README.md

---
language: ht
language_name: Haitian Creole
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: 4.271
  - name: best_isotropy
    type: isotropy
    value: 0.7588
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

# Haitian Creole - Wikilangs Models
## Comprehensive Research Report & Full Ablation Study

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Haitian Creole** 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** | 3.548x | 3.55 | 0.3624% | 230,377 |
| **16k** | 3.848x | 3.85 | 0.3931% | 212,420 |
| **32k** | 4.091x | 4.10 | 0.4179% | 199,821 |
| **64k** | 4.271x 🏆 | 4.28 | 0.4363% | 191,397 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `se yon vil Etazini. Li sitye nan leta Kentucky. Chèf-lye li se ?. Istwa Istwa Po...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁se ▁yon ▁vil ▁etazini . ▁li ▁sitye ▁nan ▁leta ▁kentucky ... (+18 more)` | 28 |
| 16k | `▁se ▁yon ▁vil ▁etazini . ▁li ▁sitye ▁nan ▁leta ▁kentucky ... (+18 more)` | 28 |
| 32k | `▁se ▁yon ▁vil ▁etazini . ▁li ▁sitye ▁nan ▁leta ▁kentucky ... (+18 more)` | 28 |
| 64k | `▁se ▁yon ▁vil ▁etazini . ▁li ▁sitye ▁nan ▁leta ▁kentucky ... (+18 more)` | 28 |

**Sample 2:** `lane nan almanak gregoryen lane nan lòt almanak yo nonm`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁lane ▁nan ▁almanak ▁gregoryen ▁lane ▁nan ▁lòt ▁almanak ▁yo ▁nonm` | 10 |
| 16k | `▁lane ▁nan ▁almanak ▁gregoryen ▁lane ▁nan ▁lòt ▁almanak ▁yo ▁nonm` | 10 |
| 32k | `▁lane ▁nan ▁almanak ▁gregoryen ▁lane ▁nan ▁lòt ▁almanak ▁yo ▁nonm` | 10 |
| 64k | `▁lane ▁nan ▁almanak ▁gregoryen ▁lane ▁nan ▁lòt ▁almanak ▁yo ▁nonm` | 10 |

**Sample 3:** `Solit se yon sibstans ki fonn nan yon solisyon. Referans`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁sol it ▁se ▁yon ▁sibstans ▁ki ▁fon n ▁nan ▁yon ... (+4 more)` | 14 |
| 16k | `▁sol it ▁se ▁yon ▁sibstans ▁ki ▁fonn ▁nan ▁yon ▁solisyon ... (+2 more)` | 12 |
| 32k | `▁solit ▁se ▁yon ▁sibstans ▁ki ▁fonn ▁nan ▁yon ▁solisyon . ... (+1 more)` | 11 |
| 64k | `▁solit ▁se ▁yon ▁sibstans ▁ki ▁fonn ▁nan ▁yon ▁solisyon . ... (+1 more)` | 11 |


### Key Findings

- **Best Compression:** 64k achieves 4.271x compression
- **Lowest UNK Rate:** 8k with 0.3624% 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 | 8,092 | 12.98 | 142,856 | 26.4% | 57.0% |
| **2-gram** | Subword | 263 🏆 | 8.04 | 4,790 | 68.2% | 99.4% |
| **3-gram** | Word | 7,226 | 12.82 | 216,416 | 28.0% | 62.3% |
| **3-gram** | Subword | 2,065 | 11.01 | 40,230 | 29.0% | 73.0% |
| **4-gram** | Word | 6,609 | 12.69 | 326,152 | 29.8% | 66.2% |
| **4-gram** | Subword | 10,042 | 13.29 | 232,790 | 16.6% | 45.8% |
| **5-gram** | Word | 3,612 | 11.82 | 221,589 | 33.1% | 73.5% |
| **5-gram** | Subword | 31,768 | 14.96 | 722,497 | 11.3% | 35.2% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `se yon` | 67,026 |
| 2 | `istwa istwa` | 34,640 |
| 3 | `kèk lyen` | 34,549 |
| 4 | `referans kèk` | 34,144 |
| 5 | `nan etazini` | 32,194 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `referans kèk lyen` | 33,800 |
| 2 | `se yon vil` | 31,899 |
| 3 | `kèk lyen nan` | 24,836 |
| 4 | `yon vil nan` | 23,512 |
| 5 | `relasyon ak ayiti` | 23,065 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `referans kèk lyen nan` | 24,833 |
| 2 | `se yon vil nan` | 23,468 |
| 3 | `relasyon ant eta sa` | 23,057 |
| 4 | `ayisyen relasyon ant eta` | 23,056 |
| 5 | `eta sa epi ayiti` | 23,056 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ant eta sa epi ayiti` | 23,056 |
| 2 | `relasyon ant eta sa epi` | 23,056 |
| 3 | `ayisyen relasyon ant eta sa` | 23,056 |
| 4 | `kominote ayisyen relasyon ant eta` | 23,056 |
| 5 | `istwa istwa relasyon ak ayiti` | 23,047 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n _` | 1,566,894 |
| 2 | `a n` | 1,400,762 |
| 3 | `e _` | 1,352,775 |
| 4 | `_ a` | 826,760 |
| 5 | `o n` | 794,450 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a n _` | 604,651 |
| 2 | `o n _` | 457,174 |
| 3 | `_ : _` | 418,125 |
| 4 | `y o n` | 400,796 |
| 5 | `_ n a` | 387,040 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `n a n _` | 364,441 |
| 2 | `_ n a n` | 363,482 |
| 3 | `y o n _` | 363,435 |
| 4 | `s y o n` | 232,709 |
| 5 | `a s y o` | 182,198 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ n a n _` | 361,168 |
| 2 | `s y o n _` | 199,085 |
| 3 | `a s y o n` | 182,181 |
| 4 | `_ y o n _` | 136,613 |
| 5 | `y o n _ a` | 111,148 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 263
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~35% 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 | 1.0080 | 2.011 | 8.45 | 250,146 | 0.0% |
| **1** | Subword | 0.9515 | 1.934 | 6.83 | 2,043 | 4.8% |
| **2** | Word | 0.3036 | 1.234 | 1.83 | 2,109,748 | 69.6% |
| **2** | Subword | 0.8151 | 1.759 | 5.68 | 13,935 | 18.5% |
| **3** | Word | 0.1123 | 1.081 | 1.22 | 3,857,526 | 88.8% |
| **3** | Subword | 0.8180 | 1.763 | 4.75 | 79,027 | 18.2% |
| **4** | Word | 0.0480 🏆 | 1.034 | 1.08 | 4,709,643 | 95.2% |
| **4** | Subword | 0.7094 | 1.635 | 3.41 | 374,804 | 29.1% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `nan mikwofòn elektwomayetik yo te regrèt ke lidè otokratik jouk lè nou yòk new york li`
2. `de paul belmondo yon vil nan pwovens pinar del rio nan eta sa te rantre nan`
3. `li yo gide ak bibliyometrik premye woman ki ra kòm luis lazo aktivite li se pi`

**Context Size 2:**

1. `se yon endikatè ph tankou fenolftalein oswa bromotimol ble vignette tès pou lide a soti nan vèb`
2. `istwa istwa relasyon ak ayiti kominote ayisyen relasyon ant eta sa epi ayiti 6 fevrye gouvènman ayis...`
3. `referans kèk lyen nan georgie nan etazini li sitye nan leta ilinwa chèf lye li se bèzbòl`

**Context Size 3:**

1. `referans kèk lyen nan new york nan etazini se yon aktris ak chantèz fransèz orijin woumèn li te`
2. `se yon vil nan eta kawolin dinò nan etazini li te genyen 26 996 abitan nan rejyon windham`
3. `kèk lyen nan habana nan kiba gade tout gwo vil yo nan kat sa pèsonalite moun sa yo`

**Context Size 4:**

1. `referans kèk lyen nan kawolin dinò nan etazini istwa istwa relasyon ak ayiti kominote ayisyen relasy...`
2. `se yon vil nan pwovens santiago de cuba nan kiba gade tout gwo vil yo nan kat sa pèsonalite`
3. `relasyon ant eta sa epi ayiti 6 fevrye gouvènman ayisyen reprann kontak ak otorite kiben 2 chanselye...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_pasyoikshanimyo`
2. `an_sawen_d'épili`
3. `e_i_keri_kizonn_`

**Context Size 2:**

1. `n_rartikaskasyo._`
2. `ans_fevitillies_k`
3. `e_latî-mageonsema`

**Context Size 3:**

1. `an_antmanuel_marti`
2. `on_lan_redracebsta`
3. `_:_maritanizatè_pl`

**Context Size 4:**

1. `nan_li_se_yon_vil_p`
2. `_nan_wyominote_ayit`
3. `yon_ak_aktivite_kas`


### Key Findings

- **Best Predictability:** Context-4 (word) with 95.2% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (374,804 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 | 121,217 |
| Total Tokens | 8,389,833 |
| Mean Frequency | 69.21 |
| Median Frequency | 4 |
| Frequency Std Dev | 1815.48 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | nan | 363,480 |
| 2 | de | 183,168 |
| 3 | li | 156,632 |
| 4 | yo | 145,429 |
| 5 | yon | 137,456 |
| 6 | se | 132,316 |
| 7 | ak | 125,166 |
| 8 | sa | 123,752 |
| 9 | te | 99,980 |
| 10 | la | 84,358 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | taman | 2 |
| 2 | meotian | 2 |
| 3 | billikens | 2 |
| 4 | stb | 2 |
| 5 | oden | 2 |
| 6 | beno | 2 |
| 7 | olimpija | 2 |
| 8 | omri | 2 |
| 9 | duny | 2 |
| 10 | robiane | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.1103 |
| R² (Goodness of Fit) | 0.998571 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 47.2% |
| Top 1,000 | 71.5% |
| Top 5,000 | 84.3% |
| Top 10,000 | 89.1% |

### Key Findings

- **Zipf Compliance:** R²=0.9986 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 47.2% of corpus
- **Long Tail:** 111,217 words needed for remaining 10.9% 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 | 0.7588 | 0.3532 | N/A | N/A |
| **mono_64d** | 64 | 0.7534 | 0.2841 | N/A | N/A |
| **mono_128d** | 128 | 0.7522 | 0.2432 | N/A | N/A |
| **aligned_32d** | 32 | 0.7588 🏆 | 0.3565 | 0.0840 | 0.3820 |
| **aligned_64d** | 64 | 0.7534 | 0.2966 | 0.1500 | 0.5020 |
| **aligned_128d** | 128 | 0.7522 | 0.2468 | 0.2020 | 0.5860 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.7588 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2967. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 20.2% 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 | **1.087** | High formulaic/idiomatic 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 |
|--------|----------|
| `-a` | anios, answé, alexandrian |
| `-s` | slimane, scholl, serpico |
| `-ma` | marell, mariton, marivi |
| `-m` | marell, métrage, mariton |
| `-b` | belencita, basse, bretonneau |
| `-p` | puzzled, polisemi, pwoteyins |
| `-d` | dezyèm, divinite, delsham |
| `-c` | clayton, cuétara, cuarto |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-e` | naville, slimane, gigolette |
| `-s` | anios, pwoteyins, disques |
| `-n` | clayton, expression, kayman |
| `-a` | cuétara, belencita, preacha |
| `-on` | clayton, expression, dèfon |
| `-es` | disques, personnages, conneries |
| `-r` | haudecoeur, quarter, messemer |
| `-t` | joyadet, briat, fiat |

### 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 |
|------|----------|------------------|----------|
| `asyo` | 2.56x | 46 contexts | rasyo, rasyon, kasyon |
| `efer` | 2.56x | 29 contexts | refer, defer, jefery |
| `ogra` | 1.88x | 95 contexts | òtograf, ekograf, pwogram |
| `ikas` | 2.84x | 15 contexts | likasi, vikash, efikas |
| `lasy` | 2.82x | 15 contexts | glasyè, plasye, glasye |
| `omin` | 1.86x | 65 contexts | comin, komin, bomin |
| `rans` | 1.85x | 63 contexts | frans, trans, transe |
| `rela` | 2.10x | 34 contexts | relay, prela, irela |
| `liti` | 2.02x | 31 contexts | litik, litij, politi |
| `ayis` | 2.34x | 18 contexts | gayis, kayis, ayisye |
| `dika` | 2.18x | 21 contexts | odikap, fadika, endikap |
| `refe` | 2.30x | 17 contexts | refer, grefe, refere |

### 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 |
|--------|--------|-----------|----------|
| `-c` | `-e` | 91 words | charente, caderousse |
| `-c` | `-s` | 87 words | coquillages, colins |
| `-p` | `-e` | 78 words | paratonnerre, pwentiye |
| `-s` | `-s` | 72 words | sannois, sabines |
| `-p` | `-s` | 72 words | panis, phénomènes |
| `-s` | `-e` | 71 words | souffre, sœurette |
| `-a` | `-e` | 66 words | ampoule, affronte |
| `-d` | `-e` | 61 words | détente, detache |
| `-c` | `-n` | 59 words | comparaison, chambrun |
| `-b` | `-e` | 59 words | burlesque, banalite |

### 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 |
|------|-----------------|------------|------|
| champigny | **`champig-n-y`** | 7.5 | `n` |
| bronchinson | **`bronchin-s-on`** | 7.5 | `s` |
| illustreret | **`illustrer-e-t`** | 7.5 | `e` |
| paristonkar | **`paristonk-a-r`** | 7.5 | `a` |
| réalisateur | **`réalisat-e-ur`** | 7.5 | `e` |
| amoureuse | **`amoureu-s-e`** | 7.5 | `s` |
| glorieuses | **`glorieu-s-es`** | 7.5 | `s` |
| mauricette | **`maurice-t-te`** | 7.5 | `t` |
| manglehorn | **`mangleho-r-n`** | 7.5 | `r` |
| merchantville | **`merchantvi-l-le`** | 7.5 | `l` |
| smithville | **`smithvi-l-le`** | 7.5 | `l` |
| pedevilla | **`pe-de-villa`** | 7.5 | `villa` |
| potpourri | **`potpour-r-i`** | 7.5 | `r` |
| colasanti | **`co-la-santi`** | 7.5 | `santi` |
| ayikodans | **`ayikod-an-s`** | 7.5 | `an` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Haitian Creole shows high morphological productivity. The subword models are significantly more efficient than word models, suggesting a rich system of affixation or compounding.

> **Note on Idiomaticity:** The high Idiomaticity Gap suggests a large number of frequent multi-word expressions or formulaic sequences that are statistically distinct from their component parts.

---
## 7. Summary & Recommendations

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

### Production Recommendations

| Component | Recommended | Rationale |
|-----------|-------------|-----------|
| Tokenizer | **64k BPE** | Best compression (4.27x) |
| N-gram | **2-gram** | Lowest perplexity (263) |
| Markov | **Context-4** | Highest predictability (95.2%) |
| 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 03:29:01*