README.md

---
language: tpi
language_name: Tok Pisin
language_family: germanic_west_anglofrisian
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-germanic_west_anglofrisian
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: 4.037
  - name: best_isotropy
    type: isotropy
    value: 0.0778
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-11
---

# Tok Pisin - Wikilangs Models
## Comprehensive Research Report & Full Ablation Study

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Tok Pisin** 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** | 3.783x | 3.79 | 0.8512% | 89,876 |
| **16k** | 4.037x 🏆 | 4.05 | 0.9083% | 84,227 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `emi wanpela taun long Soria provins, Castile na León, Spen. provins`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁emi ▁wanpela ▁taun ▁long ▁soria ▁provins , ▁castile ▁na ▁león ... (+4 more)` | 14 |
| 16k | `▁emi ▁wanpela ▁taun ▁long ▁soria ▁provins , ▁castile ▁na ▁león ... (+4 more)` | 14 |

**Sample 2:** `Kerema em i kapitol na taun bikpela tumas bilong Gulf provins long Papua Niugini...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁kerema ▁em ▁i ▁kapitol ▁na ▁taun ▁bikpela ▁tumas ▁bilong ▁gulf ... (+5 more)` | 15 |
| 16k | `▁kerema ▁em ▁i ▁kapitol ▁na ▁taun ▁bikpela ▁tumas ▁bilong ▁gulf ... (+5 more)` | 15 |

**Sample 3:** `Palermo em i wanpela taun long Sisili long kantri Itali. Em igat 678.492 manmeri...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁palermo ▁em ▁i ▁wanpela ▁taun ▁long ▁sisili ▁long ▁kantri ▁itali ... (+14 more)` | 24 |
| 16k | `▁palermo ▁em ▁i ▁wanpela ▁taun ▁long ▁sisili ▁long ▁kantri ▁itali ... (+14 more)` | 24 |


### Key Findings

- **Best Compression:** 16k achieves 4.037x compression
- **Lowest UNK Rate:** 8k with 0.8512% 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 | 765 | 9.58 | 1,782 | 41.8% | 85.7% |
| **2-gram** | Subword | 220 🏆 | 7.78 | 1,423 | 75.2% | 99.3% |
| **3-gram** | Word | 1,436 | 10.49 | 2,504 | 30.0% | 71.1% |
| **3-gram** | Subword | 1,252 | 10.29 | 7,330 | 36.8% | 80.5% |
| **4-gram** | Word | 3,719 | 11.86 | 5,474 | 17.7% | 43.3% |
| **4-gram** | Subword | 4,262 | 12.06 | 25,004 | 24.7% | 57.2% |
| **5-gram** | Word | 3,008 | 11.55 | 4,258 | 18.3% | 44.4% |
| **5-gram** | Subword | 7,473 | 12.87 | 36,235 | 20.2% | 48.1% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `em i` | 1,565 |
| 2 | `ol i` | 502 |
| 3 | `i gat` | 454 |
| 4 | `i bin` | 429 |
| 5 | `i wanpela` | 353 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `em i wanpela` | 277 |
| 2 | `em i intanet` | 170 |
| 3 | `i intanet kod` | 169 |
| 4 | `intanet kod bilong` | 168 |
| 5 | `i stap long` | 152 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `em i intanet kod` | 169 |
| 2 | `i intanet kod bilong` | 168 |
| 3 | `intanet kod bilong kantri` | 150 |
| 4 | `emi wanpela taun long` | 77 |
| 5 | `na león spen provins` | 73 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `em i intanet kod bilong` | 168 |
| 2 | `i intanet kod bilong kantri` | 150 |
| 3 | `provins castile na león spen` | 73 |
| 4 | `castile na león spen provins` | 73 |
| 5 | `wanpela taun long soria provins` | 70 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n g` | 9,784 |
| 2 | `o n` | 9,572 |
| 3 | `i _` | 8,914 |
| 4 | `l o` | 8,912 |
| 5 | `a _` | 8,788 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n g _` | 8,594 |
| 2 | `o n g` | 8,176 |
| 3 | `l o n` | 8,105 |
| 4 | `_ i _` | 4,901 |
| 5 | `_ b i` | 4,777 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `l o n g` | 8,042 |
| 2 | `o n g _` | 7,994 |
| 3 | `_ l o n` | 4,532 |
| 4 | `_ b i l` | 3,254 |
| 5 | `i l o n` | 3,199 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `l o n g _` | 7,945 |
| 2 | `_ l o n g` | 4,521 |
| 3 | `_ b i l o` | 3,195 |
| 4 | `b i l o n` | 3,195 |
| 5 | `i l o n g` | 3,194 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 220
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~48% 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 | 0.6340 | 1.552 | 3.43 | 10,055 | 36.6% |
| **1** | Subword | 0.6982 | 1.622 | 4.77 | 907 | 30.2% |
| **2** | Word | 0.2413 | 1.182 | 1.52 | 34,078 | 75.9% |
| **2** | Subword | 0.7924 | 1.732 | 4.03 | 4,305 | 20.8% |
| **3** | Word | 0.0987 | 1.071 | 1.16 | 51,273 | 90.1% |
| **3** | Subword | 0.6704 | 1.591 | 2.82 | 17,280 | 33.0% |
| **4** | Word | 0.0388 🏆 | 1.027 | 1.05 | 58,656 | 96.1% |
| **4** | Subword | 0.4282 | 1.346 | 1.86 | 48,609 | 57.2% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `i mas save pairap inglis molecule o latvijas republika latvija letonia lv sv toppdomän f bihain`
2. `long em ol kaikai long giraun papua niugini i save luksave olsem wanpela teritori bilong kantri`
3. `bilong zeus`

**Context Size 2:**

1. `em i wanpela distrik long is samar provins nau long taim ol i makim bill skate i`
2. `ol i yusim diatomit bilong wokim giaman stori bilong aeneas i gat mo rot tu tasol long`
3. `i gat biknem long lotu na bagarap na yumi igat rait long senisim asples o kantri inap`

**Context Size 3:**

1. `em i wanpela pasin bilong raitim ol tok olsem wan wan leta i makim wanpela krai dispela i`
2. `em i intanet kod bilong kantri siapan long esia 36 milion manmeri i stap abrus o waitpela manmeri`
3. `i intanet kod bilong kantri kiribas ki sv toppdomän k`

**Context Size 4:**

1. `em i intanet kod bilong kantri siamani de sv toppdomän d`
2. `i intanet kod bilong ascension insait kantri sen helena ascension na tristan da kuna ac`
3. `intanet kod bilong kantri solomon ailans slb`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_тступоваув_5976`
2. `alelutaina_binge`
3. `i_блав_le_lon_vi`

**Context Size 2:**

1. `ng_kong_van_wan_t`
2. `ong_kripenis:_лек`
3. `i_lusianwanpeleón`

**Context Size 3:**

1. `ng_holimigur_20_49`
2. `ong_mp3_familipim_`
3. `long_manmeri_inter`

**Context Size 4:**

1. `long_graun_bikpela_`
2. `ong_diksen_bilong_s`
3. `_long_haus_wanpela_`


### Key Findings

- **Best Predictability:** Context-4 (word) with 96.1% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (48,609 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 | 4,414 |
| Total Tokens | 68,197 |
| Mean Frequency | 15.45 |
| Median Frequency | 3 |
| Frequency Std Dev | 129.66 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | i | 4,945 |
| 2 | long | 4,543 |
| 3 | bilong | 3,174 |
| 4 | na | 2,044 |
| 5 | em | 2,006 |
| 6 | ol | 2,005 |
| 7 | wanpela | 937 |
| 8 | kantri | 793 |
| 9 | tok | 737 |
| 10 | olsem | 581 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | iucn | 2 |
| 2 | tudakpela | 2 |
| 3 | haitim | 2 |
| 4 | transformer | 2 |
| 5 | pletfom | 2 |
| 6 | nintendo | 2 |
| 7 | return | 2 |
| 8 | deluxe | 2 |
| 9 | allies | 2 |
| 10 | forgotten | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0374 |
| R² (Goodness of Fit) | 0.984176 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 58.9% |
| Top 1,000 | 85.6% |
| Top 5,000 | 0.0% |
| Top 10,000 | 0.0% |

### Key Findings

- **Zipf Compliance:** R²=0.9842 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 58.9% of corpus
- **Long Tail:** -5,586 words needed for remaining 100.0% 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 | 0.0778 🏆 | 0.6368 | N/A | N/A |
| **mono_64d** | 64 | 0.0142 | 0.6826 | N/A | N/A |
| **mono_128d** | 128 | 0.0027 | 0.6822 | N/A | N/A |
| **aligned_32d** | 32 | 0.0778 | 0.6434 | 0.0080 | 0.0900 |
| **aligned_64d** | 64 | 0.0142 | 0.6713 | 0.0120 | 0.0680 |
| **aligned_128d** | 128 | 0.0027 | 0.6897 | 0.0060 | 0.0560 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.0778 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.6677. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 1.2% R@1 in cross-lingual retrieval.
- **Recommendation:** 128d aligned for best cross-lingual performance

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

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

### 6.1 Productivity & Complexity

| Metric | Value | Interpretation | Recommendation |
|--------|-------|----------------|----------------|
| Productivity Index | **5.000** | High morphological productivity | Reliable analysis |
| Idiomaticity Gap | **-0.076** | 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` | sutim, stude, science |
| `-p` | ponoloji, papa, puławy |
| `-b` | bikpla, by, bringim |
| `-m` | montreal, mick, mindanao |
| `-a` | andersen, amamas, anderson |
| `-k` | katim, konversen, kainantu |
| `-t` | toledo, tuesday, territories |
| `-ma` | maui, mathew, masta |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-a` | despla, bikpla, papa |
| `-n` | circumcision, andersen, yunien |
| `-s` | opis, ogastas, territories |
| `-e` | stude, hangre, science |
| `-m` | sutim, lukautim, katim |
| `-en` | andersen, yunien, konversen |
| `-an` | giaman, independan, aislan |
| `-l` | montreal, medal, kaunsil |

### 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 |
|------|----------|------------------|----------|
| `tpel` | 1.44x | 7 contexts | etpela, retpela, sotpela |
| `inim` | 1.38x | 6 contexts | winim, minim, painim |
| `arap` | 1.37x | 6 contexts | narapla, bagarap, arapela |
| `amba` | 1.35x | 6 contexts | namba, nambafo, nambaut |
| `namb` | 1.36x | 5 contexts | namba, nambis, nambafo |

### 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` | 27 words | palawan, plen |
| `-m` | `-a` | 25 words | mipela, masta |
| `-s` | `-a` | 23 words | sevilla, sta |
| `-a` | `-n` | 22 words | andersen, anderson |
| `-s` | `-n` | 21 words | sandaun, suwisalan |
| `-s` | `-s` | 20 words | saiens, songs |
| `-a` | `-a` | 18 words | aljiria, angila |
| `-p` | `-a` | 17 words | papa, palencia |
| `-b` | `-a` | 17 words | bikpla, brata |
| `-k` | `-a` | 16 words | kaledonia, kompyuta |

### 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 |
|------|-----------------|------------|------|
| independans | **`independ-an-s`** | 7.5 | `an` |
| vientiane | **`vienti-an-e`** | 7.5 | `an` |
| pensilvania | **`pensilv-an-ia`** | 7.5 | `an` |
| filipinas | **`filipin-a-s`** | 7.5 | `a` |
| eksaminim | **`eksam-in-im`** | 7.5 | `in` |
| konstitusen | **`konstitu-s-en`** | 7.5 | `s` |
| deutschland | **`deutsch-la-nd`** | 7.5 | `la` |
| plantikain | **`planti-ka-in`** | 7.5 | `ka` |
| manmanmeri | **`m-an-manmeri`** | 7.5 | `manmeri` |
| toktokman | **`toktok-m-an`** | 7.5 | `m` |
| representim | **`re-present-im`** | 6.0 | `present` |
| periodical | **`periodic-al`** | 4.5 | `periodic` |
| champions | **`champion-s`** | 4.5 | `champion` |
| provinsel | **`provins-el`** | 4.5 | `provins` |
| internationale | **`international-e`** | 4.5 | `international` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Tok Pisin 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 | **16k BPE** | Best compression (4.04x) |
| N-gram | **2-gram** | Lowest perplexity (220) |
| Markov | **Context-4** | Highest predictability (96.1%) |
| 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},
  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: 2026-01-11 01:31:19*