README.md

---
language: jv
language_name: Javanese
language_family: austronesian_javanese
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_javanese
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: 4.770
  - name: best_isotropy
    type: isotropy
    value: 0.8468
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Javanese** 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** | 3.761x | 3.76 | 0.0624% | 367,079 |
| **16k** | 4.158x | 4.16 | 0.0690% | 332,063 |
| **32k** | 4.504x | 4.51 | 0.0747% | 306,543 |
| **64k** | 4.770x 🏆 | 4.77 | 0.0791% | 289,433 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Lawang Tamang iku désa ing Kacamatan Kapuas Hulu, Kabupatèn Kapuas, Provinsi Kal...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁lawang ▁tam ang ▁iku ▁désa ▁ing ▁kacamatan ▁kapuas ▁hulu , ... (+13 more)` | 23 |
| 16k | `▁lawang ▁tam ang ▁iku ▁désa ▁ing ▁kacamatan ▁kapuas ▁hulu , ... (+13 more)` | 23 |
| 32k | `▁lawang ▁tam ang ▁iku ▁désa ▁ing ▁kacamatan ▁kapuas ▁hulu , ... (+13 more)` | 23 |
| 64k | `▁lawang ▁tam ang ▁iku ▁désa ▁ing ▁kacamatan ▁kapuas ▁hulu , ... (+13 more)` | 23 |

**Sample 2:** `Olimpiade Innsbruck iku tegesé bisa: Olimpiade Mangsa Adhem Olimpiade Mangsa Adh...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁olimpiade ▁in ns br uck ▁iku ▁tegesé ▁bisa : ▁olimpiade ... (+13 more)` | 23 |
| 16k | `▁olimpiade ▁in ns br uck ▁iku ▁tegesé ▁bisa : ▁olimpiade ... (+13 more)` | 23 |
| 32k | `▁olimpiade ▁in ns br uck ▁iku ▁tegesé ▁bisa : ▁olimpiade ... (+13 more)` | 23 |
| 64k | `▁olimpiade ▁innsbruck ▁iku ▁tegesé ▁bisa : ▁olimpiade ▁mangsa ▁adhem ▁olimpiade ... (+7 more)` | 17 |

**Sample 3:** `Tumbang Randang iku désa ing Kacamatan Timpah, Kabupatèn Kapuas, Provinsi Kalima...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁t umbang ▁r andang ▁iku ▁désa ▁ing ▁kacamatan ▁t imp ... (+15 more)` | 25 |
| 16k | `▁tumbang ▁r andang ▁iku ▁désa ▁ing ▁kacamatan ▁t imp ah ... (+14 more)` | 24 |
| 32k | `▁tumbang ▁r andang ▁iku ▁désa ▁ing ▁kacamatan ▁t imp ah ... (+14 more)` | 24 |
| 64k | `▁tumbang ▁r andang ▁iku ▁désa ▁ing ▁kacamatan ▁timpah , ▁kabupatèn ... (+12 more)` | 22 |


### Key Findings

- **Best Compression:** 64k achieves 4.770x compression
- **Lowest UNK Rate:** 8k with 0.0624% 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 | 53,400 | 15.70 | 220,522 | 10.1% | 24.5% |
| **2-gram** | Subword | 259 🏆 | 8.01 | 15,060 | 68.6% | 99.0% |
| **3-gram** | Word | 61,400 | 15.91 | 252,205 | 10.0% | 25.8% |
| **3-gram** | Subword | 2,364 | 11.21 | 78,931 | 26.6% | 70.5% |
| **4-gram** | Word | 77,247 | 16.24 | 361,150 | 9.9% | 27.0% |
| **4-gram** | Subword | 14,956 | 13.87 | 384,924 | 13.0% | 38.3% |
| **5-gram** | Word | 47,870 | 15.55 | 237,597 | 10.3% | 31.3% |
| **5-gram** | Subword | 61,146 | 15.90 | 1,130,643 | 8.1% | 24.8% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `pranala njaba` | 22,877 |
| 2 | `ya iku` | 21,546 |
| 3 | `désa ing` | 18,151 |
| 4 | `wonten ing` | 17,934 |
| 5 | `ing kacamatan` | 17,641 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `désa ing kacamatan` | 14,588 |
| 2 | `iku désa ing` | 12,656 |
| 3 | `pranala njaba situs` | 10,259 |
| 4 | `njaba situs resmi` | 7,571 |
| 5 | `provinsi jawa tengah` | 6,585 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `iku désa ing kacamatan` | 12,424 |
| 2 | `pranala njaba situs resmi` | 7,568 |
| 3 | `provinsi jawa tengah indonésia` | 5,971 |
| 4 | `njaba situs resmi kabupatèn` | 5,917 |
| 5 | `tengah indonésia uga delengen` | 4,463 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `pranala njaba situs resmi kabupatèn` | 5,917 |
| 2 | `jawa tengah indonésia uga delengen` | 4,458 |
| 3 | `provinsi jawa tengah indonésia uga` | 4,344 |
| 4 | `delengen pratélan désa ing nurwègen` | 3,052 |
| 5 | `uga delengen pratélan désa ing` | 3,052 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a n` | 2,445,496 |
| 2 | `n g` | 2,062,677 |
| 3 | `n _` | 1,386,666 |
| 4 | `a _` | 1,357,298 |
| 5 | `i n` | 1,235,038 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n g _` | 1,062,306 |
| 2 | `a n _` | 825,330 |
| 3 | `i n g` | 754,596 |
| 4 | `a n g` | 728,138 |
| 5 | `_ k a` | 616,864 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `i n g _` | 593,220 |
| 2 | `_ i n g` | 401,502 |
| 3 | `a n g _` | 300,987 |
| 4 | `l a n _` | 237,133 |
| 5 | `_ l a n` | 214,461 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ i n g _` | 314,002 |
| 2 | `_ l a n _` | 197,495 |
| 3 | `k a n g _` | 153,856 |
| 4 | `_ k a n g` | 151,621 |
| 5 | `n g _ k a` | 91,155 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 259
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~25% 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 | 0.9308 | 1.906 | 8.58 | 468,924 | 6.9% |
| **1** | Subword | 1.1692 | 2.249 | 7.47 | 10,119 | 0.0% |
| **2** | Word | 0.2944 | 1.226 | 1.76 | 4,009,882 | 70.6% |
| **2** | Subword | 0.5600 | 1.474 | 3.25 | 75,466 | 44.0% |
| **3** | Word | 0.0884 | 1.063 | 1.15 | 7,031,088 | 91.2% |
| **3** | Subword | 0.5549 | 1.469 | 3.14 | 244,687 | 44.5% |
| **4** | Word | 0.0284 🏆 | 1.020 | 1.04 | 8,087,514 | 97.2% |
| **4** | Subword | 0.6012 | 1.517 | 2.96 | 767,333 | 39.9% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `ing wajan utawa rusa sing nganggo cithakan kanggo mesin iki uga dadi sawijining omongané kepeksa nur...`
2. `lan sawisé sawatara organisasi kabèh kalungguhan punika kanthi dipundalaken kagem nyithak karakter é...`
3. `kang béda kanggo best lonely island caribbean at cbci siro malabar rajkot sumber daya ekonomi bank`

**Context Size 2:**

1. `pranala njaba situs resmi kabupatèn kendhal pranala njaba master wewengkon ing situs bps data desemb...`
2. `ya iku 55 20 00 dalu kanthi ritual kesurupan ing pungkasanipun simran remen kaliyan rara oyi diwasa`
3. `désa ing kacamatan tapin tengah suku bangsa wong sundha kalah lan nagis bilung uga karan nagara panc...`

**Context Size 3:**

1. `désa ing kacamatan tunjungan kurang luwih 12 157 kepala kulawarga lan 67 157 jiwa nglakokaké transmi...`
2. `iku désa ing kacamatan balongpanggang kabupatèn gresik provinsi jawa wétan indonésia rujukan uga del...`
3. `pranala njaba situs resmi kabupatèn batang`

**Context Size 4:**

1. `iku désa ing kacamatan samigaluh kabupatèn kulon praga daerah istimewa yogyakarta réferènsi ing kabu...`
2. `pranala njaba situs resmi luhur ing gorontalo`
3. `njaba situs resmi kabupatèn pekalongan`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_tedhrapeyi_koke`
2. `aspingka,_serero`
3. `ng_ahecahalosung`

**Context Size 2:**

1. `antiong._katuhati`
2. `ng_bittlenting_so`
3. `n_bis_ovièrènsijs`

**Context Size 3:**

1. `ng_séjéngge_misuma`
2. `an_r._kapusahané_k`
3. `ing_kudu_dhèwèké_j`

**Context Size 4:**

1. `ing_yahya_dhésèmber`
2. `_ing_wadhisi_déné_k`
3. `ang_dibat_mliginipu`


### Key Findings

- **Best Predictability:** Context-4 (word) with 97.2% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (767,333 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 | 206,658 |
| Total Tokens | 9,650,282 |
| Mean Frequency | 46.70 |
| Median Frequency | 4 |
| Frequency Std Dev | 1053.92 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ing | 316,085 |
| 2 | lan | 198,460 |
| 3 | kang | 92,968 |
| 4 | iku | 84,366 |
| 5 | sing | 79,278 |
| 6 | saka | 66,802 |
| 7 | ingkang | 59,183 |
| 8 | iki | 55,316 |
| 9 | taun | 54,241 |
| 10 | kabupatèn | 53,392 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | kaayom | 2 |
| 2 | paridhiri | 2 |
| 3 | lakwantara | 2 |
| 4 | bebakon | 2 |
| 5 | kadyan | 2 |
| 6 | nitikira | 2 |
| 7 | piwoleh | 2 |
| 8 | llms | 2 |
| 9 | marosa | 2 |
| 10 | letan | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0368 |
| R² (Goodness of Fit) | 0.991631 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 28.5% |
| Top 1,000 | 54.2% |
| Top 5,000 | 74.0% |
| Top 10,000 | 81.1% |

### Key Findings

- **Zipf Compliance:** R²=0.9916 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 28.5% of corpus
- **Long Tail:** 196,658 words needed for remaining 18.9% 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 | 0.8468 | 0.3355 | N/A | N/A |
| **mono_64d** | 64 | 0.7745 | 0.2697 | N/A | N/A |
| **mono_128d** | 128 | 0.7659 | 0.1964 | N/A | N/A |
| **aligned_32d** | 32 | 0.8468 🏆 | 0.3396 | 0.1700 | 0.4900 |
| **aligned_64d** | 64 | 0.7745 | 0.2725 | 0.2720 | 0.6640 |
| **aligned_128d** | 128 | 0.7659 | 0.1970 | 0.4020 | 0.7520 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.8468 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2684. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 40.2% R@1 in cross-lingual retrieval.
- **Recommendation:** 128d aligned for best cross-lingual performance

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

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

### 6.1 Productivity & Complexity

| Metric | Value | Interpretation | Recommendation |
|--------|-------|----------------|----------------|
| Productivity Index | **5.000** | High morphological productivity | Reliable analysis |
| Idiomaticity Gap | **0.262** | 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 |
|--------|----------|
| `-s` | sejajar, savages, sigifredo |
| `-a` | arepé, apla, aristizábal |
| `-ka` | kakangipun, kari, kambu |
| `-k` | kinali, kakangipun, kari |
| `-ma` | mansel, mangkunegoro, matar |
| `-di` | diah, dipompa, disebutnang |
| `-m` | mesiu, michail, mansel |
| `-sa` | savages, samsat, sandler |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-n` | sokawatèn, tekukan, kakangipun |
| `-a` | rayya, apla, archuleta |
| `-e` | oise, cave, scalable |
| `-an` | tekukan, panerbitan, pegelaran |
| `-s` | fasciatus, liturgis, savages |
| `-i` | kinali, nareswari, kari |
| `-ng` | dhuwung, nonggunong, widianing |
| `-g` | dhuwung, nonggunong, widianing |

### 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 |
|------|----------|------------------|----------|
| `angk` | 1.62x | 487 contexts | angka, angké, angki |
| `puni` | 2.39x | 38 contexts | punia, punik, punis |
| `nthi` | 2.24x | 47 contexts | knthi, anthi, sonthi |
| `nten` | 1.80x | 122 contexts | enten, onten, inten |
| `angg` | 1.40x | 471 contexts | anggy, anggo, anggi |
| `ngka` | 1.47x | 336 contexts | angka, ongka, ingka |
| `enga` | 1.54x | 237 contexts | menga, denga, engau |
| `gkan` | 2.05x | 60 contexts | angkan, igkang, ngkana |
| `ingk` | 1.63x | 161 contexts | ingka, singka, ingkah |
| `angi` | 1.49x | 229 contexts | tangi, rangi, angie |
| `ngin` | 1.63x | 128 contexts | ngina, nging, angin |
| `akak` | 1.71x | 93 contexts | lakak, sakak, kakak |

### 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 |
|--------|--------|-----------|----------|
| `-s` | `-n` | 129 words | sekuningan, suwukan |
| `-pa` | `-n` | 102 words | patuan, parwanosen |
| `-k` | `-n` | 91 words | kondhan, kin |
| `-di` | `-i` | 90 words | disigèni, dipungameli |
| `-s` | `-a` | 82 words | shimojima, spinella |
| `-ka` | `-n` | 82 words | karenggan, kamawen |
| `-di` | `-é` | 75 words | diwajibaké, dijodokaké |
| `-pa` | `-an` | 72 words | patuan, parengkuan |
| `-k` | `-an` | 60 words | kondhan, kutukan |
| `-a` | `-a` | 54 words | angkawijaya, anzola |

### 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 |
|------|-----------------|------------|------|
| duryudana | **`duryud-an-a`** | 7.5 | `an` |
| banjengan | **`banje-ng-an`** | 7.5 | `ng` |
| ngrencana | **`ngrenc-an-a`** | 7.5 | `an` |
| indowebster | **`indowebs-t-er`** | 7.5 | `t` |
| tengkorake | **`tengko-ra-ke`** | 7.5 | `ra` |
| dentawyanjana | **`dentawyanj-an-a`** | 7.5 | `an` |
| dhongkrak | **`dhongk-ra-k`** | 7.5 | `ra` |
| kayubiranga | **`kayubira-ng-a`** | 7.5 | `ng` |
| kathosana | **`kathos-an-a`** | 7.5 | `an` |
| tunjungan | **`tunju-ng-an`** | 7.5 | `ng` |
| dengannya | **`dengan-n-ya`** | 7.5 | `n` |
| västergötland | **`västergötl-an-d`** | 7.5 | `an` |
| romandini | **`romandi-n-i`** | 7.5 | `n` |
| kentingan | **`kenti-ng-an`** | 7.5 | `ng` |
| çuklapaksa | **`çuklapak-s-a`** | 7.5 | `s` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Javanese 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 (4.77x) |
| N-gram | **2-gram** | Lowest perplexity (259) |
| Markov | **Context-4** | Highest predictability (97.2%) |
| 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 06:50:22*