--- language: th language_name: Thai language_family: taikadai_southwestern 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-taikadai_southwestern 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: 4.749 - name: best_isotropy type: isotropy value: 0.8475 - name: vocabulary_size type: vocab value: 0 generated: 2026-01-17 --- # Thai - Wikilangs Models ## Comprehensive Research Report & Full Ablation Study This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Thai** 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** | 3.339x | 3.36 | 0.1132% | 2,229,178 | | **16k** | 3.862x | 3.88 | 0.1309% | 1,927,473 | | **32k** | 4.323x | 4.35 | 0.1466% | 1,722,046 | | **64k** | 4.749x 🏆 | 4.78 | 0.1610% | 1,567,500 | ### Tokenization Examples Below are sample sentences tokenized with each vocabulary size: **Sample 1:** `āđ„āļĨāļŸāđŒāļāļēāļĢāđŒāļ”āđāļŦāđˆāļ‡āļŦāļēāļ”āļšāļ­āļ™āđ„āļ” āđ€āļ›āđ‡āļ™āļŠāļēāļĢāļ„āļ”āļĩāļˆāļēāļāļ­āļ­āļŠāđ€āļ•āļĢāđ€āļĨāļĩāļĒāļ™āļģāđ€āļŠāļ™āļ­āļāļēāļĢāļ—āļģāļ‡āļēāļ™āļ•āļĨāļ­āļ” 24 āļŠāļąāđˆāļ§āđ‚āļĄāļ‡āļ‚āļ­āļ‡āđ„āļĨāļŸ...` | Vocab | Tokens | Count | |-------|--------|-------| | 8k | `▁āđ„āļĨ āļŸāđŒ āļāļēāļĢāđŒ āļ” āđāļŦāđˆāļ‡ āļŦāļēāļ” āļšāļ­āļ™ āđ„āļ” â–āđ€āļ›āđ‡āļ™ āļŠāļēāļĢ ... (+24 more)` | 34 | | 16k | `▁āđ„āļĨ āļŸāđŒ āļāļēāļĢāđŒāļ” āđāļŦāđˆāļ‡ āļŦāļēāļ” āļšāļ­āļ™ āđ„āļ” â–āđ€āļ›āđ‡āļ™āļŠāļēāļĢ āļ„āļ”āļĩ āļˆāļēāļ ... (+19 more)` | 29 | | 32k | `▁āđ„āļĨāļŸāđŒ āļāļēāļĢāđŒāļ” āđāļŦāđˆāļ‡ āļŦāļēāļ” āļšāļ­āļ™ āđ„āļ” â–āđ€āļ›āđ‡āļ™āļŠāļēāļĢ āļ„āļ”āļĩ āļˆāļēāļ āļ­āļ­āļŠāđ€āļ•āļĢ ... (+18 more)` | 28 | | 64k | `▁āđ„āļĨāļŸāđŒ āļāļēāļĢāđŒāļ” āđāļŦāđˆāļ‡ āļŦāļēāļ” āļšāļ­āļ™ āđ„āļ” â–āđ€āļ›āđ‡āļ™āļŠāļēāļĢ āļ„āļ”āļĩ āļˆāļēāļ āļ­āļ­āļŠāđ€āļ•āļĢāđ€āļĨāļĩāļĒāļ™ ... (+17 more)` | 27 | **Sample 2:** `32 āļ­āļēāļˆāļŦāļĄāļēāļĒāļ–āļķāļ‡: 32 (āļ•āļąāļ§āđ€āļĨāļ‚) 32 āļāđˆāļ­āļ™āļ„āļĢāļīāļŠāļ•āļĻāļąāļāļĢāļēāļŠ, 32, āđāļĨāļ°āļ­āļ·āđˆāļ™āđ† 32 (āđ€āļžāļĨāļ‡) ,āđ€āļžāļĨāļ‡āđƒāļ™āļ›āļĩ ...` | Vocab | Tokens | Count | |-------|--------|-------| | 8k | `▁ 3 2 ▁āļ­āļēāļˆāļŦāļĄāļēāļĒāļ–āļķāļ‡ : ▁ 3 2 ▁( āļ•āļąāļ§ ... (+28 more)` | 38 | | 16k | `▁ 3 2 ▁āļ­āļēāļˆāļŦāļĄāļēāļĒāļ–āļķāļ‡ : ▁ 3 2 ▁( āļ•āļąāļ§āđ€āļĨāļ‚ ... (+27 more)` | 37 | | 32k | `▁ 3 2 ▁āļ­āļēāļˆāļŦāļĄāļēāļĒāļ–āļķāļ‡ : ▁ 3 2 ▁( āļ•āļąāļ§āđ€āļĨāļ‚ ... (+25 more)` | 35 | | 64k | `▁ 3 2 ▁āļ­āļēāļˆāļŦāļĄāļēāļĒāļ–āļķāļ‡ : ▁ 3 2 ▁( āļ•āļąāļ§āđ€āļĨāļ‚ ... (+24 more)` | 34 | **Sample 3:** `Molopanthera āđ€āļ›āđ‡āļ™āļŠāļāļļāļĨāļ‚āļ­āļ‡āļžāļ·āļŠāļ”āļ­āļāļ—āļĩāđˆāļ­āļĒāļđāđˆāđƒāļ™āļ§āļ‡āļĻāđŒ Rubiaceae. āļ–āļīāđˆāļ™āļāļģāđ€āļ™āļīāļ”āļ‚āļ­āļ‡āļĄāļąāļ™āļ„āļ·āļ­ āļšāļĢāļēāļ‹āļī...` | Vocab | Tokens | Count | |-------|--------|-------| | 8k | `▁m ol op anth era ▁āđ€āļ›āđ‡āļ™āļŠāļāļļāļĨāļ‚āļ­āļ‡ āļžāļ·āļŠāļ”āļ­āļ āļ—āļĩāđˆāļ­āļĒāļđāđˆāđƒāļ™āļ§āļ‡āļĻāđŒ ▁r ub ... (+24 more)` | 34 | | 16k | `▁mol op anthera ▁āđ€āļ›āđ‡āļ™āļŠāļāļļāļĨāļ‚āļ­āļ‡ āļžāļ·āļŠāļ”āļ­āļ āļ—āļĩāđˆāļ­āļĒāļđāđˆāđƒāļ™āļ§āļ‡āļĻāđŒ ▁rub iaceae . ▁āļ–āļīāđˆāļ™āļāđāļēāđ€āļ™āļīāļ” ... (+17 more)` | 27 | | 32k | `▁mol op anthera ▁āđ€āļ›āđ‡āļ™āļŠāļāļļāļĨāļ‚āļ­āļ‡ āļžāļ·āļŠāļ”āļ­āļ āļ—āļĩāđˆāļ­āļĒāļđāđˆāđƒāļ™āļ§āļ‡āļĻāđŒ ▁rubiaceae . ▁āļ–āļīāđˆāļ™āļāđāļēāđ€āļ™āļīāļ” āļ‚āļ­āļ‡āļĄāļąāļ™āļ„āļ·āļ­ ... (+14 more)` | 24 | | 64k | `▁mol op anthera ▁āđ€āļ›āđ‡āļ™āļŠāļāļļāļĨāļ‚āļ­āļ‡ āļžāļ·āļŠāļ”āļ­āļ āļ—āļĩāđˆāļ­āļĒāļđāđˆāđƒāļ™āļ§āļ‡āļĻāđŒ ▁rubiaceae . ▁āļ–āļīāđˆāļ™āļāđāļēāđ€āļ™āļīāļ” āļ‚āļ­āļ‡āļĄāļąāļ™āļ„āļ·āļ­ ... (+14 more)` | 24 | ### Key Findings - **Best Compression:** 64k achieves 4.749x compression - **Lowest UNK Rate:** 8k with 0.1132% 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 | 56,310 | 15.78 | 475,306 | 16.2% | 28.1% | | **2-gram** | Subword | 2,438 🏆 | 11.25 | 124,885 | 27.9% | 71.1% | | **3-gram** | Word | 160,871 | 17.30 | 713,993 | 10.6% | 19.4% | | **3-gram** | Subword | 27,338 | 14.74 | 1,000,290 | 10.1% | 31.1% | | **4-gram** | Word | 529,813 | 19.02 | 1,376,813 | 3.4% | 10.2% | | **4-gram** | Subword | 174,441 | 17.41 | 4,905,540 | 5.4% | 17.4% | | **5-gram** | Word | 577,241 | 19.14 | 1,093,587 | 2.6% | 7.1% | | **5-gram** | Subword | 676,357 | 19.37 | 11,885,834 | 3.2% | 11.3% | ### Top 5 N-grams by Size **2-grams (Word):** | Rank | N-gram | Count | |------|--------|-------| | 1 | `āļž āļĻ` | 586,670 | | 2 | `āļ„ āļĻ` | 304,560 | | 3 | `āļ­āđ‰āļēāļ‡āļ­āļīāļ‡ āđāļŦāļĨāđˆāļ‡āļ‚āđ‰āļ­āļĄāļđāļĨāļ­āļ·āđˆāļ™` | 46,447 | | 4 | `of the` | 42,755 | | 5 | `āļĻ āļž` | 32,101 | **3-grams (Word):** | Rank | N-gram | Count | |------|--------|-------| | 1 | `āļĻ āļž āļĻ` | 31,957 | | 2 | `āļž āļĻ āļž` | 27,195 | | 3 | `āļĻ āļ„ āļĻ` | 25,879 | | 4 | `āļ˜āļąāļ™āļ§āļēāļ„āļĄ āļž āļĻ` | 21,330 | | 5 | `āļ•āļļāļĨāļēāļ„āļĄ āļž āļĻ` | 21,250 | **4-grams (Word):** | Rank | N-gram | Count | |------|--------|-------| | 1 | `āļž āļĻ āļž āļĻ` | 27,071 | | 2 | `āļž āļĻ āļ„ āļĻ` | 20,164 | | 3 | `0 0 0 0` | 7,943 | | 4 | `āļ„ āļĻ āļ„ āļĻ` | 4,813 | | 5 | `āļ­āđ‰āļēāļ‡āļ­āļīāļ‡ āđāļŦāļĨāđˆāļ‡āļ‚āđ‰āļ­āļĄāļđāļĨāļ­āļ·āđˆāļ™ āļž āļĻ` | 4,336 | **5-grams (Word):** | Rank | N-gram | Count | |------|--------|-------| | 1 | `āļĻ āļž āļĻ āļž āļĻ` | 4,329 | | 2 | `āļž āļĻ āļž āļĻ āļž` | 4,251 | | 3 | `āļĻ āļž āļĻ āļ„ āļĻ` | 3,779 | | 4 | `āļž āļĻ āļž āļĻ āļ„` | 3,510 | | 5 | `0 0 0 0 0` | 3,345 | **2-grams (Subword):** | Rank | N-gram | Count | |------|--------|-------| | 1 | `āļ­ āļ‡` | 3,386,500 | | 2 | `āļē āļĢ` | 3,061,397 | | 3 | `āļ āļē` | 2,892,062 | | 4 | `āļĢ āļ°` | 2,734,121 | | 5 | `āļ™ _` | 2,476,484 | **3-grams (Subword):** | Rank | N-gram | Count | |------|--------|-------| | 1 | `āļ āļē āļĢ` | 2,154,969 | | 2 | `āđ€ āļ›āđ‡ āļ™` | 1,461,135 | | 3 | `āđ āļĨ āļ°` | 1,456,554 | | 4 | `āļ‚ āļ­ āļ‡` | 1,220,921 | | 5 | `āļ› āļĢ āļ°` | 1,178,596 | **4-grams (Subword):** | Rank | N-gram | Count | |------|--------|-------| | 1 | `. āļĻ . _` | 887,086 | | 2 | `_ āđ āļĨ āļ°` | 845,421 | | 3 | `āļž . āļĻ .` | 598,362 | | 4 | `_ āļž . āļĻ` | 554,703 | | 5 | `āļ„ āļ§ āļē āļĄ` | 480,722 | **5-grams (Subword):** | Rank | N-gram | Count | |------|--------|-------| | 1 | `āļž . āļĻ . _` | 572,671 | | 2 | `_ āļž . āļĻ .` | 553,928 | | 3 | `āļ„ . āļĻ . _` | 311,390 | | 4 | `_ āļ„ . āļĻ .` | 268,747 | | 5 | `āļ› āļĢ āļ° āđ€ āļ—` | 260,009 | ### Key Findings - **Best Perplexity:** 2-gram (subword) with 2,438 - **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 | 0.2321 | 1.175 | 2.38 | 8,268,387 | 76.8% | | **1** | Subword | 0.8922 | 1.856 | 12.22 | 37,876 | 10.8% | | **2** | Word | 0.1165 | 1.084 | 1.32 | 19,576,764 | 88.4% | | **2** | Subword | 0.6125 | 1.529 | 5.30 | 462,626 | 38.7% | | **3** | Word | 0.0518 | 1.037 | 1.11 | 25,779,145 | 94.8% | | **3** | Subword | 0.5564 | 1.471 | 3.91 | 2,452,254 | 44.4% | | **4** | Word | 0.0248 🏆 | 1.017 | 1.05 | 28,430,641 | 97.5% | | **4** | Subword | 0.4718 | 1.387 | 2.77 | 9,576,634 | 52.8% | ### Generated Text Samples (Word-based) Below are text samples generated from each word-based Markov chain model: **Context Size 1:** 1. `āļĻ 829 575b 220 āļ™āļąāđ‰āļ™āļĄāļĩāļāļēāļĢāđƒāļŠāđ‰āļāļĢāļ°āļŠāļļāļ™āļ—āļĩāđˆāļˆāļģāļāļąāļ” āļˆāļķāļ‡āļœāļĨāļīāļ•āļ›āļ·āļ™āļĢāļļāđˆāļ™āļ™āļĩāđ‰āļ­āļ­āļāļĄāļē āđāļĨāļ°āļĒāļąāļ‡āļĄāļĩāļĢāļļāđˆāļ™āļĒāđˆāļ­āļĒāļ„āļ·āļ­ āļžāļĩ āļ§āļĩāļŠāđŒāļšāļīāļāđāļ­āļ”āđ€āļ§...` 2. `āļž āļĻ 12 12 34 1 āļžāļĪāļĻāļˆāļīāļāļēāļĒāļ™ āđ€āļĄāļ·āđˆāļ­āļ§āļąāļ™āļ—āļĩāđˆ 16 āļ—āļĩāļĄāļŠāļļāļ”āļ—āđ‰āļēāļĒ 8 āđ€āļ›āđ‡āļ™āļ•āđ‰āļ™āđ„āļ› āļ„ āļĻ āđ€āļ‚āļēāļĒāļąāļ‡āđ„āļ”āđ‰āļ›āļĩāļ™āļ‚āđˆāļēāļ™āđ€āļ—āļ™āļāļĢāļĩ āđƒāļ™āļ›āļĩ` 3. `1 āļˆāļąāļ”āđ€āļ›āđ‡āļ™āļŠāļĩāđˆāļāļĨāļļāđˆāļĄ āļāļĨāļļāđˆāļĄāļĨāļ° 4 9 āļĻāļĢāļĩāļĢāļēāļŠāļē āļ­āļģāđ€āļ āļ­āļĻāļĢāļĩāļĢāļēāļŠāļē āļˆāļąāļ‡āļŦāļ§āļąāļ”āļŠāļĨāļšāļļāļĢāļĩ āđƒāļ™āļ„āļĢāļ­āļšāļ„āļĢāļąāļ§āļ—āļĩāđˆāļĄāļĩāļžāļĩāđˆāļ™āđ‰āļ­āļ‡ 5 āļžāļĪāļĐāļ āļēāļ„āļĄ āļŠāļī...` **Context Size 2:** 1. `āļž āļĻ āđ„āļ”āđ‰āļĢāļąāļšāļ­āļ™āļļāļĄāļąāļ•āļīāļˆāļēāļāļĄāļŦāļēāđ€āļ–āļĢāļŠāļĄāļēāļ„āļĄāđƒāļŦāđ‰āļ›āļĢāļąāļšāļ›āļĢāļļāļ‡āļŠāļ āļēāļžāļ§āļąāļ”āđƒāļŦāđ‰āļ”āļĩāļ‚āļķāđ‰āļ™ āļ›āļĩ āļž āļĻ āļ„ āļĻ āļžāļĢāļ°āđ€āļˆāđ‰āļēāļ­āļīāļŠāļ•āđŒāļ§āļēāļ™āļ—āļĩāđˆ 1 āļ›āļĢāļ°āđ€āļ—āļĻāļŪāļąāļ‡...` 2. `āļ„ āļĻ āļ›āļąāļˆāļˆāļļāļšāļąāļ™ āļĨāļ°āļ„āļĢāļŠāļļāļ” āļ›āļĩāđ€āļĢāļ·āđˆāļ­āļ‡āļšāļ—āļĢāđˆāļ§āļĄāļāļąāļšāļ­āļ­āļāļ­āļēāļāļēāļĻāļ­āđ‰āļēāļ‡āļ­āļīāļ‡āļž āļĻ āļ„āļ§āļēāļĄāļ—āļĢāļ‡āļˆāļģāļ—āļĩāđˆāđ„āļĄāđˆāļ­āļēāļˆāļĨāļ·āļĄ āļ•āļ­āļ™ āļšāļąāļ™āļ—āļķāļāļ—āđˆāļ­āļ‡āđ€āļ—āļĩāđˆāļĒāļ§āļ—...` 3. `of the usaf retrieved 20 october āđ€āļˆāđ‰āļēāļŠāļēāļĒāđ‚āļ—āđ‚āļĄāļŪāļīāđ‚āļ•āļ°āđāļŦāđˆāļ‡āļĄāļīāļāļēāļ‹āļ°āļŠāļīāđ‰āļ™āļžāļĢāļ°āļŠāļ™āļĄāđŒāđ€āļĄāļ·āđˆāļ­āļ§āļąāļ™āļ—āļĩāđˆ 6 āļĄāļīāļ–āļļāļ™āļēāļĒāļ™ āļž āļĻ āļ­āđ‰āļē...` **Context Size 3:** 1. `āļž āļĻ āļž āļĻ āđāļĨāļ°āļ„āļĢāļąāđ‰āļ‡āļ—āļĩāđˆāļŠāļ­āļ‡ āļ›āļĢāļ°āļĄāļēāļ“ āļž āļĻ 31 āļŠāļīāļ‡āļŦāļēāļ„āļĄ āļž āļĻ āļ§āļīāļ—āļĒāļēāļĨāļąāļĒāđ‚āļ—āļĢāļ„āļĄāļ™āļēāļ„āļĄāļ™āļ™āļ—āļšāļļāļĢāļĩ āļĢāļąāļšāļ™āļąāļāļĻāļķāļāļĐāļēāļˆāļēāļ āļāļēāļĢāļŠāļ­āļšāļ„āļąāļ”āđ€āļĨ...` 2. `āļĻ āļž āļĻ āļ„ āļĻ āļžāļĢāļ°āđ€āļˆāđ‰āļēāđāļŸāļĢāđŒāļ”āļĩāļ™āļąāļ™āļ—āđŒāļ—āļĩāđˆ 4 āđāļŦāđˆāļ‡āļŠāļēāļ§āđ‚āļĢāļĄāļąāļ™ 8 āļāļąāļ™āļĒāļēāļĒāļ™ āļ„ āļĻ āđ€āļ›āđ‡āļ™āļ—āļĩāđˆāļĢāļđāđ‰āļˆāļąāļāđƒāļ™āļŠāļ·āđˆāļ­ āđ„āļ›āđ‹ āļĨāļđāđˆ āđ€āļ›āđ‡āļ™āļ™āļąāļāđāļŠāļ”āļ‡...` 3. `āļĻ āļ„ āļĻ āļžāļĢāļ°āļ­āļ‡āļ„āđŒāđ€āļˆāđ‰āļēāļŠāļļāļ§āļžāļąāļāļ•āļĢāđŒāļ§āļīāđ„āļĨāļĒāļžāļĢāļĢāļ“ āļ›āļĢāļ°āļŠāļđāļ•āļī 2 āļžāļĪāļĐāļ āļēāļ„āļĄ āļž āļĻ āļž āļĻ āđ„āļ­ āļˆāļĩ āļ„āļ­āļĄāļĄāļīāļ§āļ™āļīāđ€āļ„āļŠāļąāļ™` **Context Size 4:** 1. `āļž āļĻ āļž āļĻ āđ€āļ›āđ‡āļ™āļ›āļĢāļēāļŠāļāđŒāđāļĨāļ°āļ™āļąāļāļ„āļīāļ”āļ­āļīāļŠāļĢāļ°āļ—āļĩāđˆāđ„āļ”āđ‰āļĢāļąāļšāļāļēāļĢāļāļĨāđˆāļēāļ§āļ–āļķāļ‡āļ­āļĒāđˆāļēāļ‡āļāļ§āđ‰āļēāļ‡āļ‚āļ§āļēāļ‡ āđ€āļĄāļ·āđˆāļ­āļĒāļąāļ‡āđ€āļ›āđ‡āļ™āđ€āļ”āđ‡āļāļŦāļ™āļļāđˆāļĄ āļ—āļēāļ‡āļŠāļĄāļēāļ„āļĄāđ€āļ—āļ§...` 2. `āļž āļĻ āļ„ āļĻ āļ”āļĩāđāļĨāļ™ āļĄāļīāļ™āđ€āļ™āđ‡āļ•āļ•āđŒ āļ™āļąāļāđāļŠāļ”āļ‡āđāļĨāļ°āļ™āļąāļāļ”āļ™āļ•āļĢāļĩāļŠāļēāļ§āļ­āđ€āļĄāļĢāļīāļāļąāļ™ āļ›āļĢāļīāļ™āļ‹āđŒ āđāļ­āļĄāļžāļ­āļ™āļ‹āļē āļ™āļąāļāļŸāļļāļ•āļšāļ­āļĨāļŠāļēāļ§āļāļēāļ™āļē āļ‹āļēāļ™āļ° āļĄāļīāļ™āļēāđ‚āļ•āļ‹āļē...` 3. `0 0 0 0 āđ„āļĄāđ„āļ”āđ‰āđ€āļ‚āđ‰āļēāļĢāđˆāļ§āļĄāđāļ‚āđˆāļ‡āļ‚āļąāļ™ 4 0 0 0 0 4 21 17 4 26 4 āļĢāļ°āļ™āļ­āļ‡ 16 6` ### Generated Text Samples (Subword-based) Below are text samples generated from each subword-based Markov chain model: **Context Size 1:** 1. `_āļ„āļĒāļ‡āļ‚_āđƒāļ™āļ—āļĩāđˆāđ€āļ›āđ‡āļ™āļāļ­āļ§āđˆāļē` 2. `āļēāļĢāļ•āļĩāļĒāļ™āļœāļīāļ”āļĨāļ°āļ„āļēāļĢāļ°āļšāđ‚āļ”` 3. `āļ™āļēāļ“āļ‘āļĨāļēāļĢāļ°āļ›āļĒāļĩāļ›āļ”āļēāļĢāļĢāļ™` **Context Size 2:** 1. `āļ­āļ‡āļ„āļĢāļąāđ‰āļ‡āļ—āļĩāđˆāļĢāļđāđ‰āļˆāļąāļāļĢāļ°āļŠāļ™āļīāļ”āļ™āļķāļ‡āļ™āļī` 2. `āļēāļĢāđ„āļ›āļŠ.āļ˜.90.0_āļ‚āļ­āļ‡āđ€` 3. `āļāļēāļĢāļŠāļŠ.āļ­āļĩāļāļ„āļĢāļąāđ‰āļ‡āļĄāļĩāļ§āđˆāļē_āđ€āļ›` **Context Size 3:** 1. `āļāļēāļĢāđƒāļ”_āđ†_broad_mete` 2. `āđ€āļ›āđ‡āļ™āļāļēāļĢāđƒāļŠāđ‰āļ­āļĒāđˆāļēāļ‡āļ•āđˆāļēāļ‡āļˆāļēāļ` 3. `āđāļĨāļ°āļ„āļĢāļąāđ‰āļ‡āđāļĢāļĄāļŠāļēāļ•āļīāļ‚āļķāđ‰āļ™āļšāļ_āđ` **Context Size 4:** 1. `.āļĻ._āļŠāļŦāļĢāļēāļŠāļ­āļēāļ“āļēāđ€āļ‚āļ•āļˆāļ•āļļāļˆāļą` 2. `_āđāļĨāļ°āđ„āļĄāđˆāļŠāļēāļĄāļēāļĢāļ–āļ›āđ‰āļ­āļ‡āļāļąāļ™āđāļĨ` 3. `āļž.āļĻ._76_(āđ„āļ—āļĒ` ### Key Findings - **Best Predictability:** Context-4 (word) with 97.5% predictability - **Branching Factor:** Decreases with context size (more deterministic) - **Memory Trade-off:** Larger contexts require more storage (9,576,634 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 | 1,276,542 | | Total Tokens | 26,332,909 | | Mean Frequency | 20.63 | | Median Frequency | 3 | | Frequency Std Dev | 1261.31 | ### Most Common Words | Rank | Word | Frequency | |------|------|-----------| | 1 | āļĻ | 920,143 | | 2 | āļž | 595,465 | | 3 | 1 | 351,475 | | 4 | āļ„ | 314,624 | | 5 | 2 | 306,676 | | 6 | 3 | 247,910 | | 7 | the | 217,279 | | 8 | 4 | 172,069 | | 9 | āđ† | 171,685 | | 10 | of | 169,227 | ### Least Common Words (from vocabulary) | Rank | Word | Frequency | |------|------|-----------| | 1 | āđ€āļšāļ­āļĢāđŒāļĨāļīāļ™āļŸāļēāļĢāđŒāļĄāļēāļ‹āļđāļ•āļīāļ„āļ­āļĨāļ­āļīāļ™āļ”āļąāļŠāļ•āļĢāļĩāđ‰ | 2 | | 2 | āļĄāļĩāđ€āļĨāļ‚āļ‹āļĩāđ€āļ—āļ™āļ‚āļąāđ‰āļ™āļ•āđˆāļģāļ—āļĩāđˆ | 2 | | 3 | āļ™āđ‰āļģāļĄāļąāļ™āļ”āļĩāđ€āļ‹āļĨāļŦāļĄāļļāļ™āđ€āļ§āļĩāļĒāļ™ | 2 | | 4 | neste | 2 | | 5 | āđ€āļŪāļāļ‹āļēāļ”āļĩāđ€āļ„āļ™ | 2 | | 6 | āđ€āļŪāļ›āļ•āļēāđ€āļĄāļ—āļīāļĨāđ‚āļ™āđ€āļ™āļ™ | 2 | | 7 | āđ€āļ„āļĢāļ·āđˆāļ­āļ‡āļ—āļ”āļŠāļ­āļšāļ„āļļāļ“āļ āļēāļžāļāļēāļĢāļˆāļļāļ”āļĢāļ°āđ€āļšāļīāļ” | 2 | | 8 | āđ€āļ„āļĢāļ·āđˆāļ­āļ‡āļĄāļ·āļ­āļ™āļĩāđ‰āđƒāļŠāđ‰āļ§āļīāļ˜āļĩāđ€āļĢāļĩāļĒāļšāļ‡āđˆāļēāļĒāļāļ§āđˆāļēāđāļĨāļ°āđāļ‚āđ‡āļ‡āđāļāļĢāđˆāļ‡āļāļ§āđˆāļēāđƒāļ™āļāļēāļĢāļ§āļąāļ”āđ€āļĨāļ‚āļ‹āļĩāđ€āļ—āļ™āđ€āļĄāļ·āđˆāļ­āđ€āļ—āļĩāļĒāļšāļāļąāļš | 2 | | 9 | āļšāđ‰āļēāļ™āđ€āļāđ‰āļēāđ€āļĨāļĩāđ‰āļĒāļ§ | 2 | | 10 | āļŠāļļāļĄāļŠāļ™āļšāđ‰āļēāļ™āđ€āļāđ‰āļēāđ€āļĨāļĩāđ‰āļĒāļ§ | 2 | ### Zipf's Law Analysis | Metric | Value | |--------|-------| | Zipf Coefficient | 0.9360 | | RÂē (Goodness of Fit) | 0.999043 | | Adherence Quality | **excellent** | ### Coverage Analysis | Top N Words | Coverage | |-------------|----------| | Top 100 | 29.7% | | Top 1,000 | 45.1% | | Top 5,000 | 57.6% | | Top 10,000 | 63.4% | ### Key Findings - **Zipf Compliance:** RÂē=0.9990 indicates excellent adherence to Zipf's law - **High Frequency Dominance:** Top 100 words cover 29.7% of corpus - **Long Tail:** 1,266,542 words needed for remaining 36.6% 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 | 0.8475 | 0.3288 | N/A | N/A | | **mono_64d** | 64 | 0.8400 | 0.2631 | N/A | N/A | | **mono_128d** | 128 | 0.8225 | 0.1868 | N/A | N/A | | **aligned_32d** | 32 | 0.8475 🏆 | 0.3296 | 0.2180 | 0.6440 | | **aligned_64d** | 64 | 0.8400 | 0.2600 | 0.4200 | 0.7840 | | **aligned_128d** | 128 | 0.8225 | 0.1907 | 0.4680 | 0.8680 | ### Key Findings - **Best Isotropy:** aligned_32d with 0.8475 (more uniform distribution) - **Semantic Density:** Average pairwise similarity of 0.2598. Lower values indicate better semantic separation. - **Alignment Quality:** Aligned models achieve up to 46.8% 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 | **-0.367** | 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 | |--------|----------| | `-āđāļĨāļ°` | āđāļĨāļ°āđ„āļ”āđ‰āļĢāļąāļšāđāļ•āđˆāļ‡āļ•āļąāđ‰āļ‡āđ€āļ›āđ‡āļ™āļĢāļąāļāļĄāļ™āļ•āļĢāļĩāļ—āļĩāđˆāđ„āļĄāđˆāļ›āļĢāļ°āļˆāļģāļāļĢāļ°āļ—āļĢāļ§āļ‡, āđāļĨāļ°āļžāļēāļ—āļĩāļĄāđ€āļ‚āđ‰āļēāļŠāļīāļ‡āļŠāļ™āļ°āđ€āļĨāļīāļĻ, āđāļĨāļ°āļ„āļĢāļīāļŠāļ•āđŒāļĻāļēāļŠāļ™āļē | | `-āđ€` | āđ€āļˆāđ‰āļēāđ€āļĄāļ·āļ­āļ‡āļ™āļ„āļĢāđ€āļ‚āļ·āđˆāļ­āļ™āļ‚āļąāļ™āļ˜āđŒāļ„āļ™āļ—āļĩāđˆ, āđ€āļ‰āļīāļ‡āđ€āļ•āļĩāđ‹āļĒāļ­āļĩ, āđ€āļ§āļĨāļ§āļīāļŠāđ€āļ‹āļĩāļĒ | | `-āđ‚` | āđ‚āļĢāļ‡āđ€āļĢāļĩāļĒāļ™āļžāļĢāļ°āļ›āļāļĄāļ§āļīāļ—āļĒāļēāļĨāļąāļĒ, āđ‚āļĢāļ‡āđ€āļĢāļĩāļĒāļ™āđ€āļ§āļĩāļĒāļ‡āļāļēāļŦāļĨāļ‡āļ§āļīāļ—āļĒāļē, āđ‚āļ­āļĄāļĄāļ­āļ | | `-āđ` | āđāļĄāđˆāđ€āļĄāđ‡āļ‡, āđāļĨāļ°āđ„āļ”āđ‰āļĢāļąāļšāđāļ•āđˆāļ‡āļ•āļąāđ‰āļ‡āđ€āļ›āđ‡āļ™āļĢāļąāļāļĄāļ™āļ•āļĢāļĩāļ—āļĩāđˆāđ„āļĄāđˆāļ›āļĢāļ°āļˆāļģāļāļĢāļ°āļ—āļĢāļ§āļ‡, āđāļĨāļ°āļžāļēāļ—āļĩāļĄāđ€āļ‚āđ‰āļēāļŠāļīāļ‡āļŠāļ™āļ°āđ€āļĨāļīāļĻ | | `-āļ­` | āļ­āļ„āļ§āļēāļ„āļĢāļ­āļŠ, āļ­āļģāđ€āļ āļ­āļ—āļļāđˆāļ‡āļĒāļēāļ‡āđāļ”āļ‡, āļ­āļļāļ”āļĄāļĻāļīāļĨāļ›āđŒ | | `-āļŠ` | āļŠāļĄāđ€āļ”āđ‡āļˆāļžāļĢāļ°āļŠāļąāļ™āļ•āļ°āļ›āļēāļ›āļēāļ˜āļĩāđ‚āļ­āļ”āļ­āļĢāđŒāļ—āļĩāđˆ, āļŠāļ–āļīāļĢāļ§āđāđ‚āļŠ, āļŠāļēāļ‚āļēāļ§āļīāļŠāļēāļĻāļīāļĨāļ›āļāļĢāļĢāļĄ | | `-āļ` | āļāļēāļĢāļĨāļ­āļšāļ†āđˆāļē, āļāļēāļāļˆāļ™āļžāļąāļ’āļ™āđŒ, āļāļēāļĢāļ›āļĢāļ°āļĄāļđāļĨāļ„āļĨāļ·āđˆāļ™ | | `-āļ„` | āļ„āļīāļ„āļēāļ§āļēāļ”āļ°, āļ„āļīāļ”āļ›āļĢāļ°āđ€āļŠāļĢāļīāļ, āļ„āļĢāļ­āļšāļ„āļĨāļļāļĄāļžāļ·āđ‰āļ™āļ—āļĩāđˆāļ•āļģāļšāļĨāļŠāļ‡āļĒāļēāļ‡āļ—āļąāđ‰āļ‡āļ•āļģāļšāļĨ | #### Productive Suffixes | Suffix | Examples | |--------|----------| | `-āļ™` | āļ—āļĩāđˆāđ‚āļ”āđˆāļ‡āļ”āļąāļ‡āđ€āļŠāđˆāļ™, āļ•āļąāđ‰āļ‡āļ­āļĒāļđāđˆāļšāļ™āđ€āļŠāļīāļ‡āđ€āļ—āļīāļ™, āļ™āļąāļāđ€āļĢāļĩāļĒāļ™āļŦāđ‰āļ­āļ‡āļāļīāļŸāļ•āđŒāļĢāļļāđˆāļ™ | | `-āļ‡` | āđāļĄāđˆāđ€āļĄāđ‡āļ‡, āđāļĨāļ°āđ„āļ”āđ‰āļĢāļąāļšāđāļ•āđˆāļ‡āļ•āļąāđ‰āļ‡āđ€āļ›āđ‡āļ™āļĢāļąāļāļĄāļ™āļ•āļĢāļĩāļ—āļĩāđˆāđ„āļĄāđˆāļ›āļĢāļ°āļˆāļģāļāļĢāļ°āļ—āļĢāļ§āļ‡, āļ­āļģāđ€āļ āļ­āļ—āļļāđˆāļ‡āļĒāļēāļ‡āđāļ”āļ‡ | | `-āļē` | āļāļēāļĢāļĨāļ­āļšāļ†āđˆāļē, āđ‚āļĢāļ‡āđ€āļĢāļĩāļĒāļ™āđ€āļ§āļĩāļĒāļ‡āļāļēāļŦāļĨāļ‡āļ§āļīāļ—āļĒāļē, āļ­āļ”āļĩāļ•āļ™āļēāļĒāļāļĢāļąāļāļĄāļ™āļ•āļĢāļĩāđāļ„āļ™āļēāļ”āļē | | `-āļĒ` | āļ™āļģāļŠāļąāļĒ, āđ‚āļĢāļ‡āđ€āļĢāļĩāļĒāļ™āļžāļĢāļ°āļ›āļāļĄāļ§āļīāļ—āļĒāļēāļĨāļąāļĒ, āļŠāļĩāļ§āļīāļ•āļŠāđˆāļ§āļ‡āļ›āļĨāļēāļĒ | | `-āļ` | āļĨāđ‰āļ­āļĄāļĢāļ­āļšāļ”āđ‰āļ§āļĒāļāļ™āļāđ€āļ›āļĨāļ§āđ€āļžāļĨāļīāļ‡āļ”āđ‰āļēāļ™āļšāļ™āļĄāļĩāļĢāļąāļĻāļĄāļĩāļ›āļĢāļ°āļāļ­āļšāđāļ›āļ”āđāļ‰āļ, āđ‚āļ­āļĄāļĄāļ­āļ, āđ€āļ™āļ·āđˆāļ­āļ‡āđƒāļ™āđ‚āļ­āļāļēāļŠāļžāļĢāļ°āļĢāļēāļŠāļžāļīāļ˜āļĩāļāļēāļāļˆāļ™āļēāļ āļīāđ€āļĐāļ | | `-āļĄ` | āļŦāļīāļĄāļēāļĨāļąāļĒāļĒāļīāļĄ, āļĄāļ­āļšāđ‚āļ”āļĒāļāļĢāļ°āļ—āļĢāļ§āļ‡āļ§āļąāļ’āļ™āļ˜āļĢāļĢāļĄ, āļŠāļēāļ‚āļēāļ§āļīāļŠāļēāļĻāļīāļĨāļ›āļāļĢāļĢāļĄ | | `-āļ­āļ‡` | āđāļĨāļ°āđƒāļ™āļ„āļ·āļ™āļ™āļąāđ‰āļ™āđ€āļ­āļ‡, āļāļēāđ€āļ­āļ•āđ‡āļ­āļ‡, āļ™āļąāļāđāļŠāļ”āļ‡āļˆāļēāļāđ€āļĢāļ·āđˆāļ­āļ‡ | | `-āļĢ` | āļŦāļ‡āļĐāđŒāļ‚āļˆāļĢ, āļˆāļ°āļžāļđāļ”āļ–āļķāļ‡āļ„āļ§āļēāļĄāļĢāļđāđ‰āļŠāļķāļāđ€āļŠāļ·āđˆāļ­āđƒāļˆāđ„āļ”āđ‰āļ­āļĒāđˆāļēāļ‡āđ„āļĢ, āđāļĨāļ°āļ™āļēāļĒāļ—āļŦāļēāļĢ | ### 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 | |------|----------|------------------|----------| | `āļāļēāļĢāđ` | 2.17x | 65 contexts | āļāļēāļĢāđāļĨāļ, āļāļēāļĢāđāļ›āļĨ, āļāļēāļĢāđāļ•āļ | | `āļ‚āļ­āļ‡āđ€` | 1.49x | 196 contexts | āļ‚āļ­āļ‡āđ€āļĨ, āļ‚āļ­āļ‡āđ€āļˆ, āļ‚āļ­āļ‡āđ€āļ­ | | `āļžāļĢāļ°āļĢ` | 2.07x | 33 contexts | āļžāļĢāļ°āļĢāļ–, āļžāļĢāļ°āļĢāļēāļĄ, āļžāļĢāļ°āļĢāļēāļŠ | | `āļāļēāļĢāđ€` | 1.55x | 93 contexts | āļāļēāļĢāđ€āļĒ, āļāļēāļĢāđ€āļ”āļ—, āļāļēāļĢāđ€āļ­āļē | | `āļĻāļēāļŠāļ•` | 1.82x | 46 contexts | āļĻāļēāļŠāļ•āļē, āļĻāļēāļŠāļ•āļĢāļē, āļĻāļēāļŠāļ•āļĢāđŒ | | `āļēāļāļēāļĢ` | 1.45x | 100 contexts | āļ­āļēāļāļēāļĢ, āļšāļēāļāļēāļĢāļĩ, āļ„āļēāļāļēāļĢāļī | | `āļ™āļāļēāļĢ` | 1.48x | 86 contexts | āļ˜āļ™āļāļēāļĢ, āđƒāļ™āļāļēāļĢ, āđāļœāļ™āļāļēāļĢ | | `āļ›āļĢāļ°āļ` | 1.46x | 84 contexts | āļ›āļĢāļ°āļāļš, āļ›āļĢāļ°āļāļēāļĢ, āļ›āļĢāļ°āļāļīāļˆ | | `āđ‚āļĢāļ‡āđ€` | 2.92x | 8 contexts | āđ‚āļĢāļ‡āđ€āļˆ, āđ‚āļĢāļ‡āđ€āļ‚āđ‰, āđ‚āļĢāļ‡āđ€āļĢāļĩāļĒ | | `āļ›āļĢāļ°āđ€` | 1.42x | 83 contexts | āļ›āļĢāļ°āđ€āļ–āļ—, āļ›āļĢāļ°āđ€āļ āļ—, āļ›āļĢāļ°āđ€āļ—āļĻ | | `āļ‡āļˆāļēāļ` | 1.44x | 72 contexts | āļšāļēāļ‡āļˆāļēāļ, āļ­āļīāļ‡āļˆāļēāļ, āļ—āļēāļ‡āļˆāļēāļ | | `āļĢāļ°āđ€āļ—` | 1.66x | 38 contexts | āļāļĢāļ°āđ€āļ—āļĒ, āļāļĢāļ°āđ€āļ—āļĻ, āļžāļĢāļ°āđ€āļ—āļž | ### 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 | |--------|--------|-----------|----------| | `-āđ€` | `-āļ™` | 101 words | āđ€āļžāļ·āđˆāļ­āđ€āļ›āđ‡āļ™āļāļēāļĢāļĢāļąāļāļĐāļēāļāļģāļĨāļąāļ‡āđāļĨāļ°āđ„āļžāļĢāđˆāļžāļĨāļ—āļŦāļēāļĢāļ‚āļ­āļ‡āļ•āļ™āđ€āļ­āļ‡āđ„āļ§āđ‰āļŠāļģāļŦāļĢāļąāļšāļāļēāļĢāļĻāļķāļāļ­āļ·āđˆāļ™, āđ€āļŦāļĒāļēāļˆāļ·āđˆāļ­āļˆāļīāļ™ | | `-āđ€` | `-āļ‡` | 84 words | āđ€āļ˜āļ­āđ„āļ”āđ‰āļ­āļ­āļāđ€āļžāļĨāļ‡, āđ€āļˆāļīāđ‰āļ™āđ€āļŠāļĩāļĒāļ‡ | | `-āđ€` | `-āļē` | 80 words | āđ€āļ‚āđ‡āļĄāļ‚āđ‰āļēāļŦāļĨāļ§āļ‡āđ€āļ”āļīāļĄāļĢāļēāļŠāļāļīāļˆāļˆāļēāļ™āļļāđ€āļšāļāļĐāļē, āđ€āļˆāđ‰āļēāļŦāļāļīāļ‡āđ‚āļĢāļĄāļēāļ™āļ­āļŸāļŠāļāļēāļĒāļē | | `-āđ€` | `-āļĒ` | 53 words | āđ€āļ›āđ‡āļ™āļ āļēāļĐāļēāđ„āļ—āļĒāļ­āļĩāļāļ”āđ‰āļ§āļĒ, āđ€āļžāļĨāļ‡āļ”āļēāļšāđāļĄāđˆāļ™āđ‰āļģāļĢāđ‰āļ­āļĒāļŠāļēāļĒ | | `-āđāļĨāļ°` | `-āļ™` | 52 words | āđāļĨāļ°āļ•āļģāļšāļĨāļšāđ‰āļēāļ™āđāļŦāļ§āļ™, āđāļĨāļ°āđ€āļ›āļĨāļĩāđˆāļĒāļ™āļŠāļ·āđˆāļ­āđ„āļ›āđ€āļ›āđ‡āļ™ | | `-āđ‚` | `-āļ‡` | 50 words | āđ‚āļ”āļĒāđƒāļŠāđ‰āđ€āļ„āļĢāļ·āđˆāļ­āļ‡āļšāļīāļ™āđ‚āļšāļ­āļīāļ‡, āđ‚āļĢāļ‡āđ€āļĢāļĩāļĒāļ™āļ•āļ°āđ‚āļāļ”āļ­āļ™āļŦāļāđ‰āļēāļ™āļēāļ‡ | | `-āđ‚` | `-āļ™` | 45 words | āđ‚āļ„āđ€āļŪāđ‡āļ™, āđ‚āļŸāļāļŠāđŒāļ§āļēāđ€āļāļ™ | | `-āļ` | `-āļ™` | 44 words | āļāļļāļĨāļ˜āļ™, āļāļēāļĢāđāļšāđˆāļ‡āļŠāļ™āļŠāļąāđ‰āļ™ | | `-āđ‚` | `-āļē` | 42 words | āđ‚āļĢāļ‡āđ€āļĢāļĩāļĒāļ™āļĻāļĢāļĩāļŠāļĄāļšāļđāļĢāļ“āđŒāļ§āļīāļ—āļĒāļē, āđ‚āļ”āļĒāļĄāļĩāļ§āļąāļ•āļ–āļļāļ›āļĢāļ°āļŠāļ‡āļ„āđŒāđ€āļžāļ·āđˆāļ­āđ€āļ›āđ‡āļ™āļŠāļ–āļēāļšāļąāļ™āļāļēāļĢāļĻāļķāļāļĐāļē | | `-āđ` | `-āļ™` | 41 words | āđāļĨāļ°āļ•āļģāļšāļĨāļšāđ‰āļēāļ™āđāļŦāļ§āļ™, āđāļĄāđˆāļŪāđˆāļ­āļ‡āļŠāļ­āļ™ | ### 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 | |------|-----------------|------------|------| | āļāļēāļĢāļ—āļģāļ„āļ°āđāļ™āļ™ | **`āļāļēāļĢāļ—āļģāļ„āļ°āđ-āļ™-āļ™`** | 7.5 | `āļ™` | | āđ€āļˆāđ‰āļēāļˆāļ­āļĄāļŠāļļāļ§āļąāļ—āļ™āļē | **`āđ€āļˆāđ‰āļēāļˆāļ­āļĄāļŠāļļāļ§āļąāļ—-āļ™-āļē`** | 7.5 | `āļ™` | | āļŦāļąāļ§āļ­āļāļŠāļēāļ§āļ™āļē | **`āļŦāļąāļ§āļ­āļāļŠāļēāļ§-āļ™-āļē`** | 7.5 | `āļ™` | | āļ­āļīāļĢāļīāļĒāļēāļ›āļ–āļšāļĢāļĢāļž | **`āļ­āļīāļĢāļīāļĒāļēāļ›āļ–āļšāļĢ-āļĢ-āļž`** | 7.5 | `āļĢ` | | āļ•āļģāļšāļĨāļžāļĨāļ§āļ‡āļŠāļ­āļ‡āļ™āļēāļ‡ | **`āļ•āļģāļšāļĨāļžāļĨāļ§āļ‡āļŠāļ­āļ‡-āļ™-āļēāļ‡`** | 7.5 | `āļ™` | | āđ€āļŠāđ‰āļ™āļ—āļēāļ‡āļ—āļĢāļ™āļ‡ | **`āđ€āļŠāđ‰āļ™āļ—āļēāļ‡āļ—āļĢ-āļ™-āļ‡`** | 7.5 | `āļ™` | | āļ­āđ€āļĨāđ‡āļāļ‹āļēāļ™āļ”āļĢāļ­āļŸāļ™āļē | **`āļ­āđ€āļĨāđ‡āļāļ‹āļēāļ™āļ”āļĢāļ­āļŸ-āļ™-āļē`** | 7.5 | `āļ™` | | āļ„āļēāļˆāļīāđ‚āļ”āļāļīāļ­āļēāļĢāđŒāļĄāļŠ | **`āļ„āļēāļˆāļīāđ‚āļ”āļāļīāļ­āļēāļĢāđŒ-āļĄ-āļŠ`** | 7.5 | `āļĄ` | | āđāļĨāļ°āđ€āļ‹āđ€āļĢāļ™āļē | **`āđāļĨāļ°āđ€āļ‹āđ€āļĢ-āļ™-āļē`** | 7.5 | `āļ™` | | āđāļĨāļ°āļ„āļđāļĨāļĨāļīāđāļ™āļ™ | **`āđāļĨāļ°āļ„āļđāļĨāļĨāļīāđ-āļ™-āļ™`** | 7.5 | `āļ™` | | āļŸāļļāļŠāļīāļāļīāļ”āļēāđ€āļ™āļ° | **`āļŸāļļāļŠāļīāļāļīāļ”āļēāđ€-āļ™-āļ°`** | 7.5 | `āļ™` | | āļ•āļģāļšāļĨāļĄāđˆāļ§āļ‡āļ‡āļēāļĄ | **`āļ•āļģāļšāļĨāļĄāđˆāļ§āļ‡-āļ‡-āļēāļĄ`** | 6.0 | `āļ•āļģāļšāļĨāļĄāđˆāļ§āļ‡` | | āđāļĨāļ°āļ•āļģāļšāļĨāļŦāļĄāļ·āđˆāļ™āđ„āļ§āļĒ | **`āđāļĨāļ°-āļ•āļģāļšāļĨāļŦāļĄāļ·āđˆāļ™āđ„āļ§āļĒ`** | 4.5 | `āļ•āļģāļšāļĨāļŦāļĄāļ·āđˆāļ™āđ„āļ§āļĒ` | | āđāļĨāļ°āđ„āļ”āđ‰āļĢāļąāļšāļŠāļĄāļāļēāļ§āđˆāļē | **`āđāļĨāļ°-āđ„āļ”āđ‰āļĢāļąāļšāļŠāļĄāļāļēāļ§āđˆāļē`** | 4.5 | `āđ„āļ”āđ‰āļĢāļąāļšāļŠāļĄāļāļēāļ§āđˆāļē` | | āđāļĨāļ°āļ›āļĢāļ°āļ—āļąāļšāļ­āļĒāļđāđˆ | **`āđāļĨāļ°-āļ›āļĢāļ°āļ—āļąāļšāļ­āļĒāļđāđˆ`** | 4.5 | `āļ›āļĢāļ°āļ—āļąāļšāļ­āļĒāļđāđˆ` | ### 6.6 Linguistic Interpretation > **Automated Insight:** The language Thai shows high morphological productivity. The subword models are significantly more efficient than word models, suggesting a rich system of affixation or compounding. --- ## 7. Summary & Recommendations ![Performance Dashboard](visualizations/performance_dashboard.png) ### Production Recommendations | Component | Recommended | Rationale | |-----------|-------------|-----------| | Tokenizer | **64k BPE** | Best compression (4.75x) | | N-gram | **2-gram** | Lowest perplexity (2,438) | | Markov | **Context-4** | Highest predictability (97.5%) | | 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. **RÂē (Coefficient of Determination)** > *Definition:* Measures how well the linear fit explains the frequency-rank relationship. Ranges from 0 to 1. > > *Intuition:* RÂē near 1.0 means the data closely follows Zipf's law; lower values indicate deviation from expected word frequency patterns. > > *What to seek:* RÂē > 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) - ðŸĪ— Models: [huggingface.co/wikilangs](https://huggingface.co/wikilangs) - 📊 Data: [wikipedia-monthly](https://huggingface.co/datasets/omarkamali/wikipedia-monthly) - ðŸ‘Ī Author: [Omar Kamali](https://huggingface.co/omarkamali) - ðŸĪ Sponsor: [Featherless AI](https://featherless.ai) --- *Generated by Wikilangs Models Pipeline* *Report Date: 2026-01-17 15:56:15*