README.md

---
language: an
language_name: Aragonese
language_family: romance_iberian
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_iberian
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: 4.275
  - name: best_isotropy
    type: isotropy
    value: 0.8232
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-03
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Aragonese** 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** | 3.559x | 3.56 | 0.1247% | 1,207,427 |
| **16k** | 3.854x | 3.85 | 0.1351% | 1,114,964 |
| **32k** | 4.092x | 4.09 | 0.1434% | 1,050,138 |
| **64k** | 4.275x 🏆 | 4.28 | 0.1498% | 1,005,070 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `Bobadilla puet estar: Bobadilla, un municipio de La Rioja. Bobadilla del Campo, ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁bob ad illa ▁puet ▁estar : ▁bob ad illa , ... (+17 more)` | 27 |
| 16k | `▁bob ad illa ▁puet ▁estar : ▁bob ad illa , ... (+17 more)` | 27 |
| 32k | `▁bob ad illa ▁puet ▁estar : ▁bob ad illa , ... (+17 more)` | 27 |
| 64k | `▁bobadilla ▁puet ▁estar : ▁bobadilla , ▁un ▁municipio ▁de ▁la ... (+11 more)` | 21 |

**Sample 2:** `Charleville-Mézières ye una localidat y comuna francesa, capital d'o departament...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁char le ville - m é zi ères ▁ye ▁una ... (+26 more)` | 36 |
| 16k | `▁char le ville - mé zi ères ▁ye ▁una ▁localidat ... (+25 more)` | 35 |
| 32k | `▁char le ville - mé zi ères ▁ye ▁una ▁localidat ... (+23 more)` | 33 |
| 64k | `▁charleville - mézières ▁ye ▁una ▁localidat ▁y ▁comuna ▁francesa , ... (+19 more)` | 29 |

**Sample 3:** `Schöngeising (en bavaro Scheegeising) ye un municipio de Bavera, Alemanya. Se tr...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁sch ön ge is ing ▁( en ▁bavaro ▁s che ... (+29 more)` | 39 |
| 16k | `▁schön ge is ing ▁( en ▁bavaro ▁sche e ge ... (+25 more)` | 35 |
| 32k | `▁schön ge ising ▁( en ▁bavaro ▁sche e ge ising ... (+20 more)` | 30 |
| 64k | `▁schön ge ising ▁( en ▁bavaro ▁sche e ge ising ... (+20 more)` | 30 |


### Key Findings

- **Best Compression:** 64k achieves 4.275x compression
- **Lowest UNK Rate:** 8k with 0.1247% 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 | 25,712 | 14.65 | 233,669 | 16.7% | 37.4% |
| **2-gram** | Subword | 257 🏆 | 8.01 | 7,000 | 68.7% | 99.3% |
| **3-gram** | Word | 87,357 | 16.41 | 461,562 | 8.3% | 23.0% |
| **3-gram** | Subword | 2,151 | 11.07 | 52,727 | 25.8% | 73.4% |
| **4-gram** | Word | 209,676 | 17.68 | 900,576 | 6.8% | 17.2% |
| **4-gram** | Subword | 12,170 | 13.57 | 289,768 | 12.6% | 39.7% |
| **5-gram** | Word | 208,007 | 17.67 | 773,213 | 6.3% | 16.4% |
| **5-gram** | Subword | 46,669 | 15.51 | 901,225 | 7.3% | 25.5% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `d a` | 107,208 |
| 2 | `d o` | 106,261 |
| 3 | `en a` | 60,798 |
| 4 | `en o` | 45,519 |
| 5 | `de l` | 37,458 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a provincia de` | 17,480 |
| 2 | `d a provincia` | 13,447 |
| 3 | `una superficie de` | 12,736 |
| 4 | `suya población ye` | 12,405 |
| 5 | `en una superficie` | 12,352 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `suya población ye de` | 12,284 |
| 2 | `en una superficie de` | 12,148 |
| 3 | `d a provincia de` | 12,141 |
| 4 | `habitants en una superficie` | 11,275 |
| 5 | `a suya población ye` | 11,250 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a suya población ye de` | 11,136 |
| 2 | `habitants en una superficie de` | 11,095 |
| 3 | `una densidat de población de` | 10,633 |
| 4 | `km con una densidat de` | 7,736 |
| 5 | `con una densidat de población` | 7,674 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a _` | 1,873,392 |
| 2 | `_ d` | 1,605,638 |
| 3 | `e _` | 1,544,207 |
| 4 | `s _` | 1,309,585 |
| 5 | `n _` | 1,215,896 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ d e` | 891,253 |
| 2 | `d e _` | 772,067 |
| 3 | `_ d '` | 491,537 |
| 4 | `e n _` | 478,088 |
| 5 | `_ e n` | 454,282 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ d e _` | 737,370 |
| 2 | `_ e n _` | 397,348 |
| 3 | `_ d ' a` | 234,868 |
| 4 | `a _ d e` | 184,900 |
| 5 | `_ c o n` | 179,093 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `a _ d e _` | 147,074 |
| 2 | `_ q u e _` | 125,472 |
| 3 | `c i ó n _` | 124,436 |
| 4 | `o _ d e _` | 123,146 |
| 5 | `_ d ' a _` | 106,742 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 257
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~25% 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 | 0.9753 | 1.966 | 7.59 | 368,549 | 2.5% |
| **1** | Subword | 0.7834 | 1.721 | 5.75 | 3,672 | 21.7% |
| **2** | Word | 0.3415 | 1.267 | 2.01 | 2,791,626 | 65.9% |
| **2** | Subword | 0.8176 | 1.763 | 5.23 | 21,123 | 18.2% |
| **3** | Word | 0.1548 | 1.113 | 1.33 | 5,610,004 | 84.5% |
| **3** | Subword | 0.7695 | 1.705 | 4.30 | 110,486 | 23.0% |
| **4** | Word | 0.0739 🏆 | 1.053 | 1.14 | 7,469,366 | 92.6% |
| **4** | Subword | 0.7129 | 1.639 | 3.37 | 474,961 | 28.7% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `de las neveras y rfef aprebó a pachina web oficial d afers son asociadas con os`
2. `d elba antiparte la provincia d as que se veiga torda collerada rafel vidaller tricas libro`
3. `a rendición de sattler torna ta partecipar en ifriquiya y cariño homenage vasallage en aragonés vinc...`

**Context Size 2:**

1. `d a ciudat de zaragoza tomo i de castiella y leyón espanya o escritor de lausbubengeschichte ye`
2. `d o reino se consolida la influyencia de l exercito estatounitesne en europa s extiende dende os`
3. `en a provincia de teruel d o cual en fan parte 4 cantons y 129 comunas lista`

**Context Size 3:**

1. `a provincia de zaragoza en a provincia de concepción y d as tres serols estando dimpués enamplato a`
2. `d a provincia de guipuzcua ta atros usos se veiga carlos ix carlos ix 27 de chunio de`
3. `una superficie de 158 60 km y una densidat de población de 346 35 hab km a suya`

**Context Size 4:**

1. `suya población ye de 81 habitants en una superficie de 194 49 km con una densidat de población de`
2. `en una superficie de 64 16 km con una densidat de población de 43 44 hab km demografía administració...`
3. `d a provincia de burgos ta atros usos se veiga fort yuma desambigación fort yuma títol orichinal en ...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_un_der_dent_ckm`
2. `as_en_as_2_tacla`
3. `en_lern_don_vitr`

**Context Size 2:**

1. `a_saus_dabinascer`
2. `_derfica_sublosti`
3. `e_manaisitau_suyo`

**Context Size 3:**

1. `_dens._val_novant,`
2. `de_319_de_fuel,_qu`
3. `_d'o_primetada_cic`

**Context Size 4:**

1. `_de_jean-jose_(naix`
2. `_en_sido_per_bueno,`
3. `_d'anglés_jean_sabi`


### Key Findings

- **Best Predictability:** Context-4 (word) with 92.6% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (474,961 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 | 183,928 |
| Total Tokens | 11,661,736 |
| Mean Frequency | 63.40 |
| Median Frequency | 4 |
| Frequency Std Dev | 2823.00 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | de | 741,521 |
| 2 | d | 497,145 |
| 3 | a | 440,622 |
| 4 | en | 410,893 |
| 5 | o | 301,627 |
| 6 | y | 247,568 |
| 7 | que | 127,976 |
| 8 | l | 109,848 |
| 9 | ye | 109,774 |
| 10 | una | 105,502 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | beljakova | 2 |
| 2 | méchaly | 2 |
| 3 | wiedemann | 2 |
| 4 | limotte | 2 |
| 5 | wlodkowski | 2 |
| 6 | taos | 2 |
| 7 | slovis | 2 |
| 8 | samaha | 2 |
| 9 | seros | 2 |
| 10 | cookeville | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0690 |
| R² (Goodness of Fit) | 0.998251 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 44.8% |
| Top 1,000 | 66.8% |
| Top 5,000 | 80.7% |
| Top 10,000 | 85.9% |

### Key Findings

- **Zipf Compliance:** R²=0.9983 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 44.8% of corpus
- **Long Tail:** 173,928 words needed for remaining 14.1% 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 | 0.8168 | 0.3517 | N/A | N/A |
| **mono_64d** | 64 | 0.8232 🏆 | 0.2779 | N/A | N/A |
| **mono_128d** | 128 | 0.8044 | 0.2016 | N/A | N/A |
| **aligned_32d** | 32 | 0.8168 | 0.3524 | 0.1520 | 0.4840 |
| **aligned_64d** | 64 | 0.8232 | 0.2773 | 0.2480 | 0.6340 |
| **aligned_128d** | 128 | 0.8044 | 0.2034 | 0.3740 | 0.7380 |

### Key Findings

- **Best Isotropy:** mono_64d with 0.8232 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2774. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 37.4% 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 | **-0.358** | 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 |
|--------|----------|
| `-co` | confrontatos, conchecturau, coluche |
| `-ca` | casartelli, camprodón, canthus |
| `-re` | reitzenstein, reformata, reinando |
| `-de` | destruyir, denasalizadas, debucourt |
| `-ma` | marktes, matosinhos, marciac |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-s` | mourvilles, iliricas, mylonas |
| `-a` | cingüenda, lecinyena, reformata |
| `-as` | iliricas, mylonas, aeneas |
| `-os` | confrontatos, agnatos, estranios |
| `-es` | mourvilles, marktes, forbes |

### 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 |
|------|----------|------------------|----------|
| `ient` | 1.70x | 176 contexts | cient, oient, dient |
| `ento` | 1.74x | 126 contexts | sento, bento, cento |
| `rago` | 2.03x | 58 contexts | arago, trago, ragot |
| `ranc` | 1.64x | 141 contexts | franc, rance, ranca |
| `ació` | 2.09x | 47 contexts | nació, ación, fació |
| `enci` | 1.53x | 164 contexts | encia, renci, oencia |
| `obla` | 1.90x | 56 contexts | robla, pobla, nobla |
| `nter` | 1.50x | 146 contexts | anter, enter, inter |
| `ncia` | 1.72x | 61 contexts | encia, uncia, oencia |
| `cion` | 1.50x | 110 contexts | scion, nacion, accion |
| `idat` | 2.00x | 28 contexts | unidat, deidat, humidat |
| `mbre` | 1.55x | 75 contexts | ambre, ombre, umbre |

### 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 |
|--------|--------|-----------|----------|
| `-co` | `-s` | 71 words | concilios, comenges |
| `-ca` | `-s` | 53 words | cabrinos, caracteres |
| `-ca` | `-a` | 49 words | cafeína, caixera |
| `-co` | `-a` | 49 words | cosida, conquiolina |
| `-ma` | `-s` | 41 words | mauriscus, mandos |
| `-ma` | `-a` | 36 words | mainila, mamma |
| `-re` | `-s` | 34 words | reprimius, rechiradors |
| `-re` | `-a` | 33 words | relochería, renacentista |
| `-de` | `-a` | 30 words | desidia, dentada |
| `-de` | `-s` | 30 words | demograficos, deverbativos |

### 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 |
|------|-----------------|------------|------|
| repoblatos | **`re-poblat-os`** | 6.0 | `poblat` |
| altoaragonesas | **`altoaragon-es-as`** | 6.0 | `altoaragon` |
| recullindo | **`re-cullindo`** | 4.5 | `cullindo` |
| reorganizar | **`re-organizar`** | 4.5 | `organizar` |
| romanticos | **`romantic-os`** | 4.5 | `romantic` |
| casellato | **`ca-sellato`** | 4.5 | `sellato` |
| discapacitatos | **`discapacitat-os`** | 4.5 | `discapacitat` |
| lexicales | **`lexical-es`** | 4.5 | `lexical` |
| monetarias | **`monetari-as`** | 4.5 | `monetari` |
| reprodución | **`re-produción`** | 4.5 | `produción` |
| deportaban | **`de-portaban`** | 4.5 | `portaban` |
| desconoixitas | **`de-sconoixit-as`** | 3.0 | `sconoixit` |
| caspolinas | **`ca-spolin-as`** | 3.0 | `spolin` |
| conservaderas | **`co-nservader-as`** | 3.0 | `nservader` |
| decimetros | **`de-cimetr-os`** | 3.0 | `cimetr` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Aragonese 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 (4.28x) |
| N-gram | **2-gram** | Lowest perplexity (257) |
| Markov | **Context-4** | Highest predictability (92.6%) |
| 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: 2026-01-03 17:05:39*