---
language: tig
language_name: Tigre
language_family: semitic_ethiopic
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-semitic_ethiopic
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: 2.463
  - name: best_isotropy
    type: isotropy
    value: 0.6615
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-11
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Tigre** 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** | 2.305x | 2.31 | 0.2982% | 879,983 |
| **16k** | 2.463x 🏆 | 2.46 | 0.3185% | 823,793 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `አልአሚን ዐብደለጢፍ - ሰር-ዘመ ን እት ፈን እድሪስ መሐመድ ዐሊ ሐጂ ሕላይ - ወድ ባሸቂር፡ ሕላይ ሻም ሕላይ - ወድ ባሸቂር...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁አልአሚን ▁ዐብደለጢፍ ▁- ▁ሰር - ዘ መ ▁ን ▁እት ▁ፈን ... (+23 more)` | 33 |
| 16k | `▁አልአሚን ▁ዐብደለጢፍ ▁- ▁ሰር - ዘመ ▁ን ▁እት ▁ፈን ▁እድሪስ ... (+17 more)` | 27 |

**Sample 2:** `ብለዕ ወስታይ መንፈዐት ሐበት-አሰውዳ ምን ቡን አክል አዪ እግል ትስቴ ብከ ሐሊብ እንሰ ቀርፈ እከለት`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ብ ለዕ ▁ወ ስታ ይ ▁መንፈዐት ▁ሐበት - አሰውዳ ▁ምን ... (+10 more)` | 20 |
| 16k | `▁ብለዕ ▁ወስታይ ▁መንፈዐት ▁ሐበት - አሰውዳ ▁ምን ▁ቡን ▁አክል ▁አዪ ... (+7 more)` | 17 |

**Sample 3:** `ኣሜሪካ (እብ ኢንግሊዝ፥ United States of America) እት ቅብለት ኣሜሪካ ለትትረከብ ዐድ ተ። እብ ቅብለት ምስል ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ኣሜሪካ ▁( እብ ▁ኢ ንግሊዝ፥ ▁un ited ▁s t at ... (+42 more)` | 52 |
| 16k | `▁ኣሜሪካ ▁( እብ ▁ኢንግሊዝ፥ ▁united ▁states ▁of ▁america ) ▁እት ... (+27 more)` | 37 |


### Key Findings

- **Best Compression:** 16k achieves 2.463x compression
- **Lowest UNK Rate:** 8k with 0.2982% 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 | 5,051 | 12.30 | 7,801 | 13.2% | 43.4% |
| **2-gram** | Subword | 1,101 🏆 | 10.10 | 11,050 | 45.6% | 78.3% |
| **3-gram** | Word | 5,036 | 12.30 | 6,311 | 11.0% | 37.6% |
| **3-gram** | Subword | 8,481 | 13.05 | 53,840 | 19.1% | 46.6% |
| **4-gram** | Word | 23,464 | 14.52 | 25,105 | 3.3% | 9.9% |
| **4-gram** | Subword | 38,109 | 15.22 | 169,447 | 10.8% | 26.2% |
| **5-gram** | Word | 21,344 | 14.38 | 22,370 | 3.0% | 9.1% |
| **5-gram** | Subword | 76,266 | 16.22 | 232,751 | 6.8% | 19.0% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ምን ገብእ` | 530 |
| 2 | `እት ልብል` | 428 |
| 3 | `ሰበት ዐለ` | 355 |
| 4 | `እንዴ ቤለ` | 325 |
| 5 | `እሊ ህዬ` | 233 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ሓምድ እድሪስ ዓዋተ` | 108 |
| 2 | `መነዘመት ምጅልስ ቅራን` | 88 |
| 3 | `ሌጠ እንዴ ኢገብእ` | 87 |
| 4 | `መቃበለት ምሰል ኬትባይ` | 72 |
| 5 | `ቅብለት ምፍጋር ጸሓይ` | 70 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ቅብለት ምፍጋር ጸሓይ ሳሕል` | 63 |
| 2 | `ሜራስ አድጋማት ትግሬ ክምኩም` | 49 |
| 3 | `ክታብ ሜራስ አድጋማት ትግሬ` | 49 |
| 4 | `አድጋማት ትግሬ ክምኩም ድግም` | 42 |
| 5 | `እብ ዶ ር አሕመድ` | 41 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ክታብ ሜራስ አድጋማት ትግሬ ክምኩም` | 49 |
| 2 | `ሜራስ አድጋማት ትግሬ ክምኩም ድግም` | 42 |
| 3 | `እብ ዶ ር አሕመድ ሐሰን` | 41 |
| 4 | `ዶ ር አሕመድ ሐሰን ድሕሊ` | 41 |
| 5 | `እት ደንጎበ ናይ እሊ ምህሮ` | 31 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ እ` | 66,028 |
| 2 | `ት _` | 57,371 |
| 3 | `ል _` | 32,446 |
| 4 | `_ ለ` | 31,481 |
| 5 | `_ አ` | 28,736 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ እ ግ` | 14,781 |
| 2 | `እ ግ ል` | 12,703 |
| 3 | `ግ ል _` | 12,617 |
| 4 | `_ እ ን` | 12,149 |
| 5 | `_ እ ት` | 10,195 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `እ ግ ል _` | 12,107 |
| 2 | `_ እ ግ ል` | 12,029 |
| 3 | `እ ን ዴ _` | 9,201 |
| 4 | `_ እ ን ዴ` | 9,099 |
| 5 | `_ እ ት _` | 8,997 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ እ ግ ል _` | 11,475 |
| 2 | `_ እ ን ዴ _` | 9,019 |
| 3 | `_ ክ ም ሰ ል` | 3,323 |
| 4 | `እ ግ ል _ ል` | 3,125 |
| 5 | `ክ ም ሰ ል _` | 3,063 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 1,101
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~19% 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 | 0.7017 | 1.626 | 4.17 | 72,666 | 29.8% |
| **1** | Subword | 2.7582 | 6.766 | 44.54 | 494 | 0.0% |
| **2** | Word | 0.1717 | 1.126 | 1.32 | 302,688 | 82.8% |
| **2** | Subword | 1.0638 | 2.090 | 6.10 | 21,999 | 0.0% |
| **3** | Word | 0.0349 | 1.024 | 1.05 | 399,907 | 96.5% |
| **3** | Subword | 0.6056 | 1.522 | 2.94 | 134,244 | 39.4% |
| **4** | Word | 0.0091 🏆 | 1.006 | 1.01 | 418,313 | 99.1% |
| **4** | Subword | 0.4078 | 1.327 | 1.90 | 395,253 | 59.2% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `እግል ሓበሮት ወምስል ገሮቡ እንዴ አግንዐ እሉ ሐንስ ተምነዎ ምሰል ሰብ ዐድ ከአፎ ለአምሩ አማኖም ቱ`
2. `እት ሐበት አሰውደ ዲብ ኤስያት ወፓስፊክ 138 ብድሆ ናይ መትከባት ክም ትበጥር ገብአት አተላሌት ለሸሪጥ እሊ`
3. `እንዴ ከዐ እቶም አውመ እተ ጽንሖ እብል ትሰአልኩዉ አይወ ገሌ መደት ሰህ ጀነራል ተድለ ዑቅቢት ዐለ`

**Context Size 2:**

1. `ምን ገብእ አባይካ እለ ሊበል እላ ሐሊብ ጅሉጥ ኢቲበለ ተ ለትብለከ እሊ ላኪን እተ ለደረርኩም ዲቡ ዐድ`
2. `እት ልብል በሊስ ለገብእ እግሉ ሐዲስ አፍካር ምን ከምከሞት ላተ ይዓረፈ እት ደንጎበ ናይ እሊ ክታብ ለወሰከዩ`
3. `ሰበት ዐለ መዓርክ እንዴ ወዕለው ጎይላታት ድራሮም እት ልትበህል ልትህደግ እቡ እብ ምልሃዮም ልትጫፈሮ ወለአጎብሎ ዐለው ሰውረት`

**Context Size 3:**

1. `ሓምድ እድሪስ ዓዋተ ዩልዮ 196 ሓምድ እብራሂም መሐመድ ዐሊ ወዑመር ከራይ አብ ሓምድ ለትህየበ ተሕዚር አእንዴ ትቃወመው ሕነ`
2. `መነዘመት ምጅልስ ቅራን እተሓድ አፍሪቀ አልጃምዐ አልዐረብየ ወሐምሲተን ዳይማት አንፋር ምጅልስ አምን እግል ልቀስብ ለዐለት ሰእየት ክምሰል ፈሽለት`
3. `ሌጠ እንዴ ኢገብእ ዲብለ ዲብ እም ኩሉ ለዐለት መጥበዐት ልትጠበዕ ለዐለ ቱ ነፈዕ ወድ ዕትማን መን ቱ ነፈዕ`

**Context Size 4:**

1. `ቅብለት ምፍጋር ጸሓይ ሳሕል እግል ትደውሸሽ ክምቱ በሸረው ክልኢቶም ሜርሐት ወሕዳቶም እንዴ መርሐው ስስ ስዖታት እብ ድማናይ ደንበር እንዴ`
2. `ሜራስ አድጋማት ትግሬ ክምኩም ድግም ለትነፈ ሐት ቆሬዕ ክታብ ሜራስ አድጋማት ትግሬ ክምኩም ድግም ባርህ ወጻልም ክታብ ሜራስ አድጋማት`
3. `ክታብ ሜራስ አድጋማት ትግሬ ክምኩም ፋል እብ ነሃቅ አድግ ፋልመ ፋላት ለገ ብእ ምን ብዞሕ ሞላድ ለዐቀሙ ቶ ወእቡ`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_መናታክሉሉ_ኣሳደረአግየ_`
2. `ት_ካር።_ግለ_ማን_ህቶምል`
3. `እ_ሶ_አክ_ብ_ብልብ_ወሐቆ`

**Context Size 2:**

1. `_እት_አግማን_ቀርደመ።_ወራ`
2. `ት_ዐለት_ልትበሀልየት_እብ_`
3. `ል_እቱ_እግለ_አዜመ_እግል_`

**Context Size 3:**

1. `_እግል_ትርእዩ፡'_እግል_እን`
2. `እግል_ልርእዩ_ከልብ_።_(ለሔ`
3. `ግል_“ገለድ_ፈናኔን_ወእብ_በ`

**Context Size 4:**

1. `እግል_ልፍገሮ_ልትጸዐነው_ሲኪን`
2. `_እግል_ወጠነ።_._._.._ወለ`
3. `እንዴ_ትየመመ_ለለአበጽሑ_ለነሐ`


### Key Findings

- **Best Predictability:** Context-4 (word) with 99.1% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (395,253 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 | 28,756 |
| Total Tokens | 406,203 |
| Mean Frequency | 14.13 |
| Median Frequency | 3 |
| Frequency Std Dev | 143.43 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | እግል | 11,614 |
| 2 | እት | 9,133 |
| 3 | እንዴ | 9,068 |
| 4 | እብ | 7,587 |
| 5 | ዲብ | 7,025 |
| 6 | ምን | 6,293 |
| 7 | ህዬ | 3,645 |
| 8 | እሊ | 3,461 |
| 9 | ቱ | 3,197 |
| 10 | ክምሰል | 3,001 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | prayer | 2 |
| 2 | ነዊሕ | 2 |
| 3 | ሓውሳ | 2 |
| 4 | ለይቲ | 2 |
| 5 | ሩግ | 2 |
| 6 | ስኩር | 2 |
| 7 | ይቤ | 2 |
| 8 | ዋናጋሽ | 2 |
| 9 | ዘርሲ | 2 |
| 10 | ትእምርታት | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 0.9964 |
| R² (Goodness of Fit) | 0.996594 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 34.4% |
| Top 1,000 | 60.7% |
| Top 5,000 | 80.2% |
| Top 10,000 | 88.2% |

### Key Findings

- **Zipf Compliance:** R²=0.9966 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 34.4% of corpus
- **Long Tail:** 18,756 words needed for remaining 11.8% 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 | 0.6615 🏆 | 0.4348 | N/A | N/A |
| **mono_64d** | 64 | 0.2662 | 0.3804 | N/A | N/A |
| **mono_128d** | 128 | 0.0675 | 0.3801 | N/A | N/A |
| **aligned_32d** | 32 | 0.6615 | 0.4156 | 0.0233 | 0.1808 |
| **aligned_64d** | 64 | 0.2662 | 0.3694 | 0.0379 | 0.2857 |
| **aligned_128d** | 128 | 0.0675 | 0.3732 | 0.0787 | 0.3294 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.6615 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.3922. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 7.9% 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 | **-0.518** | 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 |
|--------|----------|
| `-ለ` | ለምህነት, ለዐቀብለ, ለደላ |
| `-ወ` | ወአከይ, ወእግሎም, ወደገልል |
| `-አ` | አስደረ, አእጅትመዕ, አተጀሀ |
| `-ት` | ትለብስ, ትደቀበ, ትዋጅህነ |
| `-መ` | መጃልስ, መልክ, መልህይ |
| `-ል` | ልትሃጌኒ, ልትሰአሎም, ልትካሬ |
| `-ልት` | ልትሃጌኒ, ልትሰአሎም, ልትካሬ |
| `-ኢ` | ኢመጽአው, ኢትረይሐው, ኢረአው |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-ት` | ለምህነት, ጂፒት, ሸሃደት |
| `-ም` | ወእግሎም, ሸካም, ለነሰም |
| `-ን` | ማይረፎን, ሓለፈየን, ድመን |
| `-ር` | ተዐይር, ሰምሀር, ታእሲር |
| `-ይ` | ወአከይ, ኔብራይ, መልህይ |
| `-የት` | ፍሌንየት, ውላየት, ወማህየት |
| `-ቶም` | ልትከቦቶም, በጽሐቶም, ጅማዐቶም |
| `-ይት` | ሓምሳይት, ጥይት, ምግበይት |

### 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 |
|------|----------|------------------|----------|
| `መልህያ` | 1.72x | 11 contexts | መልህያም, መልህያመ, መልህያሙ |
| `ልትአመ` | 1.54x | 11 contexts | ልትአመር, ልትአመን, ልትአመሮ |
| `እርትር` | 1.65x | 9 contexts | እርትርያ, እርትርየ, እርትርያይ |
| `አርወሐ` | 1.57x | 10 contexts | አርወሐት, አርወሐቱ, አርወሐቼ |
| `ለትፈና` | 1.67x | 8 contexts | ለትፈናተ, ለትፈናታ, ወለትፈናተ |
| `ልትበህ` | 1.64x | 8 contexts | ልትበህሉ, ልትበህሎ, ልትበህል |
| `ለልትበ` | 1.45x | 11 contexts | ለልትበህለ, ለልትበሀለ, ለልትበሀሎ |
| `ኤረትር` | 1.53x | 9 contexts | ኤረትርያ, ኤረትርየ, ኤረትርዪን |
| `ትረከብ` | 1.52x | 8 contexts | ልትረከብ, ትትረከብ, ኢልትረከብ |
| `ትአመር` | 1.39x | 10 contexts | ትትአመር, ልትአመር, ኢትትአመር |
| `ብራሂም` | 1.70x | 6 contexts | አብራሂም, እብራሂም, ኢብራሂም |
| `ልትበሀ` | 1.49x | 8 contexts | ልትበሀል, ልትበሀለ, ልትበሀሎ |

### 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 |
|--------|--------|-----------|----------|
| `-ለ` | `-ም` | 12 words | ለአገርም, ለአልቃም |
| `-ወ` | `-ት` | 10 words | ወአእት, ወዝብጠት |
| `-ለ` | `-ት` | 5 words | ለምዴርየት, ለሔልየት |
| `-ለ` | `-ዮም` | 5 words | ለትሰመዐዮም, ለሐረዮም |
| `-ለ` | `-ር` | 5 words | ለሄራር, ለትቀድር |
| `-ወ` | `-ም` | 5 words | ወጸገም, ወፈሀም |
| `-ለ` | `-ን` | 4 words | ለአቅርን, ለኢልተመን |
| `-እ` | `-ት` | 4 words | እቅቡላት, እስባታት |
| `-እ` | `-የት` | 4 words | እሕሳእየት, እስብዳልየት |
| `-አ` | `-ት` | 3 words | አውለውያት, አፍዐበት |

### 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 |
|------|-----------------|------------|------|
| ወእተክምሰልሁመ | **`ወ-እተክምሰልሁመ`** | 4.5 | `እተክምሰልሁመ` |
| ወለልአስተሽህድ | **`ወ-ለ-ልአስተሽህድ`** | 3.0 | `ልአስተሽህድ` |
| ወለምትከብታይመ | **`ወ-ለ-ምትከብታይመ`** | 3.0 | `ምትከብታይመ` |
| ተወልዳዴመድህን | **`ተ-ወ-ልዳዴመድህን`** | 3.0 | `ልዳዴመድህን` |
| ኤለክትሮኒካይት | **`ኤለክትሮኒካይ-ት`** | 1.5 | `ኤለክትሮኒካይ` |
| ለሐቡሸትወአርዌተኒ | **`ለ-ሐቡሸትወአርዌተኒ`** | 1.5 | `ሐቡሸትወአርዌተኒ` |
| መሐመድአልአሚን | **`መ-ሐመድአልአሚን`** | 1.5 | `ሐመድአልአሚን` |
| ብዕራይኢረክበት | **`ብዕራይኢረክበ-ት`** | 1.5 | `ብዕራይኢረክበ` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Tigre 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 | **16k BPE** | Best compression (2.46x) |
| N-gram | **2-gram** | Lowest perplexity (1,101) |
| Markov | **Context-4** | Highest predictability (99.1%) |
| 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: 2026-01-11 00:55:27*