README.md

---
language: min
language_name: Minangkabau
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: 4.930
  - name: best_isotropy
    type: isotropy
    value: 0.7641
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Minangkabau** 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** | 4.013x | 4.02 | 1.3258% | 329,395 |
| **16k** | 4.388x | 4.39 | 1.4499% | 301,190 |
| **32k** | 4.696x | 4.70 | 1.5514% | 281,480 |
| **64k** | 4.930x 🏆 | 4.93 | 1.6290% | 268,080 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `VII Kota Ilir adolah marupoan kecamatan di , provinsi Jambi, Indonesia.`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁vii ▁kota ▁ilir ▁adolah ▁marupoan ▁kecamatan ▁di ▁, ▁provinsi ▁jambi ... (+3 more)` | 13 |
| 16k | `▁vii ▁kota ▁ilir ▁adolah ▁marupoan ▁kecamatan ▁di ▁, ▁provinsi ▁jambi ... (+3 more)` | 13 |
| 32k | `▁vii ▁kota ▁ilir ▁adolah ▁marupoan ▁kecamatan ▁di ▁, ▁provinsi ▁jambi ... (+3 more)` | 13 |
| 64k | `▁vii ▁kota ▁ilir ▁adolah ▁marupoan ▁kecamatan ▁di ▁, ▁provinsi ▁jambi ... (+3 more)` | 13 |

**Sample 2:** `Teluk Bayur adolah salah satu kelurahan nan talatak di Kecamatan Padang Selatan,...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁teluk ▁ba yur ▁adolah ▁salah ▁satu ▁kelurahan ▁nan ▁talatak ▁di ... (+11 more)` | 21 |
| 16k | `▁teluk ▁ba yur ▁adolah ▁salah ▁satu ▁kelurahan ▁nan ▁talatak ▁di ... (+11 more)` | 21 |
| 32k | `▁teluk ▁bayur ▁adolah ▁salah ▁satu ▁kelurahan ▁nan ▁talatak ▁di ▁kecamatan ... (+10 more)` | 20 |
| 64k | `▁teluk ▁bayur ▁adolah ▁salah ▁satu ▁kelurahan ▁nan ▁talatak ▁di ▁kecamatan ... (+10 more)` | 20 |

**Sample 3:** `Asembagus adolah salah satu kecamatan nan ado di kabupaten Situbondo, provinsi J...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁as em ba gus ▁adolah ▁salah ▁satu ▁kecamatan ▁nan ▁ado ... (+12 more)` | 22 |
| 16k | `▁as emba gus ▁adolah ▁salah ▁satu ▁kecamatan ▁nan ▁ado ▁di ... (+9 more)` | 19 |
| 32k | `▁as emba gus ▁adolah ▁salah ▁satu ▁kecamatan ▁nan ▁ado ▁di ... (+9 more)` | 19 |
| 64k | `▁as emba gus ▁adolah ▁salah ▁satu ▁kecamatan ▁nan ▁ado ▁di ... (+9 more)` | 19 |


### Key Findings

- **Best Compression:** 64k achieves 4.930x compression
- **Lowest UNK Rate:** 8k with 1.3258% 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 | 1,057 | 10.05 | 94,993 | 50.6% | 85.8% |
| **2-gram** | Subword | 192 🏆 | 7.59 | 4,402 | 75.5% | 99.8% |
| **3-gram** | Word | 859 | 9.75 | 112,851 | 49.9% | 90.4% |
| **3-gram** | Subword | 1,092 | 10.09 | 39,289 | 35.5% | 87.3% |
| **4-gram** | Word | 937 | 9.87 | 168,741 | 48.7% | 90.3% |
| **4-gram** | Subword | 3,273 | 11.68 | 206,748 | 22.8% | 69.6% |
| **5-gram** | Word | 950 | 9.89 | 127,468 | 47.6% | 90.3% |
| **5-gram** | Subword | 6,324 | 12.63 | 574,327 | 19.4% | 61.4% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `bagian dari` | 183,365 |
| 2 | `marupokan bagian` | 182,557 |
| 3 | `juo marupokan` | 166,571 |
| 4 | `spesies ko` | 152,353 |
| 5 | `filum arthropoda` | 147,993 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `marupokan bagian dari` | 182,519 |
| 2 | `juo marupokan bagian` | 166,458 |
| 3 | `filum arthropoda dan` | 147,988 |
| 4 | `dan kingdom animalia` | 147,982 |
| 5 | `arthropoda dan kingdom` | 147,982 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `juo marupokan bagian dari` | 166,457 |
| 2 | `filum arthropoda dan kingdom` | 147,982 |
| 3 | `arthropoda dan kingdom animalia` | 147,982 |
| 4 | `insecta filum arthropoda dan` | 146,143 |
| 5 | `marupokan bagian dari ordo` | 145,929 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `filum arthropoda dan kingdom animalia` | 147,982 |
| 2 | `insecta filum arthropoda dan kingdom` | 146,137 |
| 3 | `juo marupokan bagian dari ordo` | 145,928 |
| 4 | `ko juo marupokan bagian dari` | 135,752 |
| 5 | `spesies ko juo marupokan bagian` | 116,386 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a n` | 3,981,823 |
| 2 | `n _` | 2,034,622 |
| 3 | `o _` | 1,804,961 |
| 4 | `_ d` | 1,779,268 |
| 5 | `a r` | 1,764,250 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a n _` | 1,782,694 |
| 2 | `_ d a` | 986,653 |
| 3 | `a n g` | 887,768 |
| 4 | `_ m a` | 768,734 |
| 5 | `k a n` | 654,711 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a r i _` | 553,934 |
| 2 | `k a n _` | 509,174 |
| 3 | `_ d a r` | 455,453 |
| 4 | `d a r i` | 445,320 |
| 5 | `n a n _` | 366,988 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ d a r i` | 444,049 |
| 2 | `d a r i _` | 443,889 |
| 3 | `_ n a n _` | 338,174 |
| 4 | `o l a h _` | 271,974 |
| 5 | `a d o l a` | 266,632 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 192
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~61% 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 | 0.8491 | 1.801 | 6.60 | 276,703 | 15.1% |
| **1** | Subword | 0.8104 | 1.754 | 5.34 | 2,493 | 19.0% |
| **2** | Word | 0.2834 | 1.217 | 1.66 | 1,823,050 | 71.7% |
| **2** | Subword | 0.8288 | 1.776 | 5.53 | 13,312 | 17.1% |
| **3** | Word | 0.0784 | 1.056 | 1.13 | 3,012,393 | 92.2% |
| **3** | Subword | 0.8772 | 1.837 | 4.61 | 73,575 | 12.3% |
| **4** | Word | 0.0243 🏆 | 1.017 | 1.04 | 3,393,219 | 97.6% |
| **4** | Subword | 0.6939 | 1.618 | 3.17 | 338,906 | 30.6% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `dari spesies adolah langau dari famili dolichopodidae spesies ko ditagakan pado tanggal 20 7 ianuari...`
2. `nan mandapek the world spider catalog versi asli dan kingdom animalia evolusi kapuyuak panggali raks...`
3. `ko ditamukan pado taun dek linear di socorro pambantuakan sarupo asteroid nan indak cukuik gadang un...`

**Context Size 2:**

1. `bagian dari ordo diptera kelas insecta filum arthropoda dan kingdom animalia larva langau ko juo mar...`
2. `marupokan bagian dari ordo coleoptera kalas insecta filum arthropoda dan kingdom animalia langau ko ...`
3. `juo marupokan bagian dari asteroid apollo nan talatak di sabuak utamo asteroid ko tabantuak dari neb...`

**Context Size 3:**

1. `marupokan bagian dari ordo diptera kelas insecta filum arthropoda dan kingdom animalia larva kumbang...`
2. `juo marupokan bagian dari ordo diptera kelas insecta filum arthropoda dan kingdom animalia spesies i...`
3. `filum arthropoda dan kingdom animalia kumbang iko biasonyo panjangnyo sekitar 1 5 cm rujuakan minang`

**Context Size 4:**

1. `juo marupokan bagian dari genus sitticus dan ordo araneae namo ilmiah dari spesies ko partamo kali d...`
2. `filum arthropoda dan kingdom animalia larva larva kumbang iko biasonyo panjangnyo sekitar 1 5 cm ruj...`
3. `arthropoda dan kingdom animalia spesies iko mampunyoi insting predator nan agresif dan makanan utamo...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `auralu_s_s_kak_r`
2. `_ko_anthuram._ma`
3. `i_dangiekuilorud`

**Context Size 2:**

1. `an,_famonetesi_pa`
2. `n_pri_man_asutara`
3. `o_adoliastau_dang`

**Context Size 3:**

1. `an_g.,_25_marusan_`
2. `_daritidae_darikan`
3. `angau_ko_astera,_f`

**Context Size 4:**

1. `ari_asteroid_ko_juo`
2. `kan_spongiae._spesi`
3. `_dari_famili_cecido`


### 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 (338,906 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 | 133,774 |
| Total Tokens | 13,515,163 |
| Mean Frequency | 101.03 |
| Median Frequency | 3 |
| Frequency Std Dev | 3135.45 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | dari | 443,572 |
| 2 | nan | 338,511 |
| 3 | ko | 305,052 |
| 4 | adolah | 266,398 |
| 5 | dan | 241,396 |
| 6 | asteroid | 239,836 |
| 7 | di | 233,130 |
| 8 | langau | 216,756 |
| 9 | spesies | 197,096 |
| 10 | marupokan | 196,445 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | baurano | 2 |
| 2 | mampakontribusi | 2 |
| 3 | cajanus | 2 |
| 4 | cajan | 2 |
| 5 | barbahan | 2 |
| 6 | mamparangkan | 2 |
| 7 | antarindividu | 2 |
| 8 | manbuat | 2 |
| 9 | basiah | 2 |
| 10 | pencampuran | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.2882 |
| R² (Goodness of Fit) | 0.991583 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 61.8% |
| Top 1,000 | 85.8% |
| Top 5,000 | 92.1% |
| Top 10,000 | 94.3% |

### Key Findings

- **Zipf Compliance:** R²=0.9916 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 61.8% of corpus
- **Long Tail:** 123,774 words needed for remaining 5.7% 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 | 0.7641 | 0.3623 | N/A | N/A |
| **mono_64d** | 64 | 0.7363 | 0.3350 | N/A | N/A |
| **mono_128d** | 128 | 0.6903 | 0.2419 | N/A | N/A |
| **aligned_32d** | 32 | 0.7641 🏆 | 0.3624 | 0.0660 | 0.3420 |
| **aligned_64d** | 64 | 0.7363 | 0.3219 | 0.1680 | 0.4540 |
| **aligned_128d** | 128 | 0.6903 | 0.2438 | 0.2180 | 0.5680 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.7641 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.3112. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 21.8% 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 | **0.470** | 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` | serma, syuriah, sungaipasak |
| `-a` | asotus, amalgamasi, agrobisnis |
| `-ma` | mandeklarasian, malalak, maislamkan |
| `-b` | bapanguni, boreosignata, belgium |
| `-ba` | bapanguni, bangli, baklava |
| `-t` | toruaigiri, triceratops, tetabuhan |
| `-di` | dibaokkan, diawal, disetrika |
| `-pa` | parsamoan, pambakal, paetula |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-a` | cestonionerva, vania, serma |
| `-s` | asotus, agrobisnis, medis |
| `-n` | mandeklarasian, hulptroepen, pendanaan |
| `-an` | mandeklarasian, pendanaan, parsamoan |
| `-is` | agrobisnis, medis, internis |
| `-us` | asotus, iulius, angelus |
| `-i` | domini, amalgamasi, bapanguni |
| `-o` | naraco, kamiripannyo, maanganggapnyo |

### 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 |
|------|----------|------------------|----------|
| `anga` | 1.93x | 337 contexts | kanga, hanga, angan |
| `ster` | 2.39x | 97 contexts | stern, ester, aster |
| `aste` | 2.29x | 53 contexts | taste, astec, aster |
| `roid` | 3.10x | 18 contexts | viroid, tiroid, android |
| `mban` | 2.01x | 85 contexts | mbang, amban, lumban |
| `okan` | 2.42x | 35 contexts | pokan, tokan, rokan |
| `pter` | 3.12x | 13 contexts | aptera, pteron, ioptera |
| `eroi` | 2.58x | 18 contexts | boeroi, heroic, heroik |
| `ujua` | 1.95x | 38 contexts | mujua, jujua, rujua |
| `arup` | 2.29x | 20 contexts | swarup, sarupo, parupo |
| `ntua` | 1.91x | 34 contexts | ntuak, untua, luntua |
| `ruju` | 2.54x | 13 contexts | rujuk, rujua, rujuan |

### 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 |
|--------|--------|-----------|----------|
| `-k` | `-n` | 135 words | kesulitan, kern |
| `-s` | `-a` | 134 words | soera, sebangka |
| `-ma` | `-n` | 132 words | mangakibaikan, manuntun |
| `-a` | `-a` | 130 words | andinomyia, apocrypha |
| `-k` | `-an` | 121 words | kesulitan, kalulusan |
| `-p` | `-s` | 120 words | phrynoides, parapicalis |
| `-p` | `-a` | 119 words | pristina, pradana |
| `-ma` | `-an` | 119 words | mangakibaikan, mamaafan |
| `-s` | `-s` | 119 words | semigranosus, stenochironomus |
| `-pa` | `-n` | 116 words | parasen, padudukan |

### 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 |
|------|-----------------|------------|------|
| thienemanni | **`thienem-an-ni`** | 7.5 | `an` |
| variitibiata | **`variitibi-a-ta`** | 7.5 | `a` |
| cobaltina | **`cobalti-n-a`** | 7.5 | `n` |
| schistostephana | **`schistosteph-an-a`** | 7.5 | `an` |
| sheffordiana | **`sheffordi-an-a`** | 7.5 | `an` |
| albimanus | **`albim-an-us`** | 7.5 | `an` |
| bangunanyo | **`bangun-an-yo`** | 7.5 | `an` |
| vertebralis | **`vertebr-al-is`** | 7.5 | `al` |
| zimbabwensis | **`zimbabwen-s-is`** | 7.5 | `s` |
| pandeglang | **`pandegl-a-ng`** | 7.5 | `a` |
| dilandasi | **`dilanda-s-i`** | 7.5 | `s` |
| pangindraan | **`pangindr-a-an`** | 7.5 | `a` |
| kalashiani | **`kalashi-an-i`** | 7.5 | `an` |
| panasehaik | **`panaseh-a-ik`** | 7.5 | `a` |
| ditandotangani | **`ditandotang-an-i`** | 7.5 | `an` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Minangkabau 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 (192) |
| 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: 2026-01-10 12:05:56*