisyen, kolonisasyen |
| `-p` | `-a` | 82 words | panglima, praha |
| `-p` | `-an` | 69 words | pengkalan, persamaan |
| `-s` | `-n` | 65 words | sensasyen, sain |
| `-p` | `-i` | 64 words | perai, pagi |
| `-p` | `-ng` | 57 words | pesaing, putting |
| `-p` | `-g` | 57 words | pesaing, putting |

### 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 |
|------|-----------------|------------|------|
| malacañang | **`malacañ-a-ng`** | 7.5 | `a` |
| pengurang | **`pengu-ra-ng`** | 7.5 | `ra` |
| inchinnan | **`inchin-n-an`** | 7.5 | `n` |
| kandungan | **`kandu-ng-an`** | 7.5 | `ng` |
| pengeringat | **`pengeri-ng-at`** | 7.5 | `ng` |
| centuries | **`centur-i-es`** | 7.5 | `i` |
| pengerekai | **`pengere-ka-i`** | 7.5 | `ka` |
| pengerugi | **`penger-u-gi`** | 7.5 | `u` |
| prasekula | **`p-ra-sekula`** | 7.5 | `sekula` |
| nicholson | **`nichol-s-on`** | 7.5 | `s` |
| ngasingka | **`ngasi-ng-ka`** | 7.5 | `ng` |
| admission | **`a-d-mission`** | 7.5 | `mission` |
| inggerisjaku | **`inggerisja-k-u`** | 7.5 | `k` |
| interamna | **`interam-n-a`** | 7.5 | `n` |
| haubjerre | **`haubjer-r-e`** | 7.5 | `r` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Iban 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 (5.20x) |
| N-gram | **2-gram** | Lowest perplexity (194) |
| Markov | **Context-4** | Highest predictability (97.3%) |
| Embeddings | **100d** | Balanced semantic capture and isotropy |


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

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

### Tokenizer Metrics

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

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

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

### N-gram Model Metrics

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

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

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

### Markov Chain Metrics

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

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

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

### Vocabulary & Zipf's Law Metrics

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

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

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

### Word Embedding Metrics

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

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

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

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

### General Interpretation Guidelines

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


### Visualizations Index

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

---
## About This Project

### Data Source

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

### Project

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

### Maintainer

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

### Citation

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

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

### License

MIT License - Free for academic and commercial use.

### Links

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

*Report Date: 2026-01-10 03:50:03*