---
language: ary
language_name: Moroccan Arabic
language_family: arabic
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-arabic
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: 4.171
  - name: best_isotropy
    type: isotropy
    value: 0.8284
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-03
---

# Moroccan Arabic - Wikilangs Models
## Comprehensive Research Report & Full Ablation Study

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Moroccan Arabic** 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** | 3.480x | 3.48 | 0.0910% | 300,099 |
| **16k** | 3.753x | 3.76 | 0.0981% | 278,271 |
| **32k** | 3.983x | 3.99 | 0.1041% | 262,209 |
| **64k** | 4.171x 🏆 | 4.18 | 0.1090% | 250,397 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `هادي صفحة د التوضيح، كلمة بركان يمكن يكونو عندها هاد لمعاني: بْرْكان: مدينة مغري...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁هادي ▁صفحة ▁د ▁التوضيح ، ▁كلمة ▁بركان ▁يمكن ▁يكونو ▁عندها ... (+23 more)` | 33 |
| 16k | `▁هادي ▁صفحة ▁د ▁التوضيح ، ▁كلمة ▁بركان ▁يمكن ▁يكونو ▁عندها ... (+21 more)` | 31 |
| 32k | `▁هادي ▁صفحة ▁د ▁التوضيح ، ▁كلمة ▁بركان ▁يمكن ▁يكونو ▁عندها ... (+19 more)` | 29 |
| 64k | `▁هادي ▁صفحة ▁د ▁التوضيح ، ▁كلمة ▁بركان ▁يمكن ▁يكونو ▁عندها ... (+18 more)` | 28 |

**Sample 2:** `لْفزضاض ؤلا أفزضاض (سمية لعلمية Microcosmus sabatieri) حيوان لاسنسولي كيعيش ف لب...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁لْ ف ز ضاض ▁ؤلا ▁أف ز ضاض ▁( سمية ... (+31 more)` | 41 |
| 16k | `▁لْ ف ز ضاض ▁ؤلا ▁أف ز ضاض ▁( سمية ... (+28 more)` | 38 |
| 32k | `▁لْف ز ضاض ▁ؤلا ▁أف ز ضاض ▁( سمية ▁لعلمية ... (+25 more)` | 35 |
| 64k | `▁لْف زضاض ▁ؤلا ▁أف زضاض ▁( سمية ▁لعلمية ▁microcos mus ... (+17 more)` | 27 |

**Sample 3:** `نيلز أبراهام لانݣليت (مزيود ف 9 يوليوز - مات ف 30 مارس هوّا عالم د شّيمي سويدي. ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁نيل ز ▁أب راهام ▁ل انݣ ليت ▁( مزيود ▁ف ... (+19 more)` | 29 |
| 16k | `▁نيل ز ▁أبراهام ▁ل انݣ ليت ▁( مزيود ▁ف ▁ ... (+16 more)` | 26 |
| 32k | `▁نيلز ▁أبراهام ▁لانݣ ليت ▁( مزيود ▁ف ▁ 9 ▁يوليوز ... (+14 more)` | 24 |
| 64k | `▁نيلز ▁أبراهام ▁لانݣليت ▁( مزيود ▁ف ▁ 9 ▁يوليوز ▁- ... (+13 more)` | 23 |


### Key Findings

- **Best Compression:** 64k achieves 4.171x compression
- **Lowest UNK Rate:** 8k with 0.0910% 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 | 7,228 | 12.82 | 39,512 | 23.0% | 50.8% |
| **2-gram** | Subword | 424 🏆 | 8.73 | 5,903 | 58.0% | 96.4% |
| **3-gram** | Word | 5,655 | 12.47 | 43,555 | 27.5% | 57.1% |
| **3-gram** | Subword | 3,784 | 11.89 | 44,651 | 23.1% | 60.7% |
| **4-gram** | Word | 7,985 | 12.96 | 70,559 | 27.5% | 53.6% |
| **4-gram** | Subword | 20,064 | 14.29 | 220,807 | 12.0% | 36.0% |
| **5-gram** | Word | 7,565 | 12.89 | 58,964 | 28.5% | 52.9% |
| **5-gram** | Subword | 62,379 | 15.93 | 527,725 | 7.3% | 25.0% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `واصلة ل` | 8,540 |
| 2 | `نسبة د` | 7,170 |
| 3 | `ف لمغريب` | 6,305 |
| 4 | `ف إقليم` | 6,018 |
| 5 | `ف نسبة` | 4,265 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ف نسبة د` | 4,264 |
| 2 | `فيها مصدر و` | 3,236 |
| 3 | `و نسبة د` | 2,894 |
| 4 | `مصدر و بايت` | 2,856 |
| 5 | `اللي خدامين ف` | 2,760 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `فيها مصدر و بايت` | 2,856 |
| 2 | `نسبة نّاس اللي خدامين` | 2,705 |
| 3 | `نّاس اللي خدامين ف` | 2,594 |
| 4 | `على حساب لإحصاء الرسمي` | 2,501 |
| 5 | `حساب لإحصاء الرسمي د` | 2,500 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `نسبة نّاس اللي خدامين ف` | 2,593 |
| 2 | `ف لمغريب هاد دّوار كينتامي` | 2,500 |
| 3 | `هاد دّوار كينتامي ل مشيخة` | 2,500 |
| 4 | `لمغريب هاد دّوار كينتامي ل` | 2,500 |
| 5 | `حساب لإحصاء الرسمي د عام` | 2,500 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ا ل` | 347,466 |
| 2 | `_ ل` | 278,371 |
| 3 | `ة _` | 229,442 |
| 4 | `_ ا` | 220,960 |
| 5 | `_ م` | 156,801 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ ا ل` | 216,048 |
| 2 | `_ ف _` | 83,146 |
| 3 | `ا ت _` | 63,800 |
| 4 | `ي ة _` | 60,271 |
| 5 | `_ د _` | 59,563 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ د ي ا` | 47,798 |
| 2 | `د ي ا ل` | 47,559 |
| 3 | `ي ا ل _` | 33,039 |
| 4 | `د _ ا ل` | 32,831 |
| 5 | `_ م ن _` | 28,909 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ د ي ا ل` | 47,427 |
| 2 | `د ي ا ل _` | 32,608 |
| 3 | `_ ع ل ى _` | 19,473 |
| 4 | `_ ا ل ل ي` | 18,967 |
| 5 | `ا ل ل ي _` | 18,744 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 424
- **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 | 0.8561 | 1.810 | 5.38 | 178,865 | 14.4% |
| **1** | Subword | 1.1236 | 2.179 | 8.36 | 2,156 | 0.0% |
| **2** | Word | 0.2259 | 1.169 | 1.49 | 962,233 | 77.4% |
| **2** | Subword | 0.8160 | 1.761 | 5.10 | 18,029 | 18.4% |
| **3** | Word | 0.0618 | 1.044 | 1.10 | 1,431,084 | 93.8% |
| **3** | Subword | 0.8022 | 1.744 | 4.13 | 91,858 | 19.8% |
| **4** | Word | 0.0208 🏆 | 1.015 | 1.04 | 1,574,083 | 97.9% |
| **4** | Subword | 0.6604 | 1.581 | 2.86 | 379,445 | 34.0% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `ف لمغريب فيها 5 463 462 461 كم من غير ب شبه منقّر مكررعبد المسيح في`
2. `و أداب روسيا ف لمغريب ف وقت مابين اللغات الرسمية ديال حيزب لإستقلال تا سينيما ليها`
3. `د الناس فليبيا اكتشفو أنه يتقتل ولكن بقات كتلعب فالتيران ديال هاد الريحلة معا لمونتاخاب و`

**Context Size 2:**

1. `واصلة ل 98 6 و عدد لفاميلات تزاد ب 81 6 و نسبة د الناس و لمحيط`
2. `نسبة د الشوماج واصلة ل 21 12 نوطات مصادر ف لمغريب جّبل معروف عند الصامويين حتال ليوم`
3. `ف لمغريب هاد دّوار كينتامي ل مشيخة سدي حمد الدغوغي لي كتضم 14 د دّواور لعاداد د`

**Context Size 3:**

1. `ف نسبة د التسكويل واصلة ل 91 89 و نسبة د الشوماج واصلة ل 7 6 و لخصوبة`
2. `فيها مصدر و بايت زادهوم داريجابوت حيين مغاربا د لقرن 21 مغاربا مغاربا فيها مصدر و بايت زادهوم`
3. `و نسبة د لأمية واصلة ل 53 4 و نسبة د لأمية واصلة ل 92 5 و نسبة`

**Context Size 4:**

1. `نسبة نّاس اللي خدامين ف دّولة ولا لبيطاليين اللي سبق ليهوم خدمو 44 3 نسبة نّاس اللي خدامين ف`
2. `نّاس اللي خدامين ف لپريڤي ولا لبيطاليين اللي سبق ليهوم مصادر الدار البيضاء سطات قروية ف إقليم سطات ق...`
3. `على حساب لإحصاء الرسمي د عام إحصائيات إحصائيات عامة عدد السكان ديال أورسفان نقص ب 30 7 و عدد`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_دّرى_لجالب_لتالع`
2. `اكترن_لعاميلة_ن_`
3. `لت_پرومدي_و_ماتم`

**Context Size 2:**

1. `الرجل_بين_ماعة_لخ`
2. `_لكينو_العرفوقعوه`
3. `ة_27_نت،_خري_د_لج`

**Context Size 3:**

1. `_الروس_و_هي_ماية_ك`
2. `_ف_موقريب._الدفايي`
3. `ات_ف_البالشخصياتول`

**Context Size 4:**

1. `_ديالو._ميامينش_و_ت`
2. `ديال_أسباب_الغرب_6_`
3. `يال_تعرّض_للحزب_الوه`


### Key Findings

- **Best Predictability:** Context-4 (word) with 97.9% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (379,445 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 | 78,779 |
| Total Tokens | 2,032,841 |
| Mean Frequency | 25.80 |
| Median Frequency | 4 |
| Frequency Std Dev | 515.92 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ف | 83,458 |
| 2 | و | 59,829 |
| 3 | د | 59,731 |
| 4 | ديال | 32,565 |
| 5 | من | 29,236 |
| 6 | ل | 23,572 |
| 7 | على | 19,570 |
| 8 | لي | 18,402 |
| 9 | اللي | 17,442 |
| 10 | ب | 17,233 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | بوفوار | 2 |
| 2 | بيتسي | 2 |
| 3 | وصانعي | 2 |
| 4 | وأهميتها | 2 |
| 5 | بورديو | 2 |
| 6 | بلومر | 2 |
| 7 | مقترحة | 2 |
| 8 | anchor | 2 |
| 9 | بعصبة | 2 |
| 10 | ماڭي | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 1.0213 |
| R² (Goodness of Fit) | 0.998918 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 38.6% |
| Top 1,000 | 62.9% |
| Top 5,000 | 77.8% |
| Top 10,000 | 84.2% |

### Key Findings

- **Zipf Compliance:** R²=0.9989 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 38.6% of corpus
- **Long Tail:** 68,779 words needed for remaining 15.8% 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 | 0.8284 🏆 | 0.3330 | N/A | N/A |
| **mono_64d** | 64 | 0.8181 | 0.2588 | N/A | N/A |
| **mono_128d** | 128 | 0.7036 | 0.2093 | N/A | N/A |
| **aligned_32d** | 32 | 0.8284 | 0.3345 | 0.0180 | 0.1360 |
| **aligned_64d** | 64 | 0.8181 | 0.2550 | 0.0380 | 0.1760 |
| **aligned_128d** | 128 | 0.7036 | 0.2072 | 0.0620 | 0.2760 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.8284 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2663. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 6.2% 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 | **1.114** | 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 |
|--------|----------|
| `-ال` | الأمني, اللحظة, الفيرمات |
| `-لم` | لمتعصبين, لمحافض, لمونضامة |
| `-كا` | كاتدير, كايتحلو, كايقممو |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-ة` | سميّة, رقصة, اللحظة |
| `-ات` | سطراتيجيات, الفيرمات, لحتيفالات |
| `-ية` | الشرقية, اللاجنسية, ولوسطانية |
| `-ين` | لمتعصبين, ثنين, لمالحين |

### 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 |
|------|----------|------------------|----------|
| `انية` | 1.80x | 68 contexts | غانية, ثانية, سانية |
| `اللو` | 1.74x | 61 contexts | اللوه, اللور, اللول |
| `الات` | 1.71x | 65 contexts | تالات, حالات, صالات |
| `جماع` | 1.90x | 38 contexts | جماعي, تجماع, إجماع |
| `النا` | 1.63x | 63 contexts | الناي, النار, الناس |
| `لمغر` | 1.92x | 30 contexts | لمغرب, لمغربب, للمغرب |
| `إحصا` | 2.13x | 17 contexts | إحصاء, لإحصا, إحصائي |
| `مغري` | 2.08x | 18 contexts | مغريب, مغرية, مغريبي |
| `حصاء` | 2.24x | 14 contexts | إحصاء, لإحصاء, ليحصاء |
| `دهوم` | 2.14x | 16 contexts | ضدهوم, يردهوم, زادهوم |
| `قليم` | 2.06x | 17 contexts | فقليم, اقليم, إقليم |
| `لجوا` | 1.77x | 26 contexts | لجواب, لجواد, الجوا |

### 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 |
|--------|--------|-----------|----------|
| `-ال` | `-ة` | 280 words | الراكوبة, العمدة |
| `-ال` | `-ات` | 163 words | الشلالات, العبرات |
| `-ال` | `-ية` | 152 words | الزراعية, الطباشيرية |
| `-ال` | `-ين` | 76 words | الموحدين, الاثنين |
| `-لم` | `-ة` | 66 words | لمملكة, لمُحمدية |
| `-لم` | `-ين` | 45 words | لموناضيلين, لمعتقلين |
| `-لم` | `-ات` | 25 words | لمونضّامات, لممرات |
| `-لم` | `-ية` | 21 words | لمُحمدية, لمراكشية |
| `-كا` | `-ين` | 2 words | كايسين, كاتبين |

### 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 |
|------|-----------------|------------|------|
| التوجيهات | **`ال-توجيه-ات`** | 6.0 | `توجيه` |
| الصومالية | **`ال-صومال-ية`** | 6.0 | `صومال` |
| الپاكستانية | **`ال-پاكستان-ية`** | 6.0 | `پاكستان` |
| الدوّازات | **`ال-دوّاز-ات`** | 6.0 | `دوّاز` |
| الصالونات | **`ال-صالون-ات`** | 6.0 | `صالون` |
| التعبيرية | **`ال-تعبير-ية`** | 6.0 | `تعبير` |
| الانقلابية | **`ال-انقلاب-ية`** | 6.0 | `انقلاب` |
| لمنقارضين | **`لم-نقارض-ين`** | 6.0 | `نقارض` |
| التقليديين | **`ال-تقليدي-ين`** | 6.0 | `تقليدي` |
| لمنتاشرين | **`لم-نتاشر-ين`** | 6.0 | `نتاشر` |
| الماكينات | **`ال-ماكين-ات`** | 6.0 | `ماكين` |
| البرونزية | **`ال-برونز-ية`** | 6.0 | `برونز` |
| التكوينية | **`ال-تكوين-ية`** | 6.0 | `تكوين` |
| التعليمية | **`ال-تعليم-ية`** | 6.0 | `تعليم` |
| التلفزيونية | **`ال-تلفزيون-ية`** | 6.0 | `تلفزيون` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Moroccan Arabic 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 (4.17x) |
| N-gram | **2-gram** | Lowest perplexity (424) |
| Markov | **Context-4** | Highest predictability (97.9%) |
| 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 16:42:17*