README.md

---
language: pap
language_name: Papiamento
language_family: romance_creole
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-romance_creole
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: 4.536
  - name: best_isotropy
    type: isotropy
    value: 0.8452
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Papiamento** 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** | 3.813x | 3.82 | 0.1442% | 409,271 |
| **16k** | 4.143x | 4.15 | 0.1566% | 376,636 |
| **32k** | 4.392x | 4.39 | 0.1661% | 355,292 |
| **64k** | 4.536x 🏆 | 4.54 | 0.1715% | 343,992 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `ta un munisipio spano den provinsia di Soria. (provinsia)`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ta ▁un ▁munisipio ▁sp ano ▁den ▁provinsia ▁di ▁soria . ... (+3 more)` | 13 |
| 16k | `▁ta ▁un ▁munisipio ▁sp ano ▁den ▁provinsia ▁di ▁soria . ... (+3 more)` | 13 |
| 32k | `▁ta ▁un ▁munisipio ▁spano ▁den ▁provinsia ▁di ▁soria . ▁( ... (+2 more)` | 12 |
| 64k | `▁ta ▁un ▁munisipio ▁spano ▁den ▁provinsia ▁di ▁soria . ▁( ... (+2 more)` | 12 |

**Sample 2:** `Almazán ta un munisipio spaño den provinsia di Soria, region di Castilia i Leon....`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁alma z án ▁ta ▁un ▁munisipio ▁spaño ▁den ▁provinsia ▁di ... (+21 more)` | 31 |
| 16k | `▁alma z án ▁ta ▁un ▁munisipio ▁spaño ▁den ▁provinsia ▁di ... (+21 more)` | 31 |
| 32k | `▁almazán ▁ta ▁un ▁munisipio ▁spaño ▁den ▁provinsia ▁di ▁soria , ... (+19 more)` | 29 |
| 64k | `▁almazán ▁ta ▁un ▁munisipio ▁spaño ▁den ▁provinsia ▁di ▁soria , ... (+19 more)` | 29 |

**Sample 3:** `Tuvalu ta un pais oseatiko. E kapital di Tuvalu ta Vaiaku, Funafuti.`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁tu val u ▁ta ▁un ▁pais ▁os ea tiko . ... (+16 more)` | 26 |
| 16k | `▁tu valu ▁ta ▁un ▁pais ▁os ea tiko . ▁e ... (+14 more)` | 24 |
| 32k | `▁tuvalu ▁ta ▁un ▁pais ▁os ea tiko . ▁e ▁kapital ... (+11 more)` | 21 |
| 64k | `▁tuvalu ▁ta ▁un ▁pais ▁oseatiko . ▁e ▁kapital ▁di ▁tuvalu ... (+5 more)` | 15 |


### Key Findings

- **Best Compression:** 64k achieves 4.536x compression
- **Lowest UNK Rate:** 8k with 0.1442% 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 | 9,717 | 13.25 | 33,678 | 18.1% | 41.2% |
| **2-gram** | Subword | 238 🏆 | 7.89 | 2,724 | 71.0% | 99.3% |
| **3-gram** | Word | 25,247 | 14.62 | 49,901 | 8.0% | 24.5% |
| **3-gram** | Subword | 1,930 | 10.91 | 21,952 | 28.9% | 74.2% |
| **4-gram** | Word | 41,144 | 15.33 | 69,181 | 7.3% | 18.8% |
| **4-gram** | Subword | 10,003 | 13.29 | 104,371 | 14.9% | 42.3% |
| **5-gram** | Word | 22,273 | 14.44 | 38,166 | 11.0% | 24.1% |
| **5-gram** | Subword | 32,598 | 14.99 | 248,543 | 8.8% | 27.5% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `di e` | 14,647 |
| 2 | `el a` | 5,053 |
| 3 | `ta un` | 4,783 |
| 4 | `den e` | 4,574 |
| 5 | `e ta` | 4,109 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `un di e` | 1,033 |
| 2 | `di antias hulandes` | 757 |
| 3 | `for di e` | 740 |
| 4 | `na el a` | 652 |
| 5 | `ta e di` | 633 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `riba e kalènder gregoriano` | 548 |
| 2 | `ta un di e` | 408 |
| 3 | `yüni yüli ougùstùs sèptèmber` | 390 |
| 4 | `mei yüni yüli ougùstùs` | 385 |
| 5 | `aprel mei yüni yüli` | 384 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `riba e kalènder gregoriano ta` | 364 |
| 2 | `e kalènder gregoriano ta resta` | 364 |
| 3 | `mei yüni yüli ougùstùs sèptèmber` | 354 |
| 4 | `mart aprel mei yüni yüli` | 350 |
| 5 | `febrüari mart aprel mei yüni` | 345 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a _` | 273,635 |
| 2 | `_ d` | 174,552 |
| 3 | `i _` | 167,427 |
| 4 | `e _` | 140,158 |
| 5 | `n _` | 138,441 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ d i` | 117,044 |
| 2 | `d i _` | 106,629 |
| 3 | `_ e _` | 73,343 |
| 4 | `t a _` | 63,461 |
| 5 | `_ t a` | 56,841 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ d i _` | 103,952 |
| 2 | `_ t a _` | 38,893 |
| 3 | `n a n _` | 30,467 |
| 4 | `_ n a _` | 28,936 |
| 5 | `_ u n _` | 27,411 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ d e n _` | 20,331 |
| 2 | `o _ d i _` | 17,822 |
| 3 | `a _ d i _` | 17,622 |
| 4 | `_ d i _ e` | 17,588 |
| 5 | `n _ d i _` | 16,089 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 238
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~28% 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 | 1.0093 | 2.013 | 6.87 | 68,317 | 0.0% |
| **1** | Subword | 1.0745 | 2.106 | 8.28 | 829 | 0.0% |
| **2** | Word | 0.3505 | 1.275 | 1.93 | 468,008 | 65.0% |
| **2** | Subword | 0.9710 | 1.960 | 6.02 | 6,860 | 2.9% |
| **3** | Word | 0.1399 | 1.102 | 1.26 | 899,213 | 86.0% |
| **3** | Subword | 0.8488 | 1.801 | 4.26 | 41,291 | 15.1% |
| **4** | Word | 0.0522 🏆 | 1.037 | 1.08 | 1,126,785 | 94.8% |
| **4** | Subword | 0.6463 | 1.565 | 2.80 | 175,612 | 35.4% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `di artista boneriano e estadonan uni cu ta wordo proponi tin tambe ta pidié van hout`
2. `e estudio di prins claus den e lama durante e siguiente munisipionan monti olbia telti e`
3. `ta positive evaluation of invacion di e lista di promotor di tera di antia hulandes na`

**Context Size 2:**

1. `di e kontinente ta konta ku mas o ménos 3 km ku ta responsabel pa facilita e`
2. `el a keda publica pa prome biaha na pa martin lavallée ku tambe ta konosí komo pedro`
3. `ta un kolekshon di e peninsula di paraguaná situá den oséano pasífiko i na e klima specialmente`

**Context Size 3:**

1. `un di e sinkuenta 50 estado di merka aprel mei yüni yüli ougùstùs sèptèmber òktober novèmber desèmbe...`
2. `for di e costa submarino cu ta core for di hadicurari fishermens huts awendia sarah quita beach na`
3. `di antias hulandes un gran mayoria di estado practicamente tur estado ta parti di e cordon di serona...`

**Context Size 4:**

1. `riba e kalènder gregoriano ta resta 107 dia pa e aña terminá a sosodé mareshal deodoro da fonseca ta`
2. `ta un di e islanan sunda grandi na indonesia e ta e di tres industria di criminalidad mas grandi`
3. `yüni yüli ougùstùs sèptèmber òktober novèmber desèmber a nase yanüari febrüari 8 edgar palm músiko i...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_dita_anuliu_var`
2. `a_enamubestrona_`
3. `elon,_upas_baña_`

**Context Size 2:**

1. `a_aki,_lishonana.`
2. `_di_ta_guyty_arub`
3. `i_di_nal_di_su_ko`

**Context Size 3:**

1. `_di_un_un_henden_e`
2. `di_junichmonionnan`
3. `_e_makerkantorno_i`

**Context Size 4:**

1. `_di_59,45%_di_e_isl`
2. `_ta_wòrdu_i_eks-pro`
3. `nan_culturante_univ`


### Key Findings

- **Best Predictability:** Context-4 (word) with 94.8% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (175,612 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 | 34,175 |
| Total Tokens | 1,282,363 |
| Mean Frequency | 37.52 |
| Median Frequency | 4 |
| Frequency Std Dev | 827.80 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | di | 104,167 |
| 2 | e | 74,754 |
| 3 | ta | 39,477 |
| 4 | a | 31,746 |
| 5 | na | 29,351 |
| 6 | un | 27,802 |
| 7 | i | 24,418 |
| 8 | den | 20,552 |
| 9 | pa | 20,049 |
| 10 | ku | 16,379 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | maghalie | 2 |
| 2 | fei | 2 |
| 3 | kodirektor | 2 |
| 4 | influente | 2 |
| 5 | arubagrandis | 2 |
| 6 | struikelblok | 2 |
| 7 | recordnan | 2 |
| 8 | nacra | 2 |
| 9 | klep | 2 |
| 10 | guangdong | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0656 |
| R² (Goodness of Fit) | 0.993886 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 48.4% |
| Top 1,000 | 70.8% |
| Top 5,000 | 87.1% |
| Top 10,000 | 92.9% |

### Key Findings

- **Zipf Compliance:** R²=0.9939 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 48.4% of corpus
- **Long Tail:** 24,175 words needed for remaining 7.1% 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)


### 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 | 0.8452 | 0.3149 | N/A | N/A |
| **mono_64d** | 64 | 0.7555 | 0.2502 | N/A | N/A |
| **mono_128d** | 128 | 0.4621 | 0.2227 | N/A | N/A |
| **aligned_32d** | 32 | 0.8452 🏆 | 0.3064 | 0.0600 | 0.3160 |
| **aligned_64d** | 64 | 0.7555 | 0.2542 | 0.1520 | 0.4100 |
| **aligned_128d** | 128 | 0.4621 | 0.2259 | 0.1940 | 0.4780 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.8452 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2624. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 19.4% 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 | **-0.125** | 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` | suak, seccionnan, suleiman |
| `-a` | au, aradippou, anan |
| `-b` | bankario, be, biramento |
| `-p` | partituranan, ploaghe, placa |
| `-m` | mobilisá, missouri, magnesium |
| `-c` | citaat, cynanchum, circuito |
| `-k` | kritiká, kongregashonnan, konstruyendo |
| `-d` | depresion, dimensional, diskutí |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-n` | partituranan, kongregashonnan, seccionnan |
| `-o` | ratio, inkompleto, lazio |
| `-an` | partituranan, kongregashonnan, seccionnan |
| `-a` | uma, veterinaria, generalisa |
| `-e` | regime, be, ploaghe |
| `-on` | depresion, macron, wilson |
| `-s` | kisas, seychelles, libraries |
| `-te` | trieste, completamente, krítikamente |

### 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 |
|------|----------|------------------|----------|
| `acio` | 2.55x | 30 contexts | nacion, ignacio, ocacion |
| `asho` | 2.05x | 38 contexts | basho, nashon, pashon |
| `onan` | 1.88x | 53 contexts | conan, usonan, omonan |
| `ente` | 1.77x | 58 contexts | mente, lente, djente |
| `ento` | 1.96x | 36 contexts | lento, mento, sento |
| `amen` | 1.61x | 74 contexts | namen, samen, examen |
| `ista` | 1.81x | 44 contexts | vista, bista, lista |
| `enta` | 1.64x | 53 contexts | benta, kenta, menta |
| `ario` | 1.80x | 33 contexts | vario, mario, arion |
| `ster` | 1.61x | 49 contexts | stern, sterna, sister |
| `nter` | 1.67x | 41 contexts | inter, panter, hinter |
| `pres` | 1.54x | 56 contexts | presu, press, presa |

### 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` | 119 words | partidonan, patriarkanan |
| `-s` | `-n` | 108 words | sostenedónan, satisfaccion |
| `-p` | `-o` | 108 words | produsiendo, pensamento |
| `-k` | `-n` | 95 words | koalishon, koeiman |
| `-s` | `-o` | 93 words | spanjo, sosteniendo |
| `-a` | `-n` | 92 words | abdikashon, action |
| `-p` | `-a` | 92 words | predica, pornada |
| `-a` | `-o` | 91 words | anglicano, ansiano |
| `-d` | `-n` | 89 words | demostracion, desasternan |
| `-c` | `-a` | 88 words | cumbia, cuenca |

### 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 |
|------|-----------------|------------|------|
| analistanan | **`analist-an-an`** | 7.5 | `an` |
| silabanan | **`silab-an-an`** | 7.5 | `an` |
| proceduranan | **`procedur-an-an`** | 7.5 | `an` |
| interesnan | **`interes-n-an`** | 7.5 | `n` |
| valdeavellano | **`valdeavell-an-o`** | 7.5 | `an` |
| caracassana | **`caracass-an-a`** | 7.5 | `an` |
| canchanan | **`canch-an-an`** | 7.5 | `an` |
| kabbendans | **`kabbend-an-s`** | 7.5 | `an` |
| enkabesando | **`enkabes-an-do`** | 7.5 | `an` |
| critchley | **`critchl-e-y`** | 7.5 | `e` |
| musikante | **`musik-an-te`** | 7.5 | `an` |
| historiadornan | **`historiador-n-an`** | 7.5 | `n` |
| akshonistanan | **`akshonist-an-an`** | 7.5 | `an` |
| suramerikano | **`suramerik-an-o`** | 7.5 | `an` |
| peliculanan | **`pelicul-an-an`** | 7.5 | `an` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Papiamento 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 (4.54x) |
| N-gram | **2-gram** | Lowest perplexity (238) |
| Markov | **Context-4** | Highest predictability (94.8%) |
| 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 17:28:24*