README.md

---
language: kn
language_name: Kannada
language_family: dravidian_south
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-dravidian_south
license: mit
library_name: wikilangs
pipeline_tag: text-generation
datasets:
  - omarkamali/wikipedia-monthly
dataset_info:
  name: wikipedia-monthly
  description: Monthly snapshots of Wikipedia articles across 300+ languages
metrics:
  - name: best_compression_ratio
    type: compression
    value: 5.009
  - name: best_isotropy
    type: isotropy
    value: 0.7989
  - name: vocabulary_size
    type: vocab
    value: 0
generated: 2026-01-10
---

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

This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Kannada** 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** | 3.683x | 3.68 | 0.0965% | 1,750,699 |
| **16k** | 4.165x | 4.16 | 0.1092% | 1,548,001 |
| **32k** | 4.627x | 4.62 | 0.1213% | 1,393,713 |
| **64k** | 5.009x 🏆 | 5.01 | 0.1313% | 1,287,306 |

### Tokenization Examples

Below are sample sentences tokenized with each vocabulary size:

**Sample 1:** `ಕನ್ನಡ ಸಾಹಿತ್ಯ ಹಾಗು ಭಾಷೆಯ ಏಳ್ಗೆಗೆ ದುಡಿಯತ್ತಿರುವ ಶಿವಮೂಗ್ಗದ ಸಂಸ್ಥೆ. ಇದು ಖ್ಯಾತ ಸಾಹಿತಿ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ಕನ್ನಡ ▁ಸಾಹಿತ್ಯ ▁ಹಾಗು ▁ಭಾಷೆಯ ▁ಏ ಳ್ ಗೆ ಗೆ ▁ದು ಡಿಯ ... (+24 more)` | 34 |
| 16k | `▁ಕನ್ನಡ ▁ಸಾಹಿತ್ಯ ▁ಹಾಗು ▁ಭಾಷೆಯ ▁ಏ ಳ್ ಗೆ ಗೆ ▁ದುಡಿಯ ತ್ತ ... (+22 more)` | 32 |
| 32k | `▁ಕನ್ನಡ ▁ಸಾಹಿತ್ಯ ▁ಹಾಗು ▁ಭಾಷೆಯ ▁ಏ ಳ್ ಗೆ ಗೆ ▁ದುಡಿಯ ತ್ತ ... (+20 more)` | 30 |
| 64k | `▁ಕನ್ನಡ ▁ಸಾಹಿತ್ಯ ▁ಹಾಗು ▁ಭಾಷೆಯ ▁ಏ ಳ್ ಗೆಗೆ ▁ದುಡಿಯ ತ್ತಿರುವ ▁ಶಿವ ... (+18 more)` | 28 |

**Sample 2:** `ಪ್ರೌಡ ದೇವರಾಯ ಅಥವಾ ಪ್ರೌಡ ರಾಯ ಸ್ವಲ್ಪ ಕಾಲ ವಿಜಯನಗರ ಸಾಮ್ರಾಜ್ಯವನ್ನು ಆಳಿದವ. ಜನಪ್ರಿಯತೆ ಇ...`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ಪ್ರೌ ಡ ▁ದೇವರ ಾಯ ▁ಅಥವಾ ▁ಪ್ರೌ ಡ ▁ರಾಯ ▁ಸ್ವಲ್ಪ ▁ಕಾಲ ... (+27 more)` | 37 |
| 16k | `▁ಪ್ರೌ ಡ ▁ದೇವರ ಾಯ ▁ಅಥವಾ ▁ಪ್ರೌ ಡ ▁ರಾಯ ▁ಸ್ವಲ್ಪ ▁ಕಾಲ ... (+22 more)` | 32 |
| 32k | `▁ಪ್ರೌಡ ▁ದೇವರ ಾಯ ▁ಅಥವಾ ▁ಪ್ರೌಡ ▁ರಾಯ ▁ಸ್ವಲ್ಪ ▁ಕಾಲ ▁ವಿಜಯನಗರ ▁ಸಾಮ್ರಾಜ್ಯವನ್ನು ... (+19 more)` | 29 |
| 64k | `▁ಪ್ರೌಡ ▁ದೇವರಾಯ ▁ಅಥವಾ ▁ಪ್ರೌಡ ▁ರಾಯ ▁ಸ್ವಲ್ಪ ▁ಕಾಲ ▁ವಿಜಯನಗರ ▁ಸಾಮ್ರಾಜ್ಯವನ್ನು ▁ಆಳಿದ ... (+17 more)` | 27 |

**Sample 3:** `ದಂಡಂ ದಶಗುಣಂ ಗಣೇಶ್ ನಿರ್ಮಾಣದ ರಮ್ಯಾ ಅಭಿನಯದ ಚಿತ್ರ. ಚಲನಚಿತ್ರಗಳು ಕನ್ನಡಚಿತ್ರಗಳು`

| Vocab | Tokens | Count |
|-------|--------|-------|
| 8k | `▁ದಂಡ ಂ ▁ದಶ ಗುಣ ಂ ▁ಗಣೇಶ್ ▁ನಿರ್ಮಾಣದ ▁ರಮ ್ಯಾ ▁ಅಭಿನಯದ ... (+4 more)` | 14 |
| 16k | `▁ದಂಡ ಂ ▁ದಶ ಗುಣ ಂ ▁ಗಣೇಶ್ ▁ನಿರ್ಮಾಣದ ▁ರಮ್ಯಾ ▁ಅಭಿನಯದ ▁ಚಿತ್ರ ... (+3 more)` | 13 |
| 32k | `▁ದಂಡ ಂ ▁ದಶ ಗುಣ ಂ ▁ಗಣೇಶ್ ▁ನಿರ್ಮಾಣದ ▁ರಮ್ಯಾ ▁ಅಭಿನಯದ ▁ಚಿತ್ರ ... (+3 more)` | 13 |
| 64k | `▁ದಂಡ ಂ ▁ದಶ ಗುಣ ಂ ▁ಗಣೇಶ್ ▁ನಿರ್ಮಾಣದ ▁ರಮ್ಯಾ ▁ಅಭಿನಯದ ▁ಚಿತ್ರ ... (+3 more)` | 13 |


### Key Findings

- **Best Compression:** 64k achieves 5.009x compression
- **Lowest UNK Rate:** 8k with 0.0965% 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 | 147,664 | 17.17 | 361,190 | 3.3% | 13.3% |
| **2-gram** | Subword | 2,880 🏆 | 11.49 | 87,389 | 31.1% | 65.4% |
| **3-gram** | Word | 84,882 | 16.37 | 242,324 | 5.0% | 23.3% |
| **3-gram** | Subword | 27,323 | 14.74 | 675,163 | 12.2% | 31.7% |
| **4-gram** | Word | 177,938 | 17.44 | 505,335 | 5.0% | 21.6% |
| **4-gram** | Subword | 159,038 | 17.28 | 2,979,693 | 6.6% | 18.3% |
| **5-gram** | Word | 120,926 | 16.88 | 396,218 | 6.2% | 25.9% |
| **5-gram** | Subword | 537,405 | 19.04 | 5,882,653 | 3.7% | 11.2% |

### Top 5 N-grams by Size

**2-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ದೇವಾಲಯ ಶ್ರೀ` | 5,347 |
| 2 | `ಮತ್ತು ಇತರ` | 4,941 |
| 3 | `ಎಂದು ಕರೆಯಲಾಗುತ್ತದೆ` | 4,574 |
| 4 | `of the` | 4,518 |
| 5 | `ಕಿ ಮೀ` | 4,484 |

**3-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ಉಲ್ಲೇಖಗಳು ಬಾಹ್ಯ ಕೊಂಡಿಗಳು` | 2,093 |
| 2 | `c ಡಿಗ್ರಿ ಸೆಲ್ಸಿಯಸ್` | 1,487 |
| 3 | `nr nr nr` | 1,029 |
| 4 | `ಇ ಲರ್ನಿಂಗ್‍ನಲ್ಲಿ ತಯಾರಿಸಿದ` | 1,003 |
| 5 | `ಲರ್ನಿಂಗ್‍ನಲ್ಲಿ ತಯಾರಿಸಿದ ಲೇಖನ` | 1,003 |

**4-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ಇ ಲರ್ನಿಂಗ್‍ನಲ್ಲಿ ತಯಾರಿಸಿದ ಲೇಖನ` | 1,003 |
| 2 | `ಯುಗಾದಿ ದಸರಾ ದೀಪಾವಳಿ ನಾಗರ` | 891 |
| 3 | `ತೆರದ ಬಾವಿ ಕೊಳವೆ ಬಾವಿಯಿಂದ` | 891 |
| 4 | `ದಸರಾ ದೀಪಾವಳಿ ನಾಗರ ಪಂಚಮಿ` | 891 |
| 5 | `ಹಾಗೂ ಇತರೆ ಬೆಳೆಗಳನ್ನು ಬೆಳೆಯುತ್ತಾರೆ` | 890 |

**5-grams (Word):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ಯುಗಾದಿ ದಸರಾ ದೀಪಾವಳಿ ನಾಗರ ಪಂಚಮಿ` | 891 |
| 2 | `ಬಾವಿ ಕೊಳವೆ ಬಾವಿಯಿಂದ ನೀರಾವರಿ ಇದ್ದು` | 888 |
| 3 | `ಕೊಳವೆ ಬಾವಿಯಿಂದ ನೀರಾವರಿ ಇದ್ದು ಪ್ರಮುಖವಾಗಿ` | 888 |
| 4 | `ತೆರದ ಬಾವಿ ಕೊಳವೆ ಬಾವಿಯಿಂದ ನೀರಾವರಿ` | 887 |
| 5 | `ಗೋಧಿ ಹಾಗೂ ಇತರೆ ಬೆಳೆಗಳನ್ನು ಬೆಳೆಯುತ್ತಾರೆ` | 884 |

**2-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `. _` | 1,472,306 |
| 2 | `ದ _` | 1,245,673 |
| 3 | `_ ಅ` | 1,146,612 |
| 4 | `, _` | 1,086,756 |
| 5 | `ಲ್ ಲಿ` | 956,528 |

**3-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ನ್ ನು _` | 864,437 |
| 2 | `ಲ್ ಲಿ _` | 739,919 |
| 3 | `ತ್ ತು _` | 467,651 |
| 4 | `_ ಮ ತ್` | 466,426 |
| 5 | `ಮ ತ್ ತು` | 448,388 |

**4-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `ಮ ತ್ ತು _` | 446,552 |
| 2 | `_ ಮ ತ್ ತು` | 445,542 |
| 3 | `ದ ಲ್ ಲಿ _` | 268,843 |
| 4 | `ಳ ನ್ ನು _` | 246,796 |
| 5 | `ಗ ಳ ನ್ ನು` | 245,176 |

**5-grams (Subword):**

| Rank | N-gram | Count |
|------|--------|-------|
| 1 | `_ ಮ ತ್ ತು _` | 443,934 |
| 2 | `ಗ ಳ ನ್ ನು _` | 240,980 |
| 3 | `ತ್ ತ ದೆ . _` | 150,324 |
| 4 | `ಗ ಳ ಲ್ ಲಿ _` | 130,864 |
| 5 | `_ ಅ ವ ರು _` | 81,945 |


### Key Findings

- **Best Perplexity:** 2-gram (subword) with 2,880
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
- **Coverage:** Top-1000 patterns cover ~11% 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 | 0.7535 | 1.686 | 7.09 | 1,777,584 | 24.7% |
| **1** | Subword | 1.0667 | 2.095 | 20.14 | 10,349 | 0.0% |
| **2** | Word | 0.1931 | 1.143 | 1.43 | 12,594,272 | 80.7% |
| **2** | Subword | 1.0052 | 2.007 | 8.11 | 208,364 | 0.0% |
| **3** | Word | 0.0354 | 1.025 | 1.05 | 17,987,740 | 96.5% |
| **3** | Subword | 0.6231 | 1.540 | 3.82 | 1,690,414 | 37.7% |
| **4** | Word | 0.0089 🏆 | 1.006 | 1.01 | 18,916,474 | 99.1% |
| **4** | Subword | 0.4773 | 1.392 | 2.48 | 6,451,364 | 52.3% |

### Generated Text Samples (Word-based)

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

**Context Size 1:**

1. `ಮತ್ತು ಬ್ರಹ್ಮ ಪರಾ ಪ್ರಕೃತಿ ಚರಿತ್ರೆ ಉಲ್ಲೇಖಗಳು ಉಲ್ಲೇಖಗಳು ಕಲಾವಿದರು ಮತ್ತು ಗ್ರೀನ್ ರೂಮ್ ಪುಟ್ಟ ಬಾಕ್ಸ್ ಆಫ಼ಿಸ್ ...`
2. `ಈ ಕೆಳಕಂಡ ಅಗತ್ಯಗಳನ್ನು ಪೂರೈಸುವುದಕ್ಕೆ ವಿನಿಯೋಗಿಸಲು ಸಾಧ್ಯವಾಗಲಿಲ್ಲ ಅಮಿತ್ ಶಾ ಅವರ ಪ್ರಭುಲಿಂಗ ಗುರು ಚಾಂದಗಿ ರಾಮ್...`
3. `ಒಂದು ನಕ್ಷತ್ರದ ದಿನದಂದು ಚೆನ್ನೈ ಮೇಲೆ ಫ್ರೆಂಚ್ ಭಾಷೆಯಲ್ಲಿ 1 ಪಾಸ್‌ವರ್ಡ್ಸ್ ಕ್ರೆಡಿಟ್ ಕಾರ್ಡ್ ಮೂಲಕ ತಮ್ಮನ್ನು ತಾವ...`

**Context Size 2:**

1. `ದೇವಾಲಯ ಶ್ರೀ ದುರ್ಗಾದೇವಿ ದೇವಾಲಯ ಶ್ರೀ ಪಾಂಡುರಂಗ ದೇವಾಲಯ ಶ್ರೀ ಹಣಮಂತ ದೇವಾಲಯ ಮಸೀದಿ ಗ್ರಾಮದಲ್ಲಿ ಮುಸ್ಲಿಂ ಸಮುದಾಯ...`
2. `ಮತ್ತು ಇತರ ಕತೆಗಳು ನಡೆದು ಬಂದ ಪ್ರಬಲ ಪೈಪೋಟಿಕಠಿಣವಾಗಿತ್ತು 40ನೇ ನಿಮಿಷದಲ್ಲಿ ಪಂದ್ಯವನ್ನು ಸರಿಸಮ ಮಾಡಿಕೊಳ್ಳುವ ಕುರ...`
3. `ಎಂದು ಕರೆಯಲಾಗುತ್ತದೆ ಪರಿವಿಡಿ 1 ಆರಂಭಿಕ ರಾಜವಂಶೀಯ ಅವಧಿ ಕ್ರಿ ಪೂ ರಚಿತವಾಯಿತು ಇದರಲ್ಲಿ ಈಜಿಪ್ಟಿನ ಚಿನ್ನದ ಗಣಿಯನ್ನ...`

**Context Size 3:**

1. `ಉಲ್ಲೇಖಗಳು ಬಾಹ್ಯ ಕೊಂಡಿಗಳು ಕನ್ನಡಚಿತ್ರಗಳು ನಿರ್ಮಾಣಗೊಂಡ ಚಲನಚಿತ್ರಗಳು ಚಲನಚಿತ್ರಗಳು`
2. `c ಡಿಗ್ರಿ ಸೆಲ್ಸಿಯಸ್ ಚಳಿಗಾಲ ಮತ್ತು ಮಳೆಗಾಲ 18 c 30 c ಡಿಗ್ರಿ ಸೆಲ್ಸಿಯಸ್ ಚಳಿಗಾಲ ಮತ್ತು ಮಳೆಗಾಲ 18 c 30`
3. `nr nr nr ಕಾವಲುಗಾರ ಶ್ವಾನಗಳು ಹಮಗೇರಿಯನ್‌ ಹೌಂಡ್‌ ಹಂಗೇರಿ ಗುಂಪು 06 ವಿಭಾಗ 01 151 nr nr nr nr nr`

**Context Size 4:**

1. `ತೆರದ ಬಾವಿ ಕೊಳವೆ ಬಾವಿಯಿಂದ ನೀರಾವರಿ ಇದ್ದು ಪ್ರಮುಖವಾಗಿ ಕಬ್ಬು ಮೆಕ್ಕೆಜೋಳ ಜೋಳ ಉಳ್ಳಾಗಡ್ಡಿ ಈರುಳ್ಳಿ ನಿಂಬೆಹಣ್ಣು ...`
2. `ದಸರಾ ದೀಪಾವಳಿ ನಾಗರ ಪಂಚಮಿ ಉರಸು ಹಾಗೂ ಮೊಹರಮ್ ಹಬ್ಬಗಳನ್ನು ಆಚರಿಸುತ್ತಾರೆ ಶಿಕ್ಷಣ ಗ್ರಾಮದಲ್ಲಿ ಸರಕಾರಿ ಹಿರಿಯ ಪ್ರಾ...`
3. `ಯುಗಾದಿ ದಸರಾ ದೀಪಾವಳಿ ನಾಗರ ಪಂಚಮಿ ಉರಸು ಹಾಗೂ ಮೊಹರಮ್ ಹಬ್ಬಗಳನ್ನು ಆಚರಿಸುತ್ತಾರೆ ಶಿಕ್ಷಣ ಸರಕಾರಿ ಹಿರಿಯ ಗಂಡು ಮಕ್...`


### Generated Text Samples (Subword-based)

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

**Context Size 1:**

1. `_ಪಯೋಜನ_ಇಂಡಿಯುರೋಹಿಳೆಲ್ಲಿ_(`
2. `ರಲ್_ಪ್ರಾಜಿಲ್ಲಿ_ಅದು_ಏಕಲ್ಲ`
3. `ದ_ಪ್ರಗಗ_ಶ_ಕೈಗಾಗ್ರಂಭಿಸಿದಂ`

**Context Size 2:**

1. `._ಸಾಂಗತ್ತು_ರೋಗಗಳನ್ನು_ಜನರು`
2. `ದ_ಮೊದಲಿನ_ವಿಭವಕ್ಕೆ_ಮೊದಲ_`
3. `_ಅನ್ನಡ_ಶೆಟ್ಟಿ_ಬಿರ್ಲಾಕ್‌ಡೌನ್‌ಗೆ`

**Context Size 3:**

1. `ನ್ನು_ಅನುಷ್ಠಾನ",_"ಪ್ಲಾನೆಟೇರಿಯಂ`
2. `ಲ್ಲಿ_ಯಂತ್ರದಲ್ಲಿ_ಕಾಣಿಸಿಕೊಳುವುದು_`
3. `ತ್ತು_ವೇದಾಂತ,_ದಸರಾ,_ದೀರ್ಘ-ಪ್`

**Context Size 4:**

1. `ಮತ್ತು_ಕರಡು_ಅತ್ಯುತ್ತಮ_ಉದ್ಯಮ`
2. `_ಮತ್ತು_ಹಣಕಾಸಿನ_ಏಕಾಂಶಗಳಾಗಿದ್ದು`
3. `ದಲ್ಲಿ_ಊಹೆ_ಆಗಿದ್ದರೆ,_ರೇಡಿಯೊ_ಕೇಂ`


### 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 (6,451,364 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 | 638,198 |
| Total Tokens | 19,070,152 |
| Mean Frequency | 29.88 |
| Median Frequency | 3 |
| Frequency Std Dev | 738.66 |

### Most Common Words

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ಮತ್ತು | 447,255 |
| 2 | ಈ | 176,896 |
| 3 | ಒಂದು | 110,493 |
| 4 | ಎಂದು | 88,211 |
| 5 | ಅವರು | 84,795 |
| 6 | ಇದು | 76,215 |
| 7 | ಅಥವಾ | 75,251 |
| 8 | ಹಾಗೂ | 66,634 |
| 9 | ಅವರ | 58,212 |
| 10 | ಎಂಬ | 51,061 |

### Least Common Words (from vocabulary)

| Rank | Word | Frequency |
|------|------|-----------|
| 1 | ಆಯಾಮಗಳಾಗಲಿ | 2 |
| 2 | ಮಳೆಪಾತ | 2 |
| 3 | ಬ್ಲಾಕಾಂಗ್ | 2 |
| 4 | ಹೈಪೋಕ್ಸಾಂಥಸ್ | 2 |
| 5 | polbot | 2 |
| 6 | ಚುಬರೋವ್ | 2 |
| 7 | ಥೆರಾವಾಡ | 2 |
| 8 | ಅಮರಸೂರ್ಯ | 2 |
| 9 | ದೇಗಲ್ಡೋರುವಾ | 2 |
| 10 | ಸಲಾಖೈನ್ | 2 |

### Zipf's Law Analysis

| Metric | Value |
|--------|-------|
| Zipf Coefficient | 0.8718 |
| R² (Goodness of Fit) | 0.993113 |
| Adherence Quality | **excellent** |

### Coverage Analysis

| Top N Words | Coverage |
|-------------|----------|
| Top 100 | 15.9% |
| Top 1,000 | 35.7% |
| Top 5,000 | 55.2% |
| Top 10,000 | 64.0% |

### Key Findings

- **Zipf Compliance:** R²=0.9931 indicates excellent adherence to Zipf's law
- **High Frequency Dominance:** Top 100 words cover 15.9% of corpus
- **Long Tail:** 628,198 words needed for remaining 36.0% 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 | 0.7989 | 0.3692 | N/A | N/A |
| **mono_64d** | 64 | 0.6997 | 0.2879 | N/A | N/A |
| **mono_128d** | 128 | 0.6068 | 0.2284 | N/A | N/A |
| **aligned_32d** | 32 | 0.7989 🏆 | 0.3651 | 0.0380 | 0.2320 |
| **aligned_64d** | 64 | 0.6997 | 0.2981 | 0.0820 | 0.3480 |
| **aligned_128d** | 128 | 0.6068 | 0.2150 | 0.1180 | 0.4800 |

### Key Findings

- **Best Isotropy:** aligned_32d with 0.7989 (more uniform distribution)
- **Semantic Density:** Average pairwise similarity of 0.2939. Lower values indicate better semantic separation.
- **Alignment Quality:** Aligned models achieve up to 11.8% 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 | **1.310** | 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 |
|--------|----------|
| `-ಕ` | ಕಾನ್ಫೆಡರೇಷನ್ಸ್, ಕನ್ನೆಧಿರೆ, ಕ್ರೌಲಿ |
| `-ಸ` | ಸೌಂದರಿ, ಸೆನೋರಿಟಾ, ಸಂಗ್‌ |
| `-ಮ` | ಮಲ್ಕಾಪುರ, ಮುಖಾಮುಖಿಯಾಗಿರುವ, ಮೇಲ್ವಿನ್ಯಾಸದಲ್ಲಿ |
| `-ಬ` | ಬಾಯಿಮಾತು, ಬದಲಾಯಿಸಿರುವ, ಬಾಳಾಜಿ |
| `-ಅ` | ಅಭಿನಯಗಳಿಂದಾಗಿ, ಅವಲನ್, ಅಲೇಯ್ನ್ |
| `-ಪ` | ಪ್ರಭೇದನಗಳು, ಪರಿಕರಗಳೆಂದರೆ, ಪ್ರಚೋದನೆಯಿಲ್ಲದೆ |
| `-ನ` | ನ್ಯಾಯವನ್ನು, ನಾಶಮಾಡುವಲ್ಲಿ, ನಿರಾಳವಾಗಿ |
| `-ಹ` | ಹಾಡುಗಳಲ್ಲಿದ್ದ, ಹಂಗೆರಿಯು, ಹಾಯುವುದೇ |

#### Productive Suffixes
| Suffix | Examples |
|--------|----------|
| `-ದ` | ಹಾಡುಗಳಲ್ಲಿದ್ದ, ಸ್ಥಾನಮಾನದ, ಶಬ್ಧವಾದ |
| `-ನ` | ವಿಲ್ಸನ್‍ನ, ಮಣ್ಣಿನಲ್ಲಿನ, ಕೋಲಾರಮ್ಮನ |
| `-ವ` | ಮುಖಾಮುಖಿಯಾಗಿರುವ, ಬದಲಾಯಿಸಿರುವ, ದುಮ್ಮಿಕುವ |
| `-ಯ` | ಮ್ಯಾಲರಿಯ, ಲಿಬ್ಯ, ಏರಿಕೆಯ |
| `-ರ` | ಮಲ್ಕಾಪುರ, ಹಳೆನಗರ, ಅಷ್ಟಲಕ್ಷ್ಮೀಯರ |
| `-ಗಳ` | ಹೋಟೆಲ್‌ಗಳ, ಕಬ್ಬಿಣಗಳ, ಚತುರ್ಥಗಳ |
| `-ಳ` | ಹೋಟೆಲ್‌ಗಳ, ಕಬ್ಬಿಣಗಳ, ಚತುರ್ಥಗಳ |
| `-ಲ` | ಭಾಗಿಯಾಗಲಿಲ್ಲ, ಸಡಿಲಿಸಬಲ್ಲ, ತರಲಾಗಿಲ್ಲ |

### 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 |
|------|----------|------------------|----------|
| `atio` | 3.70x | 37 contexts | ratio, cation, mation |
| `ರಣಗಳ` | 1.49x | 96 contexts | ಮರಣಗಳ, ಚರಣಗಳ, ಕರಣಗಳ |
| `ಕರಣಗ` | 1.55x | 36 contexts | ಕರಣಗಳ, ಕರಣಗಳು, ಕರಣಗಳೂ |

### 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 |
|--------|--------|-----------|----------|
| `-ಪ` | `-ದ` | 56 words | ಪ್ರತ್ಯೇಕವಾದ, ಪ್ರಾಶಸ್ತ್ಯವಿದ್ದ |
| `-ಅ` | `-ದ` | 51 words | ಅಂತರಂಗದಿಂದ, ಅನುಭಾವದ |
| `-ಸ` | `-ದ` | 50 words | ಸ್ಲೊವೇನಿಯಾದ, ಸುಲಭವಾಗಿದ್ದರಿಂದ |
| `-ವ` | `-ದ` | 43 words | ವಿದ್ಯಾರ್ಥಿಗಳಿದ್ದ, ವಿಶ್ವಸೃಷ್ಟಿವಾದದ |
| `-ಮ` | `-ದ` | 40 words | ಮಾವನಾದ, ಮಾಳಿಗೆಗಳಿಂದ |
| `-ಕ` | `-ದ` | 39 words | ಕೋಟೆಯಲ್ಲಿದ್ದ, ಕೂರುತ್ತಿದ್ದ |
| `-ಬ` | `-ದ` | 35 words | ಬೆಳೆಸುವುದರಿಂದ, ಬಾಗಿಲವಾಡದ |
| `-ನ` | `-ದ` | 35 words | ನದೀಮುಖದ, ನ್ಯಾಯಾಧೀಶರುವಾಸ್ತವದ |
| `-ಹ` | `-ದ` | 25 words | ಹಸ್ತಾಕ್ಷರದ, ಹದ್ದುಮೀರಿದ |
| `-ಸ` | `-ಯ` | 24 words | ಸೈನೊಬ್ಯಾಕ್ಟೀರಿಯ, ಸಮಯಾವಧಿಯ |

### 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 |
|------|-----------------|------------|------|
| ಸಿದ್ಧಪಡಿಸುವುದರ | **`ಸಿದ್ಧಪಡಿಸುವು-ದ-ರ`** | 7.5 | `ದ` |
| ಬದಲಾಗುವುದರ | **`ಬದಲಾಗುವು-ದ-ರ`** | 7.5 | `ದ` |
| ವಿದ್ಯುದಾಯಸ್ಕಾಂತದ | **`ವಿದ್ಯುದಾಯಸ್ಕಾಂತ-ದ`** | 4.5 | `ವಿದ್ಯುದಾಯಸ್ಕಾಂತ` |
| ನವೀಕರಿಸುತ್ತಾನೆ | **`ನ-ವ-ೀಕರಿಸುತ್ತಾನೆ`** | 4.5 | `ೀಕರಿಸುತ್ತಾನೆ` |
| ಬ್ರಹ್ಮಣಗಳ | **`ಬ್ರಹ್ಮಣ-ಗಳ`** | 4.5 | `ಬ್ರಹ್ಮಣ` |
| ಉಸಿರಾಟಕ್ಕೂ | **`ಉ-ಸ-ಿರಾಟಕ್ಕೂ`** | 4.5 | `ಿರಾಟಕ್ಕೂ` |
| ಒಣಗಿಸುವಿಕೆಯ | **`ಒಣಗಿಸುವಿಕೆ-ಯ`** | 4.5 | `ಒಣಗಿಸುವಿಕೆ` |
| ವಾಕ್ಚಾತುರ್ಯದ | **`ವಾಕ್ಚಾತುರ್ಯ-ದ`** | 4.5 | `ವಾಕ್ಚಾತುರ್ಯ` |
| ಕತ್ತರಿಸಿಹಾಕಿದ | **`ಕತ್ತರಿಸಿಹಾಕಿ-ದ`** | 4.5 | `ಕತ್ತರಿಸಿಹಾಕಿ` |
| ನೈಟ್‌ಹುಡ್‌ನ | **`ನೈಟ್‌ಹುಡ್‌-ನ`** | 4.5 | `ನೈಟ್‌ಹುಡ್‌` |
| ಕುಮಾರಿಯವರ | **`ಕುಮಾರಿಯ-ವರ`** | 4.5 | `ಕುಮಾರಿಯ` |
| ಸೂರ್ಯಗ್ರಹಣದ | **`ಸೂರ್ಯಗ್ರಹಣ-ದ`** | 4.5 | `ಸೂರ್ಯಗ್ರಹಣ` |
| ಸೆಕೆಂಡುಗಳ | **`ಸೆಕೆಂಡು-ಗಳ`** | 4.5 | `ಸೆಕೆಂಡು` |
| ಸ್ವಾಂತಂತ್ರ್ಯದ | **`ಸ್ವಾಂತಂತ್ರ್ಯ-ದ`** | 4.5 | `ಸ್ವಾಂತಂತ್ರ್ಯ` |
| ಮದ್ಯರಾತ್ರಿ | **`ಮ-ದ-್ಯರಾತ್ರಿ`** | 4.5 | `್ಯರಾತ್ರಿ` |

### 6.6 Linguistic Interpretation

> **Automated Insight:**
The language Kannada shows high morphological productivity. The subword models are significantly more efficient than word models, suggesting a rich system of affixation or compounding.

> **Note on Idiomaticity:** The high Idiomaticity Gap suggests a large number of frequent multi-word expressions or formulaic sequences that are statistically distinct from their component parts.

---
## 7. Summary & Recommendations

![Performance Dashboard](visualizations/performance_dashboard.png)

### Production Recommendations

| Component | Recommended | Rationale |
|-----------|-------------|-----------|
| Tokenizer | **64k BPE** | Best compression (5.01x) |
| N-gram | **2-gram** | Lowest perplexity (2,880) |
| 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: 2026-01-10 11:22:23*