---
language: kge
language_name: Komering
language_family: austronesian_other
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_other
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: 3.778
  - name: best_isotropy
    type: isotropy
    value: 0.4759
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Komering** 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** | 3.083x | 3.09 | 1.8482% | 160,535 |
| **16k** | 3.454x | 3.46 | 2.0700% | 143,330 |
| **32k** | 3.778x 🏆 | 3.78 | 2.2644% | 131,028 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `ण atawa Ṇa joda da salah osay huruf say uwat dilom Aksara Diwanagori.`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ ण ▁atawa ▁ṇ a ▁joda ▁da ▁salah ▁osay ▁huruf ... (+6 more)` | 16 |
| 16k | `▁ ण ▁atawa ▁ṇa ▁joda ▁da ▁salah ▁osay ▁huruf ▁say ... (+5 more)` | 15 |
| 32k | `▁ण ▁atawa ▁ṇa ▁joda ▁da ▁salah ▁osay ▁huruf ▁say ▁uwat ... (+4 more)` | 14 |

**Sample 2:** `Jolma Inuit Greenland joda da salah osay kalompok etnis say ngaman di Greenland,...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁jolma ▁in uit ▁greenland ▁joda ▁da ▁salah ▁osay ▁kalompok ▁etnis ... (+11 more)` | 21 |
| 16k | `▁jolma ▁inuit ▁greenland ▁joda ▁da ▁salah ▁osay ▁kalompok ▁etnis ▁say ... (+10 more)` | 20 |
| 32k | `▁jolma ▁inuit ▁greenland ▁joda ▁da ▁salah ▁osay ▁kalompok ▁etnis ▁say ... (+10 more)` | 20 |

**Sample 3:** `350 atawa Tolu ratus lima puluh joda da bilangan asli diantara 349 rik 351.`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ 3 5 0 ▁atawa ▁tolu ▁ratus ▁lima ▁puluh ▁joda ... (+14 more)` | 24 |
| 16k | `▁ 3 5 0 ▁atawa ▁tolu ▁ratus ▁lima ▁puluh ▁joda ... (+14 more)` | 24 |
| 32k | `▁ 3 5 0 ▁atawa ▁tolu ▁ratus ▁lima ▁puluh ▁joda ... (+14 more)` | 24 |


### Key Findings

- **Best Compression:** 32k achieves 3.778x compression
- **Lowest UNK Rate:** 8k with 1.8482% 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 | 1,114 | 10.12 | 3,446 | 38.1% | 72.7% |
| **2-gram** | Subword | 307 🏆 | 8.26 | 3,285 | 68.9% | 96.9% |
| **3-gram** | Word | 2,200 | 11.10 | 6,072 | 31.1% | 57.0% |
| **3-gram** | Subword | 2,122 | 11.05 | 14,216 | 30.7% | 72.1% |
| **4-gram** | Word | 6,974 | 12.77 | 15,206 | 19.9% | 36.4% |
| **4-gram** | Subword | 8,682 | 13.08 | 50,196 | 18.6% | 46.5% |
| **5-gram** | Word | 7,727 | 12.92 | 14,198 | 17.3% | 33.4% |
| **5-gram** | Subword | 16,325 | 13.99 | 74,544 | 15.7% | 37.3% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `joda da` | 3,206 |
| 2 | `salah osay` | 1,901 |
| 3 | `da salah` | 1,879 |
| 4 | `say uwat` | 827 |
| 5 | `di nagara` | 713 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `joda da salah` | 1,874 |
| 2 | `da salah osay` | 1,847 |
| 3 | `salah osay huruf` | 517 |
| 4 | `say uwat dilom` | 482 |
| 5 | `osay huruf say` | 451 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `joda da salah osay` | 1,842 |
| 2 | `da salah osay huruf` | 516 |
| 3 | `salah osay huruf say` | 451 |
| 4 | `huruf say uwat dilom` | 442 |
| 5 | `osay huruf say uwat` | 439 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `joda da salah osay huruf` | 514 |
| 2 | `da salah osay huruf say` | 451 |
| 3 | `osay huruf say uwat dilom` | 439 |
| 4 | `salah osay huruf say uwat` | 439 |
| 5 | `joda da salah osay basa` | 258 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a _` | 26,179 |
| 2 | `s a` | 16,199 |
| 3 | `a n` | 15,647 |
| 4 | `_ s` | 12,853 |
| 5 | `_ d` | 11,531 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ s a` | 8,334 |
| 2 | `a y _` | 7,563 |
| 3 | `d a _` | 7,016 |
| 4 | `_ d i` | 6,107 |
| 5 | `s a y` | 6,094 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `s a y _` | 6,012 |
| 2 | `_ s a y` | 3,997 |
| 3 | `a _ s a` | 3,635 |
| 4 | `a _ d a` | 3,356 |
| 5 | `d a _ d` | 3,303 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ s a y _` | 3,977 |
| 2 | `d a _ d a` | 3,221 |
| 3 | `j o d a _` | 3,221 |
| 4 | `_ j o d a` | 3,220 |
| 5 | `a _ d a _` | 3,216 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 307
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~37% 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 | 0.5728 | 1.487 | 3.11 | 22,053 | 42.7% |
| **1** | Subword | 0.4326 | 1.350 | 3.48 | 5,154 | 56.7% |
| **2** | Word | 0.1743 | 1.128 | 1.34 | 67,279 | 82.6% |
| **2** | Subword | 0.4377 | 1.354 | 2.63 | 17,949 | 56.2% |
| **3** | Word | 0.0679 | 1.048 | 1.12 | 88,369 | 93.2% |
| **3** | Subword | 0.4188 | 1.337 | 2.19 | 47,199 | 58.1% |
| **4** | Word | 0.0336 🏆 | 1.024 | 1.06 | 96,472 | 96.6% |
| **4** | Subword | 0.3340 | 1.261 | 1.72 | 103,240 | 66.6% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `say jak apuy aktif lagi tipakay dilom pira ngabungkal huruf á ە ـە یو жj j`
2. `da salah osay huruf jopang samantara huruf say tarakhir algar partamo say uwat dilom bahasa tajik`
3. `di batusangkar joda da salah osay provinsi jawa atau 00f rgb 255 255 255 255 255`

**Context Size 2:**

1. `joda da sambahyang juma at joda da salah osay kabupaten say uwat dilom alfabet sirilik huruf ҙ`
2. `salah osay basa sey tipakay di nagara cina say nuturko bahasa malta`
3. `da salah osay basa say digunako jolma seediq di nagara bagiyan andhra pradesh di nagara ghana tiyan`

**Context Size 3:**

1. `joda da salah osay huruf bujenis aksara semisilabis say tipakay untuk nulisko pira ngabungkal basa d...`
2. `da salah osay kalompok etnis say ngaman di nagara myanmar rik bangladesh tiyan ja makay bahasa jopan...`
3. `salah osay huruf say uwat dilom alfabet latin huruf ɣ balak ɣ ronik`

**Context Size 4:**

1. `joda da salah osay kabupaten di provinsi jawa timur say ibukutana di tuban`
2. `da salah osay huruf say uwat dilom aksara diwanagori`
3. `salah osay huruf say uwat dilom alfabet latin huruf ŷ balak ŷ ronik`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_loda_ـ_talan:_s`
2. `anuripangunasaha`
3. `iat._hang_dath_d`

**Context Size 2:**

1. `a_ngkassoriyang,_`
2. `sal_turuf_buruju_`
3. `an_voki_bagah_asa`

**Context Size 3:**

1. `_salah_sa_qiangya_`
2. `ay_kuta_say_kuta_p`
3. `da_sona_kim_perú_c`

**Context Size 4:**

1. `say_basa_kikai_rik_`
2. `_say_ngama_uzbek:_s`
3. `a_salah_osay_ditutu`


### Key Findings

- **Best Predictability:** Context-4 (word) with 96.6% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (103,240 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 | 10,204 |
| Total Tokens | 135,845 |
| Mean Frequency | 13.31 |
| Median Frequency | 3 |
| Frequency Std Dev | 94.14 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | say | 3,995 |
| 2 | da | 3,328 |
| 3 | di | 3,241 |
| 4 | joda | 3,221 |
| 5 | rik | 2,433 |
| 6 | huruf | 2,283 |
| 7 | osay | 1,955 |
| 8 | salah | 1,942 |
| 9 | basa | 1,659 |
| 10 | nagara | 1,381 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | vlaicu | 2 |
| 2 | luca | 2 |
| 3 | transnistria | 2 |
| 4 | ubakna | 2 |
| 5 | kampanye | 2 |
| 6 | renamed | 2 |
| 7 | eschate | 2 |
| 8 | maharaja | 2 |
| 9 | sawai | 2 |
| 10 | leenvertaling | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0145 |
| R² (Goodness of Fit) | 0.985126 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 43.9% |
| Top 1,000 | 73.1% |
| Top 5,000 | 91.9% |
| Top 10,000 | 99.7% |

### Key Findings

- **Zipf Compliance:** R²=0.9851 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 43.9% of corpus
- **Long Tail:** 204 words needed for remaining 0.3% 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 | 0.4759 🏆 | 0.4316 | N/A | N/A |
| **mono_64d** | 64 | 0.1580 | 0.4265 | N/A | N/A |
| **mono_128d** | 128 | 0.0253 | 0.4281 | N/A | N/A |
| **aligned_32d** | 32 | 0.4759 | 0.4371 | 0.0100 | 0.1340 |
| **aligned_64d** | 64 | 0.1580 | 0.4395 | 0.0340 | 0.1880 |
| **aligned_128d** | 128 | 0.0253 | 0.4352 | 0.0600 | 0.2700 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.4759 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.4330. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 6.0% 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 | **-0.104** | 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` | sanjar, serbia, sipa |
| `-k` | kʼani, korcing, kapulauan |
| `-t` | tokoh, tujia, tsʰ |
| `-b` | baloo, buvariasi, bangli |
| `-a` | animation, atlanta, about |
| `-m` | map, moksha, malik |
| `-p` | pir, protestan, phapphā |
| `-ka` | kapulauan, kalumpukartikulasi, kaum |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-n` | animation, protestan, amazon |
| `-a` | isina, atlanta, moksha |
| `-an` | protestan, kapulauan, sabduwan |
| `-i` | kʼani, asingipabunyi, buvariasi |
| `-o` | baloo, hulontalo, euro |
| `-ng` | halimawong, korcing, badung |
| `-g` | halimawong, korcing, badung |
| `-na` | isina, saunina, buguna |

### 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 |
|------|----------|------------------|----------|
| `amba` | 1.74x | 31 contexts | hamba, tambah, sambal |
| `anda` | 1.61x | 39 contexts | tanda, danda, banda |
| `atan` | 1.88x | 20 contexts | katan, datang, ikatan |
| `ngga` | 1.67x | 29 contexts | unggas, dangga, unggak |
| `anta` | 1.85x | 18 contexts | ganta, antar, manta |
| `angg` | 1.72x | 20 contexts | anggur, dangga, manggo |
| `angk` | 1.67x | 22 contexts | angka, nangka, bangka |
| `aran` | 1.70x | 19 contexts | jarang, barani, ataran |
| `anga` | 1.68x | 19 contexts | nanga, manga, sanga |
| `ngan` | 1.75x | 16 contexts | nganak, pangan, mongan |
| `tara` | 1.79x | 12 contexts | utara, ataran, satara |
| `tang` | 1.77x | 12 contexts | datang, tangut, batang |

### 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 |
|--------|--------|-----------|----------|
| `-p` | `-n` | 92 words | pembaptisan, pangombangan |
| `-k` | `-n` | 92 words | kagiatan, kakayoan |
| `-k` | `-an` | 82 words | kagiatan, kakayoan |
| `-p` | `-an` | 81 words | pembaptisan, pangombangan |
| `-t` | `-o` | 77 words | tihaluko, timanfaatko |
| `-s` | `-a` | 48 words | sirkasia, sinhala |
| `-b` | `-n` | 46 words | bualiran, butopatan |
| `-p` | `-a` | 45 words | posisina, persiya |
| `-m` | `-a` | 42 words | maha, moldova |
| `-s` | `-n` | 40 words | shan, swan |

### 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 |
|------|-----------------|------------|------|
| mazanderani | **`mazander-an-i`** | 7.5 | `an` |
| anggotana | **`anggot-an-a`** | 7.5 | `an` |
| kamungkinan | **`kamungki-n-an`** | 7.5 | `n` |
| padananna | **`padan-an-na`** | 7.5 | `an` |
| pakngapuluh | **`pa-k-ngapuluh`** | 7.5 | `ngapuluh` |
| vientiane | **`vienti-an-e`** | 7.5 | `an` |
| kahikanna | **`kahik-an-na`** | 7.5 | `an` |
| panggunaan | **`panggu-na-an`** | 7.5 | `na` |
| batanghari | **`batangh-a-ri`** | 7.5 | `a` |
| rangkaian | **`rangka-i-an`** | 7.5 | `i` |
| sinngguway | **`s-in-ngguway`** | 7.5 | `ngguway` |
| louisiana | **`louisi-an-a`** | 7.5 | `an` |
| pamarintahanna | **`pamarintah-an-na`** | 7.5 | `an` |
| panghubung | **`pa-ng-hubung`** | 7.5 | `hubung` |
| homorganik | **`homorg-an-ik`** | 7.5 | `an` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Komering 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 | **32k BPE** | Best compression (3.78x) |
| N-gram | **2-gram** | Lowest perplexity (307) |
| Markov | **Context-4** | Highest predictability (96.6%) |
| 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 07:36:21*