README.md

---
language: ilo
language_name: Iloko
language_family: austronesian_philippine_northern
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_philippine_northern
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: 4.543
  - name: best_isotropy
    type: isotropy
    value: 0.8576
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Iloko** 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** | 3.747x | 3.75 | 0.1290% | 366,711 |
| **16k** | 4.060x | 4.06 | 0.1397% | 338,491 |
| **32k** | 4.334x | 4.34 | 0.1492% | 317,024 |
| **64k** | 4.543x 🏆 | 4.55 | 0.1564% | 302,462 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Ti tawen idi ket kadawyan a tawen a nangrugi iti Martes (iparang ti silpo ti nap...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ti ▁tawen ▁idi ▁ket ▁kadawyan ▁a ▁tawen ▁a ▁nangrugi ▁iti ... (+19 more)` | 29 |
| 16k | `▁ti ▁tawen ▁idi ▁ket ▁kadawyan ▁a ▁tawen ▁a ▁nangrugi ▁iti ... (+19 more)` | 29 |
| 32k | `▁ti ▁tawen ▁idi ▁ket ▁kadawyan ▁a ▁tawen ▁a ▁nangrugi ▁iti ... (+19 more)` | 29 |
| 64k | `▁ti ▁tawen ▁idi ▁ket ▁kadawyan ▁a ▁tawen ▁a ▁nangrugi ▁iti ... (+19 more)` | 29 |

**Sample 2:** `Ti tawen idi ket kadawyan a tawen a nangrugi iti Domingo (iparang ti silpo ti na...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ti ▁tawen ▁idi ▁ket ▁kadawyan ▁a ▁tawen ▁a ▁nangrugi ▁iti ... (+19 more)` | 29 |
| 16k | `▁ti ▁tawen ▁idi ▁ket ▁kadawyan ▁a ▁tawen ▁a ▁nangrugi ▁iti ... (+19 more)` | 29 |
| 32k | `▁ti ▁tawen ▁idi ▁ket ▁kadawyan ▁a ▁tawen ▁a ▁nangrugi ▁iti ... (+19 more)` | 29 |
| 64k | `▁ti ▁tawen ▁idi ▁ket ▁kadawyan ▁a ▁tawen ▁a ▁nangrugi ▁iti ... (+19 more)` | 29 |

**Sample 3:** `Ti tawen idi ket kadawyan a tawen a nangrugi iti Domingo (iparang ti silpo ti na...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ti ▁tawen ▁idi ▁ket ▁kadawyan ▁a ▁tawen ▁a ▁nangrugi ▁iti ... (+19 more)` | 29 |
| 16k | `▁ti ▁tawen ▁idi ▁ket ▁kadawyan ▁a ▁tawen ▁a ▁nangrugi ▁iti ... (+19 more)` | 29 |
| 32k | `▁ti ▁tawen ▁idi ▁ket ▁kadawyan ▁a ▁tawen ▁a ▁nangrugi ▁iti ... (+19 more)` | 29 |
| 64k | `▁ti ▁tawen ▁idi ▁ket ▁kadawyan ▁a ▁tawen ▁a ▁nangrugi ▁iti ... (+19 more)` | 29 |


### Key Findings

- **Best Compression:** 64k achieves 4.543x compression
- **Lowest UNK Rate:** 8k with 0.1290% 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 | 9,671 | 13.24 | 49,471 | 18.3% | 45.5% |
| **2-gram** | Subword | 205 🏆 | 7.68 | 3,758 | 74.6% | 99.5% |
| **3-gram** | Word | 23,415 | 14.52 | 90,863 | 12.7% | 32.8% |
| **3-gram** | Subword | 1,534 | 10.58 | 27,777 | 35.7% | 77.2% |
| **4-gram** | Word | 42,394 | 15.37 | 148,452 | 11.4% | 27.1% |
| **4-gram** | Subword | 7,324 | 12.84 | 148,845 | 21.9% | 51.1% |
| **5-gram** | Word | 29,789 | 14.86 | 103,807 | 12.6% | 30.6% |
| **5-gram** | Subword | 21,347 | 14.38 | 384,749 | 15.0% | 38.4% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `dagiti nagibasaran` | 11,555 |
| 2 | `maysa a` | 10,904 |
| 3 | `ket ti` | 10,192 |
| 4 | `a kas` | 9,434 |
| 5 | `daytoy ket` | 8,282 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `akinruar a silpo` | 7,499 |
| 2 | `dagiti akinruar a` | 7,494 |
| 3 | `dagiti nagibasaran dagiti` | 4,617 |
| 4 | `nagibasaran dagiti akinruar` | 4,453 |
| 5 | `ket maysa a` | 3,557 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `dagiti akinruar a silpo` | 7,484 |
| 2 | `nagibasaran dagiti akinruar a` | 4,453 |
| 3 | `dagiti nagibasaran dagiti akinruar` | 4,434 |
| 4 | `mula iti pamilia ti` | 2,523 |
| 5 | `ket ti sebbangan ti` | 2,099 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `nagibasaran dagiti akinruar a silpo` | 4,449 |
| 2 | `dagiti nagibasaran dagiti akinruar a` | 4,434 |
| 3 | `demograpia dagiti nagibasaran dagiti akinruar` | 1,659 |
| 4 | `ti mula iti pamilia ti` | 1,601 |
| 5 | `sebbangan ti mula iti pamilia` | 1,520 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a _` | 526,058 |
| 2 | `i _` | 499,068 |
| 3 | `t i` | 477,610 |
| 4 | `_ a` | 378,705 |
| 5 | `a n` | 376,214 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `t i _` | 426,029 |
| 2 | `_ a _` | 234,251 |
| 3 | `_ t i` | 225,448 |
| 4 | `i t i` | 197,634 |
| 5 | `a n _` | 128,518 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ t i _` | 218,539 |
| 2 | `i t i _` | 190,580 |
| 3 | `_ i t i` | 103,474 |
| 4 | `a g i t` | 91,630 |
| 5 | `d a g i` | 91,219 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ i t i _` | 102,341 |
| 2 | `d a g i t` | 90,946 |
| 3 | `a g i t i` | 87,738 |
| 4 | `g i t i _` | 87,510 |
| 5 | `_ k e t _` | 71,576 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 205
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~38% 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 | 0.7561 | 1.689 | 4.87 | 138,794 | 24.4% |
| **1** | Subword | 0.8199 | 1.765 | 5.06 | 2,636 | 18.0% |
| **2** | Word | 0.3125 | 1.242 | 1.94 | 673,934 | 68.7% |
| **2** | Subword | 0.7122 | 1.638 | 4.45 | 13,337 | 28.8% |
| **3** | Word | 0.1495 | 1.109 | 1.34 | 1,305,896 | 85.0% |
| **3** | Subword | 0.7826 | 1.720 | 4.17 | 59,341 | 21.7% |
| **4** | Word | 0.0702 🏆 | 1.050 | 1.12 | 1,742,668 | 93.0% |
| **4** | Subword | 0.7028 | 1.628 | 3.04 | 247,602 | 29.7% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `a pagbeddengan ti madang ti gunglo ti limba românăroronrum ron langeveld ti sistema sistema ti punto`
2. `ti pagsasao a mangiada ti turko nga antenada ken ti habitat dagiti nagibasaran triandra kadawyan iti`
3. `iti bukel kalapsan ti populasionna maysa kadagiti bukodda nga idi pimmusay otto warburg e daytoy ket`

**Context Size 2:**

1. `dagiti nagibasaran dagiti akinruar a silpo opisial a pagurasan ti nagbanagan daytoy a panagusar iti ...`
2. `maysa a maika 3 a klase nga ili iti probinsia ti cebu ket isu idi idiay estados`
3. `ket ti siudad ti tsina bagi ti ioc ti rambakan nga aldaw a kalendario iti kalendario a`

**Context Size 3:**

1. `dagiti akinruar a silpo ili ti quirino ti maddela nagtipunan cabarroguis aglipay ken diffun kaaduan ...`
2. `akinruar a silpo naenara opisial a portal ti gobierno opisial a sitio ti turismo ti karabakh siudad ...`
3. `dagiti nagibasaran dagiti akinruar a silpo siudad ti mehiko ciudad de méxico ken ti maika 7 a meridi...`

**Context Size 4:**

1. `dagiti akinruar a silpo directory of current japanese city leaders and outline of system japans evol...`
2. `nagibasaran dagiti akinruar a silpo website ti siudad ti san pablo siudad ti san pedro population 57...`
3. `dagiti nagibasaran dagiti akinruar a silpo opisial a website ti andaman ken nicobar grupo ti etniko ...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `a;_mangima_saket`
2. `_ril,_ng_ka-a_na`
3. `ikaba_kerapipa_m`

**Context Size 2:**

1. `a_aca_a_mūrīshimb`
2. `i_ngpo_ket_da_kam`
3. `ti_a_demics._mawe`

**Context Size 3:**

1. `ti_kaman_zimbahnam`
2. `_a_heogress:_annak`
3. `_ti_dagiti_filia_h`

**Context Size 4:**

1. `_ti_dua_nga_engling`
2. `iti_agarup_a_karaka`
3. `_iti_karl_edisiesto`


### Key Findings

- **Best Predictability:** Context-4 (word) with 93.0% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (247,602 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 | 60,623 |
| Total Tokens | 2,400,884 |
| Mean Frequency | 39.60 |
| Median Frequency | 4 |
| Frequency Std Dev | 1521.44 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | a | 237,485 |
| 2 | ti | 233,421 |
| 3 | iti | 103,626 |
| 4 | ket | 71,830 |
| 5 | dagiti | 62,492 |
| 6 | nga | 53,917 |
| 7 | ken | 48,636 |
| 8 | kadagiti | 24,755 |
| 9 | idi | 21,971 |
| 10 | maysa | 16,740 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | mainom | 2 |
| 2 | epektoda | 2 |
| 3 | medulla | 2 |
| 4 | nainom | 2 |
| 5 | pannakarimon | 2 |
| 6 | kannabinoide | 2 |
| 7 | agsarsarua | 2 |
| 8 | alingget | 2 |
| 9 | emetopilia | 2 |
| 10 | emetopobia | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0920 |
| R² (Goodness of Fit) | 0.998298 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 53.3% |
| Top 1,000 | 74.1% |
| Top 5,000 | 86.4% |
| Top 10,000 | 91.0% |

### Key Findings

- **Zipf Compliance:** R²=0.9983 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 53.3% of corpus
- **Long Tail:** 50,623 words needed for remaining 9.0% 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 | 0.8576 🏆 | 0.3336 | N/A | N/A |
| **mono_64d** | 64 | 0.8049 | 0.2671 | N/A | N/A |
| **mono_128d** | 128 | 0.6566 | 0.2245 | N/A | N/A |
| **aligned_32d** | 32 | 0.8576 | 0.3327 | 0.1020 | 0.4140 |
| **aligned_64d** | 64 | 0.8049 | 0.2687 | 0.1940 | 0.5560 |
| **aligned_128d** | 128 | 0.6566 | 0.2322 | 0.2440 | 0.6020 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.8576 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2765. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 24.4% 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 | **-0.203** | 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 |
|--------|----------|
| `-ma` | magnificent, matapos, mackinven |
| `-a` | abc, annonaceae, agtengtenggel |
| `-s` | saklawen, segregate, sinaugoro |
| `-na` | naipagpagarup, naipabaro, na2o |
| `-pa` | pagsasaoe, pait, pannakamatmati |
| `-b` | basle, bisitaen, begawan |
| `-ka` | katres, kalidasa, kababa |
| `-p` | pisinniflora, pagsasaoe, puesto |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-a` | curia, pisinniflora, daremdemda |
| `-n` | saklawen, positron, tatalan |
| `-o` | kodigo, naipabaro, puesto |
| `-s` | katres, matapos, oxus |
| `-an` | tatalan, begawan, tinwtawagan |
| `-na` | lehitimadona, pinarmekna, arrubayanna |
| `-e` | me, rourke, pagsasaoe |
| `-g` | temburong, dulong, aliping |

### 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 |
|------|----------|------------------|----------|
| `angi` | 1.91x | 71 contexts | angin, mangi, sangi |
| `dayt` | 2.60x | 17 contexts | dayty, dayta, dayto |
| `sion` | 1.93x | 43 contexts | pasion, bision, sesion |
| `asao` | 2.34x | 20 contexts | masao, sasao, wasao |
| `adag` | 2.23x | 21 contexts | nadag, adaga, kadagit |
| `ngga` | 1.78x | 42 contexts | ingga, anggal, rongga |
| `agsa` | 1.61x | 53 contexts | agsao, agsapa, bagsak |
| `aipa` | 1.65x | 41 contexts | naipa, maipa, taipa |
| `aika` | 1.76x | 29 contexts | maika, baikal, taikat |
| `abag` | 1.76x | 27 contexts | tabag, abaga, kabag |
| `abae` | 1.92x | 20 contexts | babae, babaen, ababaen |
| `silp` | 2.05x | 16 contexts | silpo, isilpo, insilpo |

### 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 |
|--------|--------|-----------|----------|
| `-na` | `-n` | 142 words | naminduan, nailawlawagan |
| `-pa` | `-n` | 123 words | pasuruan, patubuan |
| `-pa` | `-a` | 122 words | pannakakita, pagsinaenna |
| `-a` | `-a` | 117 words | agrepresenta, agdumaduma |
| `-na` | `-a` | 108 words | naipatulodda, nailata |
| `-na` | `-an` | 105 words | naminduan, nailawlawagan |
| `-s` | `-a` | 98 words | sinasina, sanana |
| `-pa` | `-an` | 97 words | pasuruan, patubuan |
| `-b` | `-a` | 85 words | biskleta, bella |
| `-ma` | `-a` | 83 words | malabarica, maipanunotanda |

### 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 |
|------|-----------------|------------|------|
| kalatakanna | **`kalatak-an-na`** | 7.5 | `an` |
| manggandat | **`manggan-da-t`** | 7.5 | `da` |
| matarigagay | **`matariga-g-ay`** | 7.5 | `g` |
| cavacoana | **`cavaco-an-a`** | 7.5 | `an` |
| nagunggunaan | **`nagunggu-na-an`** | 7.5 | `na` |
| gungunana | **`gungun-an-a`** | 7.5 | `an` |
| khoonmengiana | **`khoonmengi-an-a`** | 7.5 | `an` |
| kutubuano | **`kutubu-an-o`** | 7.5 | `an` |
| resultana | **`result-an-a`** | 7.5 | `an` |
| pransiskano | **`pransisk-an-o`** | 7.5 | `an` |
| stephanus | **`steph-an-us`** | 7.5 | `an` |
| tanghalan | **`tangh-al-an`** | 7.5 | `al` |
| mabaeoides | **`mabaeoi-d-es`** | 7.5 | `d` |
| kabasalan | **`kabas-al-an`** | 7.5 | `al` |
| binukbukodanna | **`binukbukod-an-na`** | 7.5 | `an` |

### 6.6 Linguistic Interpretation

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

---
## 7. Summary & Recommendations

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

### Production Recommendations

| Component | Recommended | Rationale |
|-----------|-------------|-----------|
| Tokenizer | **64k BPE** | Best compression (4.54x) |
| N-gram | **2-gram** | Lowest perplexity (205) |
| Markov | **Context-4** | Highest predictability (93.0%) |
| 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 04:17:29*