README.md

---
language: bcl
language_name: Central Bikol
language_family: austronesian_philippine_central
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_central
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: 4.810
  - name: best_isotropy
    type: isotropy
    value: 0.8247
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-03
---

# Central Bikol - Wikilangs Models
## Comprehensive Research Report & Full Ablation Study

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Central Bikol** 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** | 3.957x | 3.96 | 0.0152% | 354,491 |
| **16k** | 4.291x | 4.29 | 0.0165% | 326,860 |
| **32k** | 4.572x | 4.58 | 0.0176% | 306,791 |
| **64k** | 4.810x 🏆 | 4.81 | 0.0185% | 291,605 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `An sarong taon sa Gregoryanong kalendaryo. Enero Pebrero Marso Abril Mayo Hunyo ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁an ▁sarong ▁taon ▁sa ▁gregoryanong ▁kalendaryo . ▁enero ▁pebrero ▁marso ... (+9 more)` | 19 |
| 16k | `▁an ▁sarong ▁taon ▁sa ▁gregoryanong ▁kalendaryo . ▁enero ▁pebrero ▁marso ... (+9 more)` | 19 |
| 32k | `▁an ▁sarong ▁taon ▁sa ▁gregoryanong ▁kalendaryo . ▁enero ▁pebrero ▁marso ... (+9 more)` | 19 |
| 64k | `▁an ▁sarong ▁taon ▁sa ▁gregoryanong ▁kalendaryo . ▁enero ▁pebrero ▁marso ... (+9 more)` | 19 |

**Sample 2:** `Si Donald James "Donny" Lucas (Montreal) dating sarong Amerikanong entertainer.`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁si ▁d onald ▁james ▁" don ny " ▁luc as ... (+10 more)` | 20 |
| 16k | `▁si ▁donald ▁james ▁" don ny " ▁lucas ▁( mont ... (+7 more)` | 17 |
| 32k | `▁si ▁donald ▁james ▁" don ny " ▁lucas ▁( mont ... (+7 more)` | 17 |
| 64k | `▁si ▁donald ▁james ▁" don ny " ▁lucas ▁( mont ... (+7 more)` | 17 |

**Sample 3:** `An Yenon sarong baryo sa Abi na lugar kan gobyerno lokal sa Cross River State, N...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁an ▁y en on ▁sarong ▁baryo ▁sa ▁ab i ▁na ... (+18 more)` | 28 |
| 16k | `▁an ▁y en on ▁sarong ▁baryo ▁sa ▁ab i ▁na ... (+17 more)` | 27 |
| 32k | `▁an ▁yen on ▁sarong ▁baryo ▁sa ▁abi ▁na ▁lugar ▁kan ... (+15 more)` | 25 |
| 64k | `▁an ▁yen on ▁sarong ▁baryo ▁sa ▁abi ▁na ▁lugar ▁kan ... (+15 more)` | 25 |


### Key Findings

- **Best Compression:** 64k achieves 4.810x compression
- **Lowest UNK Rate:** 8k with 0.0152% 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 | 29,762 | 14.86 | 139,543 | 13.5% | 31.1% |
| **2-gram** | Subword | 215 🏆 | 7.75 | 6,829 | 72.7% | 99.3% |
| **3-gram** | Word | 81,081 | 16.31 | 219,146 | 7.5% | 19.3% |
| **3-gram** | Subword | 1,801 | 10.81 | 46,307 | 33.2% | 73.8% |
| **4-gram** | Word | 128,131 | 16.97 | 304,782 | 9.2% | 17.0% |
| **4-gram** | Subword | 10,353 | 13.34 | 249,114 | 18.9% | 43.8% |
| **5-gram** | Word | 55,135 | 15.75 | 164,721 | 16.0% | 24.8% |
| **5-gram** | Subword | 39,111 | 15.26 | 711,663 | 11.0% | 29.6% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `sa mga` | 30,516 |
| 2 | `an mga` | 27,434 |
| 3 | `kan mga` | 22,662 |
| 4 | `iyo an` | 17,275 |
| 5 | `nin mga` | 16,825 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `panluwas na takod` | 5,506 |
| 2 | `mga panluwas na` | 4,909 |
| 3 | `toltolan mga panluwas` | 2,791 |
| 4 | `para sa mga` | 2,778 |
| 5 | `igwa ining sukol` | 2,227 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `mga panluwas na takod` | 4,613 |
| 2 | `toltolan mga panluwas na` | 2,791 |
| 3 | `igwa ining sukol na` | 2,139 |
| 4 | `philippine standard geographic code` | 1,751 |
| 5 | `sa sensus kan igwa` | 1,728 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `toltolan mga panluwas na takod` | 2,656 |
| 2 | `sa sensus kan igwa ining` | 1,724 |
| 3 | `standard geographic code local governance` | 1,722 |
| 4 | `com philippine standard geographic code` | 1,722 |
| 5 | `philatlas com philippine standard geographic` | 1,722 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a n` | 1,358,991 |
| 2 | `a _` | 1,303,105 |
| 3 | `n _` | 1,232,546 |
| 4 | `_ s` | 834,968 |
| 5 | `n a` | 797,325 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a n _` | 702,654 |
| 2 | `_ n a` | 541,439 |
| 3 | `_ s a` | 524,860 |
| 4 | `n g _` | 465,207 |
| 5 | `_ k a` | 378,564 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ s a _` | 337,217 |
| 2 | `_ n a _` | 333,981 |
| 3 | `k a n _` | 236,687 |
| 4 | `_ k a n` | 232,949 |
| 5 | `_ a n _` | 213,433 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ k a n _` | 225,191 |
| 2 | `_ m g a _` | 166,824 |
| 3 | `_ n i n _` | 131,940 |
| 4 | `a s i n _` | 125,892 |
| 5 | `_ a s i n` | 125,534 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 215
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~30% 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 | 0.7779 | 1.715 | 6.29 | 329,127 | 22.2% |
| **1** | Subword | 0.9163 | 1.887 | 5.39 | 7,145 | 8.4% |
| **2** | Word | 0.3186 | 1.247 | 1.99 | 2,064,138 | 68.1% |
| **2** | Subword | 0.5336 | 1.448 | 3.35 | 38,469 | 46.6% |
| **3** | Word | 0.1355 | 1.098 | 1.28 | 4,087,355 | 86.5% |
| **3** | Subword | 0.6380 | 1.556 | 3.61 | 128,967 | 36.2% |
| **4** | Word | 0.0498 🏆 | 1.035 | 1.08 | 5,215,534 | 95.0% |
| **4** | Subword | 0.6487 | 1.568 | 3.06 | 465,409 | 35.1% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `sa tipan an apod na dinadalihigan kan taon kan komputasyon asin ipagbabalik sa tahaw kan taon`
2. `na nag eeksister an mga mimetikong kalibangbang patag dakol na coronet an pahayag tanganing ipabisto...`
3. `an mga komposisyon kan kompositor asin ngapit iyong watawat ang halaman asin gurutom suya sumo mga`

**Context Size 2:**

1. `sa mga minasunod the crucifixion saint anthony wisconsin si gross sarong multi partidong estado kata...`
2. `an mga osipon sarong babaeng kustomer ining lalaki winaki siya nin labing 300 bilyon historya si jam...`
3. `kan mga aldaw bago ini ibugtak sa sitwasyon kan halawig na kasaysayan asin sarong best seller asin`

**Context Size 3:**

1. `panluwas na takod opisyal na websityo toltolan paadalan sa kabikolan`
2. `mga panluwas na takod philatlas com philippine standard geographic code local governance performance...`
3. `toltolan mga panluwas na takod philatlas com philippine standard geographic code local governance pe...`

**Context Size 4:**

1. `mga panluwas na takod agi agi kan kawat na scrabblre kinua 06 11 16 mga bagay bagay dapit sa`
2. `toltolan mga panluwas na takod si iu sa universal music japan koreanong artista`
3. `igwa ining sukol na 173 70 kilometro kwadrado na kadagaan asin namumugtak sa ikaduwang distrito an d...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_c_naco'a_nimgho`
2. `asinarosy_sig-em`
3. `n,_kursud_wanari`

**Context Size 2:**

1. `anta_pincion_they`
2. `a_cagkan_kabong_i`
3. `n_an_kahabaharopi`

**Context Size 3:**

1. `an_sa_laog,_asin_l`
2. `_na_at_sa_unra_san`
3. `_sa_na_lugang_nin_`

**Context Size 4:**

1. `_sa_kastian_communi`
2. `_na_dormasya_sa_pag`
3. `kan_iban.[3]_an_sa_`


### Key Findings

- **Best Predictability:** Context-4 (word) with 95.0% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (465,409 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 | 132,282 |
| Total Tokens | 5,940,352 |
| Mean Frequency | 44.91 |
| Median Frequency | 4 |
| Frequency Std Dev | 1779.06 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | sa | 339,632 |
| 2 | na | 337,250 |
| 3 | an | 230,137 |
| 4 | kan | 225,822 |
| 5 | mga | 168,493 |
| 6 | nin | 132,058 |
| 7 | asin | 125,726 |
| 8 | sarong | 62,546 |
| 9 | si | 54,313 |
| 10 | the | 42,923 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | akkuly | 2 |
| 2 | sucuk | 2 |
| 3 | zhaparova | 2 |
| 4 | altynbekov | 2 |
| 5 | wanatabe | 2 |
| 6 | kordon | 2 |
| 7 | sobringaran | 2 |
| 8 | khanid | 2 |
| 9 | ganish | 2 |
| 10 | niceno | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0205 |
| R² (Goodness of Fit) | 0.994695 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 43.3% |
| Top 1,000 | 63.7% |
| Top 5,000 | 79.4% |
| Top 10,000 | 85.4% |

### Key Findings

- **Zipf Compliance:** R²=0.9947 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 43.3% of corpus
- **Long Tail:** 122,282 words needed for remaining 14.6% 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 | 0.8247 | 0.3483 | N/A | N/A |
| **mono_64d** | 64 | 0.8238 | 0.2714 | N/A | N/A |
| **mono_128d** | 128 | 0.8094 | 0.1968 | N/A | N/A |
| **aligned_32d** | 32 | 0.8247 🏆 | 0.3494 | 0.2280 | 0.5780 |
| **aligned_64d** | 64 | 0.8238 | 0.2693 | 0.3700 | 0.7100 |
| **aligned_128d** | 128 | 0.8094 | 0.1977 | 0.4780 | 0.8080 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.8247 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2722. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 47.8% 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 | **-0.162** | 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 |
|--------|----------|
| `-pa` | pagraranggo, pandapog, pananakop |
| `-na` | naquit, nagana, nagmamato |
| `-ma` | maghelang, malos, mangyans |
| `-pag` | pagraranggo, pagrehistro, pagsasalin |
| `-pi` | pigrorokyaw, pigsaladawan, pigpapainitan |
| `-nag` | nagana, nagmamato, nagashino |
| `-ka` | kajaman, kalipunan, kambodya |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-n` | pigsaladawan, pigpapainitan, esperidion |
| `-a` | smegma, emanuela, estrela |
| `-ng` | maghelang, gyalwang, gansing |
| `-an` | pigsaladawan, pigpapainitan, kajaman |
| `-on` | esperidion, pasteurization, oryentasyon |
| `-ong` | silensyong, mapabulong, otong |
| `-ang` | maghelang, gyalwang, tatabang |

### 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 |
|------|----------|------------------|----------|
| `agka` | 1.94x | 108 contexts | pagka, nagka, magka |
| `inak` | 2.14x | 67 contexts | inakô, inaka, inakò |
| `atio` | 2.24x | 51 contexts | ratio, patio, matios |
| `syon` | 2.04x | 72 contexts | mosyon, nasyon, losyon |
| `agpa` | 1.87x | 88 contexts | ragpa, agpay, magpa |
| `hili` | 2.23x | 39 contexts | hilig, chili, hilir |
| `asyo` | 2.00x | 57 contexts | basyo, rasyo, nasyo |
| `ista` | 1.67x | 114 contexts | istar, bista, istat |
| `ndan` | 1.73x | 78 contexts | indan, ndang, andan |
| `agin` | 1.84x | 44 contexts | sagin, magin, nagin |
| `nagp` | 2.05x | 26 contexts | nagpe, nagpa, nagpur |
| `embr` | 2.14x | 22 contexts | membro, embryo, myembro |

### 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 |
|--------|--------|-----------|----------|
| `-pi` | `-n` | 77 words | pinagkukuanan, pinagkakaputan |
| `-pa` | `-n` | 75 words | paluan, painiton |
| `-ka` | `-n` | 75 words | kakagaton, katangaan |
| `-na` | `-a` | 74 words | nagbabareta, nagsaranga |
| `-pi` | `-an` | 72 words | pinagkukuanan, pinagkakaputan |
| `-pa` | `-a` | 67 words | pamareta, padilla |
| `-ka` | `-an` | 67 words | katangaan, kagadanan |
| `-na` | `-n` | 66 words | naiisihan, nahaman |
| `-ma` | `-a` | 64 words | manusela, mababareta |
| `-na` | `-an` | 56 words | naiisihan, nahaman |

### 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 |
|------|-----------------|------------|------|
| pinakamalumoy | **`pi-na-ka-ma-lumoy`** | 9.0 | `lumoy` |
| pinakamakosog | **`pi-na-ka-ma-kosog`** | 9.0 | `kosog` |
| pinakagrabeng | **`pi-na-ka-grabe-ng`** | 9.0 | `grabe` |
| pinakaposibleng | **`pi-na-ka-posible-ng`** | 9.0 | `posible` |
| pinakadarakula | **`pi-na-ka-darakula`** | 7.5 | `darakula` |
| pagpapasakit | **`pag-pa-pa-sakit`** | 7.5 | `sakit` |
| nakakasakop | **`na-ka-ka-sakop`** | 7.5 | `sakop` |
| nakakahimo | **`na-ka-ka-himo`** | 7.5 | `himo` |
| pinakasikat | **`pi-na-ka-sikat`** | 7.5 | `sikat` |
| nakakalihis | **`na-ka-ka-lihis`** | 7.5 | `lihis` |
| pagkakamukna | **`pag-ka-ka-mukna`** | 7.5 | `mukna` |
| nagpapaluwas | **`nag-pa-pa-luwas`** | 7.5 | `luwas` |
| pinakaligtas | **`pi-na-ka-ligtas`** | 7.5 | `ligtas` |
| nagpapamidbid | **`nag-pa-pa-midbid`** | 7.5 | `midbid` |
| nakakalayog | **`na-ka-ka-layog`** | 7.5 | `layog` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Central Bikol 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 (4.81x) |
| N-gram | **2-gram** | Lowest perplexity (215) |
| Markov | **Context-4** | Highest predictability (95.0%) |
| 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: 2026-01-03 18:57:54*