README.md

---
language: bbc
language_name: BBC
language_family: austronesian_batak
tags:
  - wikilangs
  - nlp
  - tokenizer
  - embeddings
  - n-gram
  - markov
  - wikipedia
  - monolingual
  - family-austronesian_batak
license: mit
library_name: wikilangs
pipeline_tag: feature-extraction
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: 4.433
  - name: best_isotropy
    type: isotropy
    value: 0.8253
  - name: vocabulary_size
    type: vocab
    value: 24711
generated: 2025-12-28
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **BBC** 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-gram)
- Markov chains (context of 1, 2, 3 and 4)
- Subword N-gram and Markov chains
- Embeddings in various sizes and dimensions
- 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. Summary & Recommendations](#6-summary--recommendations)
- [Metrics Glossary](#appendix-metrics-glossary--interpretation-guide)
- [Visualizations Index](#visualizations-index)

---
## 1. Tokenizer Evaluation

![Tokenizer Compression](visualizations/tokenizer_compression.png)

### Results

| Vocab Size | Compression | Avg Token Len | UNK Rate | Total Tokens |
|------------|-------------|---------------|----------|--------------|
| **8k** | 3.867x | 3.83 | 0.1235% | 1,466,873 |
| **16k** | 4.118x | 4.08 | 0.1315% | 1,377,727 |
| **32k** | 4.304x | 4.27 | 0.1375% | 1,318,061 |
| **64k** | 4.433x 🏆 | 4.39 | 0.1416% | 1,279,829 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Panjunan i ma sada huta  na adong di Kecamatan Petarukan, Kabupaten Pemalang, Pr...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁panj unan ▁i ▁ma ▁sada ▁huta ▁na ▁adong ▁di ▁kecamatan ... (+11 more)` | 21 |
| 16k | `▁panj unan ▁i ▁ma ▁sada ▁huta ▁na ▁adong ▁di ▁kecamatan ... (+11 more)` | 21 |
| 32k | `▁panj unan ▁i ▁ma ▁sada ▁huta ▁na ▁adong ▁di ▁kecamatan ... (+11 more)` | 21 |
| 64k | `▁panjunan ▁i ▁ma ▁sada ▁huta ▁na ▁adong ▁di ▁kecamatan ▁petarukan ... (+10 more)` | 20 |

**Sample 2:** `Ampapaga (Surat Batak:ᯀᯔ᯲ᯇᯇᯎ) i ma sada suansuanan na tubu di gadu ni hauma.

 P...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁amp ap aga ▁( surat ▁batak : ᯀ ᯔ᯲ ᯇ ... (+20 more)` | 30 |
| 16k | `▁amp ap aga ▁( surat ▁batak : ᯀ ᯔ᯲ᯇ ᯇ ... (+18 more)` | 28 |
| 32k | `▁amp apaga ▁( surat ▁batak : ᯀ ᯔ᯲ᯇᯇᯎ ) ▁i ... (+15 more)` | 25 |
| 64k | `▁ampapaga ▁( surat ▁batak : ᯀᯔ᯲ᯇᯇᯎ ) ▁i ▁ma ▁sada ... (+13 more)` | 23 |

**Sample 3:** `Sungapan i ma sada huta  na adong di Kecamatan Pemalang, Kabupaten Pemalang, Pro...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁sung apan ▁i ▁ma ▁sada ▁huta ▁na ▁adong ▁di ▁kecamatan ... (+11 more)` | 21 |
| 16k | `▁sung apan ▁i ▁ma ▁sada ▁huta ▁na ▁adong ▁di ▁kecamatan ... (+11 more)` | 21 |
| 32k | `▁sung apan ▁i ▁ma ▁sada ▁huta ▁na ▁adong ▁di ▁kecamatan ... (+11 more)` | 21 |
| 64k | `▁sungapan ▁i ▁ma ▁sada ▁huta ▁na ▁adong ▁di ▁kecamatan ▁pemalang ... (+10 more)` | 20 |


### Key Findings

- **Best Compression:** 64k achieves 4.433x compression
- **Lowest UNK Rate:** 8k with 0.1235% 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 Coverage](visualizations/ngram_coverage.png)

### Results

| N-gram | Perplexity | Entropy | Unique N-grams | Top-100 Coverage | Top-1000 Coverage |
|--------|------------|---------|----------------|------------------|-------------------|
| **2-gram** | 7,191 🏆 | 12.81 | 30,496 | 17.8% | 49.6% |
| **2-gram** | 209 🏆 | 7.70 | 2,868 | 75.2% | 99.1% |
| **3-gram** | 25,989 | 14.67 | 62,993 | 8.7% | 26.3% |
| **3-gram** | 1,399 | 10.45 | 19,851 | 36.8% | 80.6% |
| **4-gram** | 63,619 | 15.96 | 114,847 | 4.7% | 15.6% |
| **4-gram** | 6,375 | 12.64 | 81,359 | 18.9% | 52.8% |

### Top 5 N-grams by Size

**2-grams:**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `, jala` | 8,780 |
| 2 | `i ,` | 6,997 |
| 3 | `ᯬ ᯲` | 4,573 |
| 4 | `angka na` | 4,409 |
| 5 | `dung i` | 4,328 |

**3-grams:**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `anak ni si` | 1,611 |
| 2 | `, angka na` | 1,528 |
| 3 | `. 2 :` | 1,079 |
| 4 | `, anak ni` | 1,069 |
| 5 | `ᯰ ᯄ ᯦` | 1,063 |

**4-grams:**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `, anak ni si` | 906 |
| 2 | `ᯀ ᯰ ᯄ ᯦` | 686 |
| 3 | `ᯘ ᯪ ᯀᯉ ᯲` | 457 |
| 4 | `ᯬ ᯂᯖ ᯬ ᯲` | 432 |
| 5 | `on do hata ni` | 421 |


### Key Findings

- **Best Perplexity:** 2-gram with 209
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~53% of corpus
- **Recommendation:** 4-gram or 5-gram for best predictive performance

---
## 3. Markov Chain Evaluation

![Markov Entropy](visualizations/markov_entropy.png)

![Markov Branching](visualizations/markov_branching.png)

### Results

| Context | Avg Entropy | Perplexity | Branching Factor | Unique Contexts | Predictability |
|---------|-------------|------------|------------------|-----------------|----------------|
| **1** | 0.8281 | 1.775 | 6.09 | 50,531 | 17.2% |
| **1** | 1.0960 | 2.138 | 7.35 | 1,187 | 0.0% |
| **2** | 0.3983 | 1.318 | 2.17 | 307,642 | 60.2% |
| **2** | 0.8091 | 1.752 | 4.70 | 8,725 | 19.1% |
| **3** | 0.1982 | 1.147 | 1.41 | 667,901 | 80.2% |
| **3** | 0.7878 | 1.726 | 3.57 | 41,004 | 21.2% |
| **4** | 0.0988 🏆 | 1.071 | 1.16 | 943,940 | 90.1% |
| **4** | 0.5526 🏆 | 1.467 | 2.40 | 146,535 | 44.7% |

### Generated Text Samples

Below are text samples generated from each Markov chain model:

**Context Size 1:**

1. `, jala tu si wasti , jala dipadeakdeak hamu sian i jala peakkononna tu palaspalas pamurunan`
2. `. 12 : songon i hatangku tu bagasan saluhut angka ari puasa bintang di jolo i`
3. `: ida ma angka na margoar milo dohot paredangedangan , 2 : 35 : 25 pelean`

**Context Size 2:**

1. `, jala tudoshon gumora angka anak ni si joab soara ni sarune i laho ma ho ,`
2. `i , naung pinauli ni tangan ni halak pangarupa umpogo upa ni pambahenannasida . 99 : 7`
3. `ᯬ ᯲ ᯘᯞ ᯮ ᯂᯖ ᯮ ᯲ ᯑ ᯩ ᯇᯔ ᯲ ᯅᯂ ᯩ ᯉᯉ ᯲ ᯔ ᯉ`

**Context Size 3:**

1. `anak ni si ahilud do panuturi . 18 : 13 tung sura ahu parohon begu masa tu luat`
2. `, angka na so bangso hian ; marhite sian bangso na asing , jala marbarita goarmu di betlehem`
3. `. 2 : 15 alai tarrimas situtu ma si abner dohot di halak na marroha pangansion , ndang`

**Context Size 4:**

1. `, anak ni si ammiel . 3 : 6 ndada tu torop bangso , angka parhata bobang , manang`
2. `ᯀ ᯰ ᯄ ᯦ ᯂᯖ ᯉ ᯪ ᯘ ᯪ ᯅ ᯩ ᯀ ᯩ ᯒ ᯪ ᯑ ᯪ ᯉᯘ ᯪ`
3. `ᯘ ᯪ ᯀᯉ ᯲ ᯂᯔ ᯪ ᯐ ᯮ ᯔ ᯬ ᯞ ᯬ ᯉ ᯪ ᯑ ᯩ ᯅᯖ 2 :`


### Key Findings

- **Best Predictability:** Context-4 with 90.1% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (146,535 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 | 24,711 |
| Total Tokens | 1,019,541 |
| Mean Frequency | 41.26 |
| Median Frequency | 4 |
| Frequency Std Dev | 565.94 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ni | 35,101 |
| 2 | na | 34,088 |
| 3 | i | 33,056 |
| 4 | ma | 26,769 |
| 5 | di | 26,053 |
| 6 | tu | 20,450 |
| 7 | do | 19,163 |
| 8 | angka | 17,428 |
| 9 | jala | 14,598 |
| 10 | dohot | 13,609 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ᯝᯇᯞ | 2 |
| 2 | kayo | 2 |
| 3 | uttar | 2 |
| 4 | ltr | 2 |
| 5 | ebrima | 2 |
| 6 | 290px | 2 |
| 7 | td | 2 |
| 8 | height | 2 |
| 9 | 260px | 2 |
| 10 | 22251 | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.1956 |
| R² (Goodness of Fit) | 0.996705 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 52.8% |
| Top 1,000 | 78.6% |
| Top 5,000 | 91.7% |
| Top 10,000 | 95.9% |

### Key Findings

- **Zipf Compliance:** R²=0.9967 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 52.8% of corpus
- **Long Tail:** 14,711 words needed for remaining 4.1% 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)

### Model Comparison

| Model | Vocab Size | Dimension | Avg Norm | Std Norm | Isotropy |
|-------|------------|-----------|----------|----------|----------|
| **mono_32d** | 15,079 | 32 | 3.458 | 0.818 | 0.8253 🏆 |
| **mono_64d** | 15,079 | 64 | 3.886 | 0.738 | 0.7641 |
| **mono_128d** | 15,079 | 128 | 4.143 | 0.691 | 0.4668 |
| **embeddings_enhanced** | 0 | 0 | 0.000 | 0.000 | 0.0000 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.8253 (more uniform distribution)
- **Dimension Trade-off:** Higher dimensions capture more semantics but reduce isotropy
- **Vocabulary Coverage:** All models cover 15,079 words
- **Recommendation:** 100d for balanced semantic capture and efficiency

---
## 6. Summary & Recommendations

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

### Production Recommendations

| Component | Recommended | Rationale |
|-----------|-------------|-----------|
| Tokenizer | **32k BPE** | Best compression (4.43x) with low UNK rate |
| N-gram | **5-gram** | Lowest perplexity (209) |
| Markov | **Context-4** | Highest predictability (90.1%) |
| 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},
  publisher = {HuggingFace},
  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)
---
*Generated by Wikilangs Models Pipeline*

*Report Date: 2025-12-28 00:12:38*