README.md

---
language: guc
language_name: Wayuu
language_family: american_arawak
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-american_arawak
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: 5.025
  - name: best_isotropy
    type: isotropy
    value: 0.4101
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Wayuu** 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** | 3.998x | 4.00 | 0.1729% | 272,950 |
| **16k** | 4.380x | 4.38 | 0.1895% | 249,121 |
| **32k** | 4.743x | 4.75 | 0.2052% | 230,070 |
| **64k** | 5.025x 🏆 | 5.03 | 0.2173% | 217,163 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Alhuliya Sawara Arabu Demokratika (RASD), nisqaqa Aphrikapi huk mama llaqtam, ll...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁al hu li ya ▁sawa ra ▁ara bu ▁de mo ... (+26 more)` | 36 |
| 16k | `▁al hu liya ▁sawa ra ▁ara bu ▁de mo k ... (+15 more)` | 25 |
| 32k | `▁alhuliya ▁sawara ▁ara bu ▁demokratika ▁( rasd ), ▁nisqaqa ▁aphrikapi ... (+7 more)` | 17 |
| 64k | `▁alhuliya ▁sawara ▁arabu ▁demokratika ▁( rasd ), ▁nisqaqa ▁aphrikapi ▁huk ... (+6 more)` | 16 |

**Sample 2:** `Tü Mma'ipakat Toliima (Alijunaiki: Departamento Tolima) jiia wanee mma'ipakat sa...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁tü ▁mma ' ipakat ▁to li ima ▁( alijunaiki : ... (+25 more)` | 35 |
| 16k | `▁tü ▁mma ' ipakat ▁toliima ▁( alijunaiki : ▁departamento ▁to ... (+23 more)` | 33 |
| 32k | `▁tü ▁mma ' ipakat ▁toliima ▁( alijunaiki : ▁departamento ▁tolima ... (+22 more)` | 32 |
| 64k | `▁tü ▁mma ' ipakat ▁toliima ▁( alijunaiki : ▁departamento ▁tolima ... (+22 more)` | 32 |

**Sample 3:** `Tarüjeeta ajuyaajia (alijunaiki: tarjeta de crédito)`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ta rü jee ta ▁aju yaa jia ▁( alijunaiki : ... (+11 more)` | 21 |
| 16k | `▁ta rüjee ta ▁ajuyaa jia ▁( alijunaiki : ▁ta r ... (+7 more)` | 17 |
| 32k | `▁tarüjee ta ▁ajuyaajia ▁( alijunaiki : ▁tarjeta ▁de ▁crédito )` | 10 |
| 64k | `▁tarüjeeta ▁ajuyaajia ▁( alijunaiki : ▁tarjeta ▁de ▁crédito )` | 9 |


### Key Findings

- **Best Compression:** 64k achieves 5.025x compression
- **Lowest UNK Rate:** 8k with 0.1729% 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 | 1,485 | 10.54 | 2,732 | 29.7% | 71.2% |
| **2-gram** | Subword | 206 🏆 | 7.68 | 1,229 | 73.2% | 99.9% |
| **3-gram** | Word | 1,502 | 10.55 | 2,148 | 24.9% | 69.9% |
| **3-gram** | Subword | 1,420 | 10.47 | 7,905 | 32.1% | 80.6% |
| **4-gram** | Word | 2,089 | 11.03 | 2,754 | 19.8% | 54.1% |
| **4-gram** | Subword | 6,489 | 12.66 | 32,863 | 16.4% | 49.5% |
| **5-gram** | Word | 965 | 9.91 | 1,313 | 29.3% | 83.0% |
| **5-gram** | Subword | 18,510 | 14.18 | 70,413 | 10.3% | 32.4% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `sulu u` | 663 |
| 2 | `de la` | 582 |
| 3 | `u tü` | 336 |
| 4 | `otta müsia` | 255 |
| 5 | `sukua ipa` | 255 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `sulu u tü` | 189 |
| 2 | `shi ipajee sukua` | 89 |
| 3 | `ipajee sukua ipa` | 88 |
| 4 | `no u chi` | 88 |
| 5 | `tü mma ipakat` | 70 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `shi ipajee sukua ipa` | 88 |
| 2 | `wanee mma ipakat saakaje` | 63 |
| 3 | `jiia wanee mma ipakat` | 61 |
| 4 | `shi ipajee sukuwa ipa` | 53 |
| 5 | `no u chi juyakai` | 44 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `jiia wanee mma ipakat saakaje` | 61 |
| 2 | `apünüin shiiki sumaa piama mma` | 32 |
| 3 | `shiiki sumaa piama mma ipakat` | 32 |
| 4 | `32 apünüin shiiki sumaa piama` | 32 |
| 5 | `wanee mma ipakat saakaje 32` | 31 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a _` | 43,575 |
| 2 | `_ s` | 25,414 |
| 3 | `k a` | 22,304 |
| 4 | `i n` | 19,881 |
| 5 | `n a` | 18,513 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `i n _` | 11,521 |
| 2 | `a k a` | 8,886 |
| 3 | `_ w a` | 8,759 |
| 4 | `a _ s` | 8,142 |
| 5 | `_ s ü` | 8,067 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ t ü _` | 6,482 |
| 2 | `a i n _` | 4,294 |
| 3 | `ü i n _` | 3,910 |
| 4 | `a y u u` | 3,806 |
| 5 | `_ s h i` | 3,758 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ w a y u` | 3,416 |
| 2 | `w a y u u` | 3,413 |
| 3 | `_ w a n e` | 2,552 |
| 4 | `n a i n _` | 2,226 |
| 5 | `a _ t ü _` | 2,171 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 206
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~32% 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 | 0.5726 | 1.487 | 3.22 | 34,317 | 42.7% |
| **1** | Subword | 1.3184 | 2.494 | 11.14 | 212 | 0.0% |
| **2** | Word | 0.1658 | 1.122 | 1.33 | 110,087 | 83.4% |
| **2** | Subword | 1.1483 | 2.217 | 6.13 | 2,362 | 0.0% |
| **3** | Word | 0.0458 | 1.032 | 1.06 | 145,663 | 95.4% |
| **3** | Subword | 0.8483 | 1.800 | 3.79 | 14,457 | 15.2% |
| **4** | Word | 0.0123 🏆 | 1.009 | 1.01 | 154,297 | 98.8% |
| **4** | Subword | 0.6394 | 1.558 | 2.55 | 54,759 | 36.1% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `tü niipüsekat ramón paz neima cacerio los ciudadanos no tener cola de la península guajira atunkulee`
2. `de extinción iniciativa de la primera sesión de aguas marinas pueden no u wakuwa ipa shia`
3. `wayuu laaulakai`

**Context Size 2:**

1. `sulu u wanee mma yawaasirü sünülia apütaa müsia tü eekalü süpa apünaa mmakalü wajiira soo opünaa nak...`
2. `de la madre ya que solamente hay garantía de venta en el año cuando se unifican todas`
3. `u tü wanuikikalü sünain waikale erüin tü nikirajaaka anain naja laje erüin tü ta yataainka jüpüla to...`

**Context Size 3:**

1. `sulu u tü mmakat maliro ulia cha wopumuin chawaishii juyapo ulu otta jouktaleolu u chawaishi kepian ...`
2. `shi ipajee sukua ipa wayuuirua`
3. `no u chi juyakai akumajüi jee aashajaai saa u akua ipa tü oushiikat aikalaasü süchoin namaa süikeeyu...`

**Context Size 4:**

1. `wanee mma ipakat saakaje 24 piama shiiki sumaa pienchi mma ipakat yaa ekuwatoorü`
2. `jiia wanee mma ipakat saakaje 23 piama shiiki sumaa apünüin mma ipakat yaa wenesueela wenesueela`
3. `shi ipajee sukua ipa otta mürülü`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `a_see_ive'ictü_,`
2. `_ma'waijolulasü_`
3. `ine_pümona_ayanü`

**Context Size 2:**

1. `a_naiwayuunain_lo`
2. `_su_ya_jute_recto`
3. `ka_hiin_o'u_juu_y`

**Context Size 3:**

1. `in_toolomwia;_y_co`
2. `akaa_pies._eekalü_`
3. `_wajirapüla_lin_sü`

**Context Size 4:**

1. `_tü_pütchirain_müts`
2. `ain_naya_sünain_nup`
3. `üin_shimuunain_jalo`


### Key Findings

- **Best Predictability:** Context-4 (word) with 98.8% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (54,759 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 | 11,490 |
| Total Tokens | 139,617 |
| Mean Frequency | 12.15 |
| Median Frequency | 3 |
| Frequency Std Dev | 94.29 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | tü | 6,686 |
| 2 | de | 2,914 |
| 3 | wayuu | 2,549 |
| 4 | u | 2,074 |
| 5 | a | 2,048 |
| 6 | la | 1,823 |
| 7 | otta | 1,430 |
| 8 | süpüla | 1,331 |
| 9 | shia | 1,323 |
| 10 | sünain | 1,278 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | jintu | 2 |
| 2 | sünee | 2 |
| 3 | eekiraja | 2 |
| 4 | nushi | 2 |
| 5 | aajat | 2 |
| 6 | outon | 2 |
| 7 | joloo | 2 |
| 8 | lunes | 2 |
| 9 | nien | 2 |
| 10 | rimikukai | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 0.9694 |
| R² (Goodness of Fit) | 0.991868 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 44.8% |
| Top 1,000 | 70.2% |
| Top 5,000 | 89.4% |
| Top 10,000 | 97.9% |

### Key Findings

- **Zipf Compliance:** R²=0.9919 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 44.8% of corpus
- **Long Tail:** 1,490 words needed for remaining 2.1% 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 | 0.4101 🏆 | 0.4040 | N/A | N/A |
| **mono_64d** | 64 | 0.0872 | 0.4056 | N/A | N/A |
| **mono_128d** | 128 | 0.0142 | 0.4419 | N/A | N/A |
| **aligned_32d** | 32 | 0.4101 | 0.4057 | 0.0200 | 0.1260 |
| **aligned_64d** | 64 | 0.0872 | 0.4093 | 0.0240 | 0.1680 |
| **aligned_128d** | 128 | 0.0142 | 0.4428 | 0.0340 | 0.1740 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.4101 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.4182. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 3.4% 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 | **1.563** | 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 |
|--------|----------|
| `-su` | supone, sukuiappa, supüshuwa |
| `-wa` | waraittaa, wapüshi, wayü |
| `-sü` | sümioushe, süsika, sümülüin |
| `-ka` | kariñas, kanuliakat, kashiiwai |
| `-an` | antüna, ancestros, anainjanit |
| `-ma` | mariia, malu, maicao |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-a` | mariia, universitaria, waraittaa |
| `-n` | nashatüin, neraajüin, llegaron |
| `-in` | nashatüin, neraajüin, epijain |
| `-ka` | erajunaka, süsika, isashiika |
| `-aa` | waraittaa, naa, atunkaa |
| `-aka` | erajunaka, ipaka, ataka |
| `-üin` | nashatüin, neraajüin, sümülüin |
| `-sü` | akanajasü, kojutsü, akumujunüsü |

### 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 |
|------|----------|------------------|----------|
| `ajaa` | 1.63x | 48 contexts | rajaa, kajaa, ajaaya |
| `chik` | 1.40x | 59 contexts | achiku, achiki, süchiku |
| `ainj` | 1.47x | 47 contexts | ainja, aainja, ainjaa |
| `inja` | 1.68x | 27 contexts | ainja, aainja, ainjaa |
| `tuma` | 1.64x | 23 contexts | atuma, tumas, watuma |
| `akal` | 1.48x | 30 contexts | akalü, jakala, makalü |
| `uuka` | 1.67x | 18 contexts | suuka, ayuuka, jouukai |
| `kuwa` | 1.60x | 20 contexts | kuwai, akuwa, nükuwa |
| `amüi` | 1.65x | 16 contexts | amüin, tamüin, wamüin |
| `anee` | 1.48x | 22 contexts | wanee, aneerü, taanee |
| `shik` | 1.35x | 26 contexts | shiki, shika, shikat |
| `hika` | 1.32x | 26 contexts | shika, shikat, shikaa |

### 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 |
|--------|--------|-----------|----------|
| `-wa` | `-a` | 107 words | wapüshua, wayuukaluirua |
| `-su` | `-a` | 105 words | suluupuna, suichikana |
| `-sü` | `-a` | 96 words | süpulaa, sümüsainka |
| `-ka` | `-a` | 80 words | kaaraipia, kama |
| `-ma` | `-a` | 57 words | maalia, maracaaya |
| `-sü` | `-n` | 56 words | sümuin, sülamain |
| `-sü` | `-in` | 55 words | sümuin, sülamain |
| `-an` | `-a` | 47 words | anapajirawaa, anulia |
| `-su` | `-n` | 45 words | sulapuin, sukumajünuin |
| `-wa` | `-n` | 41 words | wain, waneeyan |

### 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 |
|------|-----------------|------------|------|
| suwanajain | **`su-wa-naja-in`** | 7.5 | `naja` |
| anashanain | **`an-asha-na-in`** | 7.5 | `asha` |
| suichikana | **`su-ichi-ka-na`** | 7.5 | `ichi` |
| pasanainsü | **`pasa-na-in-sü`** | 7.5 | `pasa` |
| laülayuukana | **`laülayuu-ka-na`** | 6.0 | `laülayuu` |
| layuukana | **`layuu-ka-na`** | 6.0 | `layuu` |
| kachirasü | **`ka-chira-sü`** | 6.0 | `chira` |
| ewaliikana | **`ewalii-ka-na`** | 6.0 | `ewalii` |
| nüpülainka | **`nüpüla-in-ka`** | 6.0 | `nüpüla` |
| upayuukana | **`upayuu-ka-na`** | 6.0 | `upayuu` |
| tepichikana | **`tepichi-ka-na`** | 6.0 | `tepichi` |
| sugolfoin | **`su-golfo-in`** | 6.0 | `golfo` |
| sumainwaa | **`su-ma-inwaa`** | 6.0 | `inwaa` |
| wachukumuinkana | **`wa-chukumu-in-ka-na`** | 6.0 | `chukumu` |
| kamanakat | **`ka-ma-nakat`** | 6.0 | `nakat` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Wayuu 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 (5.02x) |
| N-gram | **2-gram** | Lowest perplexity (206) |
| Markov | **Context-4** | Highest predictability (98.8%) |
| 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 00:33:28*