---
language: ti
language_name: Tigrinya
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: 3.058
  - name: best_isotropy
    type: isotropy
    value: 0.1219
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-11
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Tigrinya** 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** | 2.515x | 2.52 | 0.2599% | 148,897 |
| **16k** | 2.779x | 2.78 | 0.2872% | 134,751 |
| **32k** | 3.058x 🏆 | 3.06 | 0.3160% | 122,449 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `ኢጣልያ (፣ ) ብወግዒ ኢጣልያዊት ሪፓብሊክ ()፣ ኣባልን መስራቲትን ኤውሮጳዊ ሕብረት፣ ስግረ-ኣህጉር ልኡላዊት ሃገር እያ። ር...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ኢጣልያ ▁( ፣ ▁) ▁ብወግዒ ▁ኢጣልያ ዊት ▁ሪፓብሊክ ▁() ፣ ... (+18 more)` | 28 |
| 16k | `▁ኢጣልያ ▁( ፣ ▁) ▁ብወግዒ ▁ኢጣልያ ዊት ▁ሪፓብሊክ ▁() ፣ ... (+17 more)` | 27 |
| 32k | `▁ኢጣልያ ▁( ፣ ▁) ▁ብወግዒ ▁ኢጣልያዊት ▁ሪፓብሊክ ▁() ፣ ▁ኣባልን ... (+14 more)` | 24 |

**Sample 2:** `ኣርጀንቲና (፣ )፣ ብወግዒ ሪፓብሊክ ኣርጀንቲና (፣ )፣ ኣብ ደቡባዊ ሸነኽ ናይ ደቡብ ኣመሪካ እትርከብ ምስ ኣትላንቲካዊ ውቅ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ኣርጀንቲና ▁( ፣ ▁) ፣ ▁ብወግዒ ▁ሪፓብሊክ ▁ኣርጀንቲና ▁( ፣ ... (+28 more)` | 38 |
| 16k | `▁ኣርጀንቲና ▁( ፣ ▁) ፣ ▁ብወግዒ ▁ሪፓብሊክ ▁ኣርጀንቲና ▁( ፣ ... (+25 more)` | 35 |
| 32k | `▁ኣርጀንቲና ▁( ፣ ▁) ፣ ▁ብወግዒ ▁ሪፓብሊክ ▁ኣርጀንቲና ▁( ፣ ... (+22 more)` | 32 |

**Sample 3:** `ማቲው ስቲቨን ሹልዘ (Matthew Steven «Matt» Schulze) ኣሜሪካዊ ተዋሳኣይ ፊልም እዩ። ኣብ ሚዙሪ እዩ ተወሊዱ።...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ማ ቲ ው ▁ስቲቨን ▁ሹ ልዘ ▁( mat th ew ... (+40 more)` | 50 |
| 16k | `▁ማቲው ▁ስቲቨን ▁ሹልዘ ▁( mat th ew ▁steven ▁« matt ... (+29 more)` | 39 |
| 32k | `▁ማቲው ▁ስቲቨን ▁ሹልዘ ▁( matthew ▁steven ▁« matt » ▁schulze ... (+22 more)` | 32 |


### Key Findings

- **Best Compression:** 32k achieves 3.058x compression
- **Lowest UNK Rate:** 8k with 0.2599% 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 | 674 | 9.40 | 936 | 35.2% | 100.0% |
| **2-gram** | Subword | 1,449 | 10.50 | 6,000 | 36.6% | 74.2% |
| **3-gram** | Word | 494 🏆 | 8.95 | 653 | 38.5% | 100.0% |
| **3-gram** | Subword | 7,666 | 12.90 | 20,589 | 14.3% | 42.7% |
| **4-gram** | Word | 1,390 | 10.44 | 1,640 | 18.2% | 67.9% |
| **4-gram** | Subword | 19,863 | 14.28 | 45,780 | 8.8% | 28.2% |
| **5-gram** | Word | 1,166 | 10.19 | 1,246 | 17.6% | 82.5% |
| **5-gram** | Subword | 24,432 | 14.58 | 45,809 | 6.5% | 24.0% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ኩዕሶ እግሪ` | 161 |
| 2 | `ከምኡ ውን` | 138 |
| 3 | `0 1` | 105 |
| 4 | `upright 0` | 103 |
| 5 | `frameless upright` | 103 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `upright 0 1` | 103 |
| 2 | `frameless upright 0` | 103 |
| 3 | `ቅድሚ ልደተ ክርስቶስ` | 28 |
| 4 | `ሰለላሁ ዓለይሂ ወሰለም` | 23 |
| 5 | `ሙሓመድ ሰለላሁ ዓለይሂ` | 23 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `frameless upright 0 1` | 103 |
| 2 | `ሙሓመድ ሰለላሁ ዓለይሂ ወሰለም` | 23 |
| 3 | `ነቢይ ሙሓመድ ሰለላሁ ዓለይሂ` | 21 |
| 4 | `ንዓኻ ንዓኻ ንዓኻ ንዓኻ` | 16 |
| 5 | `ፕሮፌሽናል ተጻዋታይ ኩዕሶ እግሪ` | 15 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ነቢይ ሙሓመድ ሰለላሁ ዓለይሂ ወሰለም` | 21 |
| 2 | `ንዓኻ ንዓኻ ንዓኻ ንዓኻ ንዓኻ` | 15 |
| 3 | `ፕሮፌሽናል ተጻዋታይ ኩዕሶ እግሪ ኮይኑ` | 13 |
| 4 | `p q r s t` | 10 |
| 5 | `5 frameless upright 0 1` | 10 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ ኣ` | 7,078 |
| 2 | `ት _` | 6,640 |
| 3 | `ን _` | 6,434 |
| 4 | `ብ _` | 5,376 |
| 5 | `_ እ` | 4,167 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ ኣ ብ` | 3,209 |
| 2 | `ኣ ብ _` | 2,860 |
| 3 | `ታ ት _` | 1,640 |
| 4 | `_ ካ ብ` | 965 |
| 5 | `_ ና ይ` | 961 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ ኣ ብ _` | 2,832 |
| 2 | `_ ና ይ _` | 750 |
| 3 | `_ ካ ብ _` | 731 |
| 4 | `_ ድ ማ _` | 658 |
| 5 | `_ እ ዩ ።` | 577 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ እ ዩ ። _` | 522 |
| 2 | `። _ ኣ ብ _` | 424 |
| 3 | `፡ _ ኣ ብ _` | 350 |
| 4 | `_ ኣ ብ _ መ` | 297 |
| 5 | `ኢ ት ዮ ጵ ያ` | 264 |


### Key Findings

- **Best Perplexity:** 3-gram (word) with 494
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~24% 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 | 0.6172 | 1.534 | 3.00 | 20,182 | 38.3% |
| **1** | Subword | 1.7048 | 3.260 | 18.92 | 788 | 0.0% |
| **2** | Word | 0.1201 | 1.087 | 1.20 | 60,235 | 88.0% |
| **2** | Subword | 0.8301 | 1.778 | 4.02 | 14,892 | 17.0% |
| **3** | Word | 0.0269 | 1.019 | 1.04 | 71,825 | 97.3% |
| **3** | Subword | 0.5079 | 1.422 | 2.25 | 59,764 | 49.2% |
| **4** | Word | 0.0074 🏆 | 1.005 | 1.01 | 74,088 | 99.3% |
| **4** | Subword | 0.2614 | 1.199 | 1.48 | 134,188 | 73.9% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `ኣብ ዓንቀጻት ኮንፈደረሽን ኩዕሶ እግሪ ክለብ ኮይና ኣስታት 115 ኪሎመተር ሪሒቓ ትርከብ አብ ሰሜን ኣህጉር ኣል`
2. `ናይ ጭንቀት ኣብ ድማ ሓደ ኣርእስቲ ካብቲ ቦታ ብኢንፎርሜሽን እና ቱማስ ሆሎፔይንየን ኣብ ዝኾነ ቁርኣን ብስም`
3. `እዩ ሊኢኽዎም ገለ ካብቶም ብብዝሒ ተተኰስትን ማረኸ እዚ ካልኣይ ደረጃ ብምሓዝ ንብዙሓት ኣዝዩ ቅዱስ ብትግርኛ መጻሕፍቲ`

**Context Size 2:**

1. `ኩዕሶ እግሪ ክለብ እያ ኣብ ህንዲ ካብ ዘለዋ ዓበይቲ ደገፍቲ ሓንቲ እያ እታ ክለብ ኣብ ከተማ ዓድ`
2. `ከምኡ ውን እቲ ዓሚል ክፍሊት ንኽገብር ዝሕግዙ ኣማራጺታት ይሕብር ሓደ ዓሚል ንኣቕሑ ንምልዋጥ ወይ ድሕሪ ምፍንጃር ምስትንፋስ`
3. `0 1 ሪፓብሊክ ኮንጎ 2 344 858 30 5 frameless upright 0 1 ኡጋንዳ ሪፓብሊክ ኡጋንዳ 241`

**Context Size 3:**

1. `frameless upright 0 1 ላትቭያ ሪፓብሊክ ላትቭያ 64 589 1 925 800 34 3 frameless upright 0 1`
2. `upright 0 1 ኤርትራ ሃገረ ኤርትራ 117 600 5 869 869 37 frameless upright 0 1 ስዊዘርላንድ ኮንፈደረሽን`
3. `ቅድሚ ልደተ ክርስቶስ ብኣካሜኒድ ገዛኢ ቂሮስ ዓቢ ዝጠፍኡ ጥንታዊነት ዘመነ ሄለኒስትን ዘመነ ቢዛንታይንን ሰፈራታት ኤዮልያን ኣዮንያንን ግሪኽን ብሰፊሑ`

**Context Size 4:**

1. `frameless upright 0 1 ቱርኪ ሪፓብሊክ ቱርኪ 783 356 105 frameless upright 0 1 ስዋዚላንድ ንግስነት ስዋዚላንድ 17 364`
2. `ሙሓመድ ሰለላሁ ዓለይሂ ወሰለም ድማ ነቲ ዘይተማለአ ሕግታት ብምጽፋፍ ንኹሉ መዳያት ህይወት ሓደ ብሓደ ዝትንክፍ ጎደሎ ዘይብሉ ሃብታምን ውዱእን`
3. `ነቢይ ሙሓመድ ሰለላሁ ዓለይሂ ወሰለም ብ ህላወ መላእኽቲ ኣላህ ክንኣምን እውን ኣዚዙና ኢዩ ካብቶም ዝጠቐስናዮም ሽዱሽተ ዓንድታት እምነት ድሕሪ`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_ኣበይ_ሰለ-ት_ና_anc_`
2. `ንኩባ።_ሓዘ_፡_ኣካ_ሕ_ግ`
3. `ብ_ክር_ፋጭንግራት፣_ደ_ና`

**Context Size 2:**

1. `_ኣብ_ፊን_ብህይወት_ስሞም_`
2. `ት_ሱፐር_ዝወድአ_።_ነይራ_`
3. `ን_16._171_ግዜ_ብግቡኡ`

**Context Size 3:**

1. `_ኣብኡ_ድማ፡_ኣሃዱታት_7_ዋ`
2. `ኣብ_ዝነበረን_ዝኣዘዘ’ሞ፡_ከ`
3. `ታት_ንምእማን_ኣይሁድን_ና_ያ`

**Context Size 4:**

1. `_ኣብ_ኢትዮጵያዊ_ኣወሃሃዲ_ሙዚ`
2. `_ናይ_መጀመርታ_ሰፈራታት_ዝኾነ`
3. `_ካብ_ዝምዕብላ_ዘለዋ_እንትኸው`


### Key Findings

- **Best Predictability:** Context-4 (word) with 99.3% predictability
- **Branching Factor:** Decreases with context size (more deterministic)
- **Memory Trade-off:** Larger contexts require more storage (134,188 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 | 7,251 |
| Total Tokens | 64,854 |
| Mean Frequency | 8.94 |
| Median Frequency | 3 |
| Frequency Std Dev | 43.70 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ኣብ | 2,873 |
| 2 | እዩ | 820 |
| 3 | ናይ | 807 |
| 4 | ካብ | 750 |
| 5 | ድማ | 704 |
| 6 | እቲ | 554 |
| 7 | ምስ | 433 |
| 8 | ከም | 405 |
| 9 | እዚ | 370 |
| 10 | ሓደ | 339 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ቻንቻን | 2 |
| 2 | ጳውሎስ | 2 |
| 3 | ኮቺን | 2 |
| 4 | ፕሪንስተን | 2 |
| 5 | ኣሉቫ | 2 |
| 6 | ኮታያም | 2 |
| 7 | ቫርጌስ | 2 |
| 8 | ዶክትሬት | 2 |
| 9 | ኮሚቴን | 2 |
| 10 | ምኽትል | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 0.9129 |
| R² (Goodness of Fit) | 0.984365 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 31.6% |
| Top 1,000 | 66.4% |
| Top 5,000 | 93.1% |
| Top 10,000 | 0.0% |

### Key Findings

- **Zipf Compliance:** R²=0.9844 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 31.6% of corpus
- **Long Tail:** -2,749 words needed for remaining 100.0% 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 | 0.1219 🏆 | 0.5907 | N/A | N/A |
| **mono_64d** | 64 | 0.0304 | 0.6195 | N/A | N/A |
| **mono_128d** | 128 | 0.0069 | 0.6350 | N/A | N/A |
| **aligned_32d** | 32 | 0.1219 | 0.6074 | 0.0108 | 0.2703 |
| **aligned_64d** | 64 | 0.0304 | 0.6287 | 0.0216 | 0.2973 |
| **aligned_128d** | 128 | 0.0069 | 0.6320 | 0.0486 | 0.4054 |

### Key Findings

- **Best Isotropy:** mono_32d with 0.1219 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.6189. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 4.9% 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 | **2.433** | 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 |
|--------|----------|
| `-ኣ` | ኣለው, ኣስዒቡ, ኣገዳሲት |
| `-ዝ` | ዝዓቐኑ, ዝባን, ዝነብሩላ |
| `-ብ` | ብ19, ብምቁጽጻር, ብሕቲ |
| `-ን` | ንዖኦም, ንዋትን, ንቁጠባ |
| `-ተ` | ተቘጻጸራኦ, ተቖጺሮም, ተርጓሚ |
| `-ም` | ምስተለኽፈ, ምስሊ, ምትሓዝ |
| `-መ` | መርዓውን, መንጎ, መዓስከር |
| `-ክ` | ክርስትያናዊት, ክትዓት, ክምረዙ |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-ን` | ዝባን, ከረጋግጽን, ንዋትን |
| `-ት` | ፓራስፖርት, ለካቲት, ጸብጻባት |
| `-ትን` | ንዋትን, ሰባትን, ስርዓትን |
| `-ያን` | ሰልጁካውያን, ኣህጉራውያን, ሽወደናውያን |
| `-ታት` | ስጉምታት, ባህልታት, ጠንቅታት |
| `-ነት` | መንግስታዊነት, ንኣብነት, ንእስነት |
| `-ንን` | ሱዳንን, ተለቪዥንን, ግሪኻውያንን |

### 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.

*No significant bound stems detected.*


### 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 |
|--------|--------|-----------|----------|
| `-ኣ` | `-ን` | 20 words | ኣዝየን, ኣህጉራውያን |
| `-ኣ` | `-ያን` | 9 words | ኣህጉራውያን, ኣውስትርያን |
| `-መ` | `-ን` | 8 words | መርዓውን, መታን |
| `-ብ` | `-ን` | 6 words | ብፌደሬሽን, ብዙሃን |
| `-መ` | `-ት` | 5 words | መንግስታዊነት, መስመራት |
| `-ም` | `-ን` | 5 words | ምምቕቓልን, ምቕራብን |
| `-መ` | `-ትን` | 5 words | መግብታትን, መምርሒታትን |
| `-መ` | `-ታት` | 4 words | መጥቃዕቲታት, መልእኽትታት |
| `-ክ` | `-ት` | 3 words | ክርስትያናዊት, ክትዓት |
| `-ብ` | `-ት` | 3 words | ብፕረዚደንት, ብምኽንያት |

### 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 |
|------|-----------------|------------|------|
| ቱርክመኒስታንን | **`ቱርክመኒስታን-ን`** | 1.5 | `ቱርክመኒስታን` |
| ኣሰላሙዓለይኩም | **`ኣ-ሰላሙዓለይኩም`** | 1.5 | `ሰላሙዓለይኩም` |
| ኣውስትራሊያውያን | **`ኣውስትራሊያውያ-ን`** | 1.5 | `ኣውስትራሊያውያ` |
| ኢንሳይክሎፔድያን | **`ኢንሳይክሎፔድያ-ን`** | 1.5 | `ኢንሳይክሎፔድያ` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Tigrinya 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 | **32k BPE** | Best compression (3.06x) |
| N-gram | **3-gram** | Lowest perplexity (494) |
| Markov | **Context-4** | Highest predictability (99.3%) |
| 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 00:50:27*