Upload all models and assets for chr (20251001)

README.md

@@ -23,14 +23,14 @@ dataset_info:
 metrics:
   - name: best_compression_ratio
     type: compression
-    value: 3.838
+    value: 3.573
   - name: best_isotropy
     type: isotropy
-    value: 0.2339
+    value: 0.2005
   - name: vocabulary_size
     type: vocab
-    value: 4475
-generated: 2025-12-28
+    value: 0
+generated: 2026-01-03
 ---
 
 # CHR - Wikilangs Models
@@ -44,12 +44,13 @@ We analyze tokenizers, n-gram models, Markov chains, vocabulary statistics, and
 ### Models & Assets
 
 - Tokenizers (8k, 16k, 32k, 64k)
-- N-gram models (2, 3, 4-gram)
-- Markov chains (context of 1, 2, 3 and 4)
+- 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
+- Embeddings in various sizes and dimensions (aligned and unaligned)
 - Language Vocabulary
 - Language Statistics
+
 ![Performance Dashboard](visualizations/performance_dashboard.png)
 
 ### Analysis and Evaluation
@@ -59,7 +60,8 @@ We analyze tokenizers, n-gram models, Markov chains, vocabulary statistics, and
 - [3. Markov Chain Evaluation](#3-markov-chain-evaluation)
 - [4. Vocabulary Analysis](#4-vocabulary-analysis)
 - [5. Word Embeddings Evaluation](#5-word-embeddings-evaluation)
-- [6. Summary & Recommendations](#6-summary--recommendations)
+- [6. Morphological Analysis (Experimental)](#6-morphological-analysis)
+- [7. Summary & Recommendations](#7-summary--recommendations)
 - [Metrics Glossary](#appendix-metrics-glossary--interpretation-guide)
 - [Visualizations Index](#visualizations-index)
 
@@ -68,58 +70,53 @@ We analyze tokenizers, n-gram models, Markov chains, vocabulary statistics, and
 
 ![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** | 3.087x | 2.96 | 0.1137% | 95,883 |
-| **16k** | 3.454x | 3.32 | 0.1272% | 85,691 |
-| **32k** | 3.838x 🏆 | 3.68 | 0.1414% | 77,111 |
+| **8k** | 2.938x | 2.95 | 0.1472% | 83,560 |
+| **16k** | 3.361x | 3.37 | 0.1684% | 73,046 |
+| **32k** | 3.573x 🏆 | 3.59 | 0.1790% | 68,709 |
 
 ### Tokenization Examples
 
 Below are sample sentences tokenized with each vocabulary size:
 
-**Sample 1:** `ᏍᎪᏯ ᏗᏃᎷᏩᏘᏍᎩ"Consortium Word List." 2016-04-15. (sgoya dinoluwatisgi)
-
- ᏓᏓᏚᎬ ᎪᏪᎵ ...`
+**Sample 1:** `ᎠᏥᏂᏘᏂᎠ ᏙᏱᏗᏢᎦᏚᎲ ᏧᎦᎾᏮ ᎠᎹᏰᏟ-Ꮒ. Buenos Aires ᎠᏰᎵᏗᎦᎳᏫᎢᏍᏗ ᎦᏚᎲᎢ. ᏓᏓᏚᎬ ᎪᏪᎵ ᏙᏯᏗᏢ ᏗᏕᎬᏔᏛ ᎠᎹ...`
 
 | Vocab | Tokens | Count |
 |-------|--------|-------|
-| 8k | `▁ᏍᎪᏯ ▁ᏗᏃ ᎷᏩᏘᏍᎩ " consortium ▁word ▁list ." ▁ 2 ... (+31 more)` | 41 |
-| 16k | `▁ᏍᎪᏯ ▁ᏗᏃ ᎷᏩᏘᏍᎩ " consortium ▁word ▁list ." ▁ 2 ... (+30 more)` | 40 |
-| 32k | `▁ᏍᎪᏯ ▁ᏗᏃᎷᏩᏘᏍᎩ " consortium ▁word ▁list ." ▁ 2 0 ... (+26 more)` | 36 |
+| 8k | `▁ᎠᏥᏂᏘᏂᎠ ▁ᏙᏱᏗᏢᎦᏚᎲ ▁ᏧᎦᎾᏮ ▁ᎠᎹᏰᏟ - Ꮒ . ▁buenos ▁aires ▁ᎠᏰᎵᏗᎦᎳᏫᎢᏍᏗ ... (+10 more)` | 20 |
+| 16k | `▁ᎠᏥᏂᏘᏂᎠ ▁ᏙᏱᏗᏢᎦᏚᎲ ▁ᏧᎦᎾᏮ ▁ᎠᎹᏰᏟ - Ꮒ . ▁buenos ▁aires ▁ᎠᏰᎵᏗᎦᎳᏫᎢᏍᏗ ... (+10 more)` | 20 |
+| 32k | `▁ᎠᏥᏂᏘᏂᎠ ▁ᏙᏱᏗᏢᎦᏚᎲ ▁ᏧᎦᎾᏮ ▁ᎠᎹᏰᏟ - Ꮒ . ▁buenos ▁aires ▁ᎠᏰᎵᏗᎦᎳᏫᎢᏍᏗ ... (+10 more)` | 20 |
 
-**Sample 2:** `ᎤᎧᏲᏗ ᎡᎶᎯᏱ ...
-
-Category:ᎠᏓᏕᏲᎲᏍᎩ ᎡᎶᎯ
-Category:To be checked`
+**Sample 2:** `ᎠᎷᏆ"Tsalagi Anagalisgi." ᏙᏱᏗᏢᎦᏚᎲ ᎾᏍᎩ ᏁᏛᎳᏂ-Ꮒ. Oranjestad ᎠᏰᎵᏗᎦᎳᏫᎢᏍᏗ ᎦᏚᎲᎢ. ᏓᏓᏚᎬ ᎪᏪ...`
 
 | Vocab | Tokens | Count |
 |-------|--------|-------|
-| 8k | `▁ᎤᎧ ᏲᏗ ▁ᎡᎶᎯ Ᏹ ▁... ▁category : ᎠᏓᏕᏲᎲᏍᎩ ▁ᎡᎶᎯ ▁category ... (+4 more)` | 14 |
-| 16k | `▁ᎤᎧᏲᏗ ▁ᎡᎶᎯ Ᏹ ▁... ▁category : ᎠᏓᏕᏲᎲᏍᎩ ▁ᎡᎶᎯ ▁category : ... (+3 more)` | 13 |
-| 32k | `▁ᎤᎧᏲᏗ ▁ᎡᎶᎯᏱ ▁... ▁category : ᎠᏓᏕᏲᎲᏍᎩ ▁ᎡᎶᎯ ▁category : to ... (+2 more)` | 12 |
-
-**Sample 3:** `ᎤᏢᏅᏛ"Consortium Word List." 2016-04-15. (tlvnvdv)
-
- ᏓᏓᏚᎬ ᎪᏪᎵ 
-
- ᏙᏯᏗᏢ ᏗᏕᎬᏔᏛ 
+| 8k | `▁ᎠᎷ Ꮖ " tsalagi ▁anagalisgi ." ▁ᏙᏱᏗᏢᎦᏚᎲ ▁ᎾᏍᎩ ▁ᏁᏛᎳᏂ - ... (+18 more)` | 28 |
+| 16k | `▁ᎠᎷᏆ " tsalagi ▁anagalisgi ." ▁ᏙᏱᏗᏢᎦᏚᎲ ▁ᎾᏍᎩ ▁ᏁᏛᎳᏂ - Ꮒ ... (+14 more)` | 24 |
+| 32k | `▁ᎠᎷᏆ " tsalagi ▁anagalisgi ." ▁ᏙᏱᏗᏢᎦᏚᎲ ▁ᎾᏍᎩ ▁ᏁᏛᎳᏂ - Ꮒ ... (+13 more)` | 23 |
 
-Cat...`
+**Sample 3:** `ᎴᎹᏂ (Lemani) ᎠᏓᎶᏂᎨᎢ ᎤᏓᏔᏅᎯ Ꭰ���ᏍᏗ ᎨᏐᎢ. ᎠᎩᏍᏗ be checked`
 
 | Vocab | Tokens | Count |
 |-------|--------|-------|
-| 8k | `▁ᎤᏢ ᏅᏛ " consortium ▁word ▁list ." ▁ 2 0 ... (+26 more)` | 36 |
-| 16k | `▁ᎤᏢᏅᏛ " consortium ▁word ▁list ." ▁ 2 0 1 ... (+25 more)` | 35 |
-| 32k | `▁ᎤᏢᏅᏛ " consortium ▁word ▁list ." ▁ 2 0 1 ... (+24 more)` | 34 |
+| 8k | `▁Ꮄ ᎹᏂ ▁( le mani ) ▁ᎠᏓᎶᏂᎨᎢ ▁ᎤᏓᏔᏅᎯ ▁ᎠᎩᏍᏗ ▁ᎨᏐᎢ ... (+4 more)` | 14 |
+| 16k | `▁ᎴᎹᏂ ▁( lemani ) ▁ᎠᏓᎶᏂᎨᎢ ▁ᎤᏓᏔᏅᎯ ▁ᎠᎩᏍᏗ ▁ᎨᏐᎢ . ▁ᎠᎩᏍᏗ ... (+2 more)` | 12 |
+| 32k | `▁ᎴᎹᏂ ▁( lemani ) ▁ᎠᏓᎶᏂᎨᎢ ▁ᎤᏓᏔᏅᎯ ▁ᎠᎩᏍᏗ ▁ᎨᏐᎢ . ▁ᎠᎩᏍᏗ ... (+2 more)` | 12 |
 
 
 ### Key Findings
 
-- **Best Compression:** 32k achieves 3.838x compression
-- **Lowest UNK Rate:** 8k with 0.1137% unknown tokens
+- **Best Compression:** 32k achieves 3.573x compression
+- **Lowest UNK Rate:** 8k with 0.1472% unknown tokens
 - **Trade-off:** Larger vocabularies improve compression but increase model size
 - **Recommendation:** 32k vocabulary provides optimal balance for production use
 
@@ -128,57 +125,89 @@ Cat...`
 
 ![N-gram Perplexity](visualizations/ngram_perplexity.png)
 
+![N-gram Unique](visualizations/ngram_unique.png)
+
 ![N-gram Coverage](visualizations/ngram_coverage.png)
 
 ### Results
 
-| N-gram | Perplexity | Entropy | Unique N-grams | Top-100 Coverage | Top-1000 Coverage |
-|--------|------------|---------|----------------|------------------|-------------------|
-| **2-gram** | 257 🏆 | 8.00 | 1,020 | 64.1% | 99.6% |
-| **2-gram** | 893 🏆 | 9.80 | 3,626 | 42.9% | 85.8% |
-| **3-gram** | 292 | 8.19 | 1,275 | 62.4% | 96.1% |
-| **3-gram** | 3,674 | 11.84 | 14,102 | 27.3% | 56.4% |
-| **4-gram** | 539 | 9.07 | 2,691 | 53.0% | 83.5% |
-| **4-gram** | 7,268 | 12.83 | 31,632 | 25.8% | 46.0% |
+| N-gram | Variant | Perplexity | Entropy | Unique N-grams | Top-100 Coverage | Top-1000 Coverage |
+|--------|---------|------------|---------|----------------|------------------|-------------------|
+| **2-gram** | Word | 160 🏆 | 7.32 | 494 | 67.3% | 100.0% |
+| **2-gram** | Subword | 938 | 9.87 | 3,277 | 40.0% | 86.1% |
+| **3-gram** | Word | 222 | 7.80 | 667 | 63.1% | 100.0% |
+| **3-gram** | Subword | 4,484 | 12.13 | 12,903 | 21.8% | 52.1% |
+| **4-gram** | Word | 508 | 8.99 | 1,308 | 48.2% | 89.2% |
+| **4-gram** | Subword | 9,908 | 13.27 | 28,877 | 18.5% | 39.2% |
 
 ### Top 5 N-grams by Size
 
-**2-grams:**
+**2-grams (Word):**
+
+| Rank | N-gram | Count |
+|------|--------|-------|
+| 1 | `be checked` | 850 |
+| 2 | `ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ` | 578 |
+| 3 | `ꮣꮣꮪꭼ ꭺꮺꮅ` | 472 |
+| 4 | `ꭺꮺꮅ ꮩꮿꮧꮲ` | 431 |
+| 5 | `word list` | 345 |
+
+**3-grams (Word):**
 
 | Rank | N-gram | Count |
 |------|--------|-------|
-| 1 | `category :` | 2,341 |
-| 2 | `be checked` | 958 |
-| 3 | `to be` | 958 |
-| 4 | `: to` | 958 |
-| 5 | `ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ` | 609 |
+| 1 | `ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ` | 431 |
+| 2 | `ꮣꮣꮪꭼ ꭺꮺꮅ ꮩꮿꮧꮲ` | 431 |
+| 3 | `consortium word list` | 343 |
+| 4 | `ꮧꮥꭼꮤꮫ be checked` | 227 |
+| 5 | `ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ be` | 216 |
 
-**3-grams:**
+**4-grams (Word):**
 
 | Rank | N-gram | Count |
 |------|--------|-------|
-| 1 | `to be checked` | 958 |
-| 2 | `: to be` | 958 |
-| 3 | `category : to` | 958 |
-| 4 | `ꮧꮥꭼꮤꮫ category :` | 596 |
-| 5 | `ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ category` | 566 |
+| 1 | `ꮣꮣꮪꭼ ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ` | 431 |
+| 2 | `ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ be checked` | 216 |
+| 3 | `ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ be` | 163 |
+| 4 | `ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ ꭰꭶꮞꮝꮤꮕ` | 96 |
+| 5 | `ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ ꭰꭶꮞꮝꮤꮕ be` | 96 |
 
-**4-grams:**
+**2-grams (Subword):**
 
 | Rank | N-gram | Count |
 |------|--------|-------|
-| 1 | `category : to be` | 958 |
-| 2 | `: to be checked` | 958 |
-| 3 | `ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ category :` | 566 |
-| 4 | `ꮣꮣꮪꭼ ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ` | 444 |
-| 5 | `ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ category` | 428 |
+| 1 | `_ ꭰ` | 5,396 |
+| 2 | `_ ꭴ` | 3,446 |
+| 3 | `ꮧ _` | 2,866 |
+| 4 | `. _` | 2,628 |
+| 5 | `, _` | 2,120 |
+
+**3-grams (Subword):**
+
+| Rank | N-gram | Count |
+|------|--------|-------|
+| 1 | `ꮝ ꮧ _` | 1,406 |
+| 2 | `_ c h` | 987 |
+| 3 | `_ ꭰ ꮄ` | 975 |
+| 4 | `c h e` | 965 |
+| 5 | `ꭰ ꮄ _` | 899 |
+
+**4-grams (Subword):**
+
+| Rank | N-gram | Count |
+|------|--------|-------|
+| 1 | `_ c h e` | 918 |
+| 2 | `_ ꭰ ꮄ _` | 892 |
+| 3 | `e _ c h` | 857 |
+| 4 | `_ b e _` | 851 |
+| 5 | `c k e d` | 850 |
 
 
 ### Key Findings
 
-- **Best Perplexity:** 2-gram with 257
+- **Best Perplexity:** 2-gram (word) with 160
 - **Entropy Trend:** Decreases with larger n-grams (more predictable)
-- **Coverage:** Top-1000 patterns cover ~46% of corpus
+- **Coverage:** Top-1000 patterns cover ~39% of corpus
 - **Recommendation:** 4-gram or 5-gram for best predictive performance
 
 ---
@@ -186,55 +215,86 @@ Cat...`
 
 ![Markov Entropy](visualizations/markov_entropy.png)
 
+![Markov Contexts](visualizations/markov_contexts.png)
+
 ![Markov Branching](visualizations/markov_branching.png)
 
 ### Results
 
-| Context | Avg Entropy | Perplexity | Branching Factor | Unique Contexts | Predictability |
-|---------|-------------|------------|------------------|-----------------|----------------|
-| **1** | 0.4793 | 1.394 | 2.62 | 13,897 | 52.1% |
-| **1** | 1.6917 | 3.230 | 17.49 | 449 | 0.0% |
-| **2** | 0.1273 | 1.092 | 1.24 | 36,338 | 87.3% |
-| **2** | 1.0597 | 2.084 | 4.84 | 7,851 | 0.0% |
-| **3** | 0.0430 | 1.030 | 1.08 | 45,090 | 95.7% |
-| **3** | 0.5746 | 1.489 | 2.30 | 38,007 | 42.5% |
-| **4** | 0.0217 🏆 | 1.015 | 1.04 | 48,532 | 97.8% |
-| **4** | 0.2783 🏆 | 1.213 | 1.46 | 87,252 | 72.2% |
+| Context | Variant | Avg Entropy | Perplexity | Branching Factor | Unique Contexts | Predictability |
+|---------|---------|-------------|------------|------------------|-----------------|----------------|
+| **1** | Word | 0.4942 | 1.409 | 2.32 | 13,243 | 50.6% |
+| **1** | Subword | 1.6130 | 3.059 | 16.20 | 448 | 0.0% |
+| **2** | Word | 0.0935 | 1.067 | 1.16 | 30,660 | 90.6% |
+| **2** | Subword | 1.0024 | 2.003 | 4.68 | 7,258 | 0.0% |
+| **3** | Word | 0.0291 | 1.020 | 1.05 | 35,274 | 97.1% |
+| **3** | Subword | 0.5824 | 1.497 | 2.33 | 33,930 | 41.8% |
+| **4** | Word | 0.0141 🏆 | 1.010 | 1.02 | 36,782 | 98.6% |
+| **4** | Subword | 0.2770 | 1.212 | 1.46 | 79,071 | 72.3% |
+
+### Generated Text Samples (Word-based)
+
+Below are text samples generated from each word-based Markov chain model:
+
+**Context Size 1:**
+
+1. `ꭰꮄ ꭿꭰ ꮳꮃꭹ ꮧꭶꮳꭶꮈᏼ ꮧꮜꮓ ꮪꮎꮜꮓꭾꭲ ꭰꮒꮝꭶꮿꮓ ꭰꮗꮱꮝꮧ ꭰꮏꮼꭲ ꭵꮝꮚ ꭰꭺꮩꮝꭶ ꭴꮒꭺꮫ ᏼꮻ ꭴꮕꮤꮒꮣꮝꮩꮧ v ꮝ`
+2. `be checked ꭿꭺꮹꮤ ꮎꮝꭶ ꮎꮝꮗ ꭰꮝꮪꭲꮫ ꭿꭰ ꭰꮣꮄꮒꮝꭼ ꭰꮫꮝꭼ ꭰᏸꮈ ꭴꮣꮄꮴꮂ consortium word list sdagoi ꮣꮣꮪꭼ`
+3. `ꭿꭰ ꮳꮃꭹ ꭴꭼꮻᏻꭿ ꭸꮢ ꮎꮏ ꮌꮞꮋꮗꭹ ꮎ ꮎꮝꭹ ꮑꮫꮃꮒ ꮪꮎꮩꮲꮹꮧꮢ be checked ꭱꮃꮧꮬ thumb none thumb`
+
+**Context Size 2:**
+
+1. `ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ be checked ꮪꮎꮩꮲꮹꮧꮢ`
+2. `ꮣꮣꮪꭼ ꭺꮺꮅ ꭴꮣꮞꭶꮴꮧ ꮧꮥꭼꮤꮫ ꮣꮆꮒꭶꮝꮫ ᏻꮃꮫ ꭼꮏꭸꮝꮫ ꮷᏼꮲ ꭰꮉᏸꮯ be checked ꭱꮃꮧꮬ ꮖꮗꭲꮴꭹꭲꮒ`
+3. `ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ be checked ꮤꮝꮊꮒꮿ ꮷꮑꭿ ꭰꮄ ꭿꮈꮝꮧ ꭰᏺꮅꭸ ꭰꮉᏸꮅ ꭷꮓꭾꭽ ꭰꮄ ꮶꭲ ꮏꮎ ꭸꮢ yolngu`
+
+**Context Size 3:**
+
+1. `ꮣꮣꮪꭼ ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ ꮪꮎꮩꮲꮹꮧꮢ be checked`
+2. `ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ be checked ꭱꮃꮧꮬ ꮖꮗꭲꮴꭹꭲꮒ`
+3. `consortium word list dikaneisd digoluwatisgi ꮣꮣꮪꭼ ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ ꮔꮫꮏꮥꭼ be checked`
+
+**Context Size 4:**
+
+1. `ꮣꮣꮪꭼ ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ ꭰꭶꮞꮝꮩꮧ be checked`
+2. `ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ be checked ꮷᏼꮲ ꭰꮉᏸꮯ ᏻꮃꮫ ꮣꮆꮒꭶꮝꮫ ꭼꮏꭸꮝꮫ`
+3. `ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ ꭰꭶꮞꮝꮤꮕ be checked`
+
 
-### Generated Text Samples
+### Generated Text Samples (Subword-based)
 
-Below are text samples generated from each Markov chain model:
+Below are text samples generated from each subword-based Markov chain model:
 
 **Context Size 1:**
 
-1. `. ꮳꮃꭹ ᏼꮻ . ᏹꮒꭼꮫꮎ , 537 the united nations statistics ꭸꮎꮣ ꭹ ꭹꭲꮒ ( ꮧꮡꮻꮝꮧrobinson`
-2. `: to be checked category : to be checked ꭿꭺꮹꮤ ꮎꮝꭶ ꮎꮝꮗ ꭶꮒꮭꭲ . " consortium`
-3. `, 863 , bekasi ) ꮔ nu ꮕ nv ) ꮣꮣꮪꭼ ꭺꮺꮅ ꮧꭷꮑꮝꮩꮧ " consortium word`
+1. `_ꭻꮹꮧꮳꮕᏹꭹꮸꮟꮅꭳꮜꭲꮧ_`
+2. `ꮧꮲ_ꮣꮞꭲᏻꭿꮵꮅ_blipe`
+3. `ꮝꮩꮲꮹꮢ_ꭰ_ꭴꮓꭿᏻꭿꭰ_ꮣ`
 
 **Context Size 2:**
 
-1. `category : to be checked category : ꭼꮓꮣ ꭰꭶꮞꮝꮩꮧ category : to be checked category : ꮒꭶꭵ`
-2. `to be checked category : ꮳꮃꭹ category : ꭰꮒꮩꮃ category : ꭰᏸꮅ ꮪꮎꮩꮲꮹꮧꮢ category : ꮧꭶꭺꭿꮧ category`
-3. `: to be checked category : ꮳꮃꭹ category : ꭶꮎꮭꭲ ( ꮷꭶꮓꮾ ꭰꮊꮅꭶ ) category : ꮣꮆꮒꭶꮝꮫ`
+1. `_ꭰꮄ_ꭴꮎꮴꮅ_ꮧꭶꮎꮕꮟꮣꮑꮈ`
+2. `_ꭴꮓꮈꮤꮕꭲ,_.._tr>_<`
+3. `ꮧ_list."_(iaꭰꮒꭺꮫ_`
 
 **Context Size 3:**
 
-1. `category : to be checked category : ꭴꮬꮹꮣ category : ꭶꮎꮭꭲ ( ꭱꮃꮧꮬ ) category : ꭶꮎꮭꭲ (`
-2. `: to be checked nds : land # länner sv : världsgeografi # lista över länder`
-3. `ꮧꮥꭼꮤꮫ category : ꭼꮓꮣ ꭰꭶꮞꮝꮩꮧ category : to be checked category : ꭶꮎꮭꭲ ( ꮷꭶꮓꮾ ꭰꮊꮅꭶ ) category`
+1. `ꮝꮧ_ble._galv)._ꮣꮣꮪ`
+2. `_checked_(gila)_ꮗꮅ`
+3. `_ꭰꮄ_ꭴꮣꮔꮦᏺꮂ_ꮞꮅꮎ_seq`
 
 **Context Size 4:**
 
-1. `category : to be checked category : ꭶꮎꮭꭲ ( ᏻꮃꮫ ) category : ꭶꮎꮭꭲ ( ꮷꭶꮓꮾ ꭰꮊꮅꭶ ) category`
-2. `ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ category : ꭼꮓꮣ ꭰꭶꮞꮝꮩꮧ category : to be checked nds : land # länner sv : världsgeografi`
-3. `ꮣꮣꮪꭼ ꭺꮺꮅ ꮩꮿꮧꮲ ꮧꮥꭼꮤꮫ category : ꭴꮎꮩꮲꭿ category : to be checked category : ꭰᏸꮅ ꮪꮎꮩꮲꮹꮧꮢ category : to`
+1. `_checked_(ꭱꮃꮧꮭꭲ_ꭰꮣꭿ`
+2. `_ꭰꮄ_ꭰꮣꭿꭿ_polydeuces`
+3. `e_checked_(ꭱꮃꮧꮬ._re`
 
 
 ### Key Findings
 
-- **Best Predictability:** Context-4 with 97.8% predictability
+- **Best Predictability:** Context-4 (word) with 98.6% predictability
 - **Branching Factor:** Decreases with context size (more deterministic)
-- **Memory Trade-off:** Larger contexts require more storage (87,252 contexts)
+- **Memory Trade-off:** Larger contexts require more storage (79,071 contexts)
 - **Recommendation:** Context-3 or Context-4 for text generation
 
 ---
@@ -250,26 +310,26 @@ Below are text samples generated from each Markov chain model:
 
 | Metric | Value |
 |--------|-------|
-| Vocabulary Size | 4,475 |
-| Total Tokens | 42,573 |
-| Mean Frequency | 9.51 |
+| Vocabulary Size | 4,236 |
+| Total Tokens | 35,186 |
+| Mean Frequency | 8.31 |
 | Median Frequency | 3 |
-| Frequency Std Dev | 53.38 |
+| Frequency Std Dev | 35.59 |
 
 ### Most Common Words
 
 | Rank | Word | Frequency |
 |------|------|-----------|
-| 1 | category | 2,341 |
-| 2 | to | 976 |
-| 3 | be | 960 |
-| 4 | checked | 958 |
-| 5 | ꭰꮄ | 903 |
-| 6 | ꭿꭰ | 772 |
-| 7 | ꮧꮥꭼꮤꮫ | 649 |
-| 8 | ꮩꮿꮧꮲ | 611 |
-| 9 | ꭺꮺꮅ | 548 |
-| 10 | ꮣꮣꮪꭼ | 506 |
+| 1 | ꭰꮄ | 903 |
+| 2 | be | 852 |
+| 3 | checked | 850 |
+| 4 | ꭿꭰ | 797 |
+| 5 | ꮧꮥꭼꮤꮫ | 611 |
+| 6 | ꮩꮿꮧꮲ | 580 |
+| 7 | ꭺꮺꮅ | 524 |
+| 8 | ꮣꮣꮪꭼ | 482 |
+| 9 | ꮳꮃꭹ | 475 |
+| 10 | word | 346 |
 
 ### Least Common Words (from vocabulary)
 
@@ -290,24 +350,24 @@ Below are text samples generated from each Markov chain model:
 
 | Metric | Value |
 |--------|-------|
-| Zipf Coefficient | 0.8972 |
-| R² (Goodness of Fit) | 0.986656 |
+| Zipf Coefficient | 0.8714 |
+| R² (Goodness of Fit) | 0.984546 |
 | Adherence Quality | **excellent** |
 
 ### Coverage Analysis
 
 | Top N Words | Coverage |
 |-------------|----------|
-| Top 100 | 44.9% |
-| Top 1,000 | 76.0% |
+| Top 100 | 39.8% |
+| Top 1,000 | 74.2% |
 | Top 5,000 | 0.0% |
 | Top 10,000 | 0.0% |
 
 ### Key Findings
 
-- **Zipf Compliance:** R²=0.9867 indicates excellent adherence to Zipf's law
-- **High Frequency Dominance:** Top 100 words cover 44.9% of corpus
-- **Long Tail:** -5,525 words needed for remaining 100.0% coverage
+- **Zipf Compliance:** R²=0.9845 indicates excellent adherence to Zipf's law
+- **High Frequency Dominance:** Top 100 words cover 39.8% of corpus
+- **Long Tail:** -5,764 words needed for remaining 100.0% coverage
 
 ---
 ## 5. Word Embeddings Evaluation
@@ -320,24 +380,85 @@ Below are text samples generated from each Markov chain model:
 
 ![t-SNE Sentences](visualizations/tsne_sentences.png)
 
-### Model Comparison
 
-| Model | Vocab Size | Dimension | Avg Norm | Std Norm | Isotropy |
-|-------|------------|-----------|----------|----------|----------|
-| **mono_32d** | 1,242 | 32 | 2.681 | 0.618 | 0.2339 🏆 |
-| **mono_64d** | 1,242 | 64 | 2.624 | 0.626 | 0.0550 |
-| **mono_128d** | 1,242 | 128 | 2.563 | 0.654 | 0.0093 |
-| **embeddings_enhanced** | 0 | 0 | 0.000 | 0.000 | 0.0000 |
+### 5.1 Cross-Lingual Alignment
+
+> *Note: Multilingual alignment visualization not available for this language.*
+
+
+### 5.2 Model Comparison
+
+| Model | Dimension | Isotropy | Semantic Density | Alignment R@1 | Alignment R@10 |
+|-------|-----------|----------|------------------|---------------|----------------|
+| **mono_32d** | 32 | 0.2005 🏆 | 0.5045 | N/A | N/A |
+| **mono_64d** | 64 | 0.0662 | 0.4736 | N/A | N/A |
+| **mono_128d** | 128 | 0.0103 | 0.4890 | N/A | N/A |
 
 ### Key Findings
 
-- **Best Isotropy:** mono_32d with 0.2339 (more uniform distribution)
-- **Dimension Trade-off:** Higher dimensions capture more semantics but reduce isotropy
-- **Vocabulary Coverage:** All models cover 1,242 words
-- **Recommendation:** 100d for balanced semantic capture and efficiency
+- **Best Isotropy:** mono_32d with 0.2005 (more uniform distribution)
+- **Semantic Density:** Average pairwise similarity of 0.4890. Lower values indicate better semantic separation.
+- **Alignment Quality:** No aligned models evaluated in this run.
+- **Recommendation:** 128d aligned for best cross-lingual performance
 
 ---
-## 6. Summary & Recommendations
+## 6.  Morphological Analysis (Experimental)
+
+> ⚠️ **Warning:** This language shows low morphological productivity. The statistical signals used for this analysis may be noisy or less reliable than for morphologically rich languages.
+
+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 | **0.000** | Low morphological productivity | ⚠️ Likely unreliable |
+| Idiomaticity Gap | **-1.000** | 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 |
+|--------|----------|
+
+#### Productive Suffixes
+| Suffix | Examples |
+|--------|----------|
+| `-ꮝꮧ` | ꮵꮣꮯꮆꮝꮧ, ꭴꮺꮕꮝꮧ, ꭰꭶꮤꮂꮝꮧ |
+
+### 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.
+
+*No significant affix co-occurrences detected.*
+
+
+### 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 |
+|------|-----------------|------------|------|
+| ꮒꭶꮅꮝꮧꮝꭸꮝꮧ | **`ꮒꭶꮅꮝꮧꮝꭸ-ꮝꮧ`** | 1.5 | `ꮒꭶꮅꮝꮧꮝꭸ` |
+| ꭰᏸꮅꮧꭶꮃꮻꭲꮝꮧ | **`ꭰᏸꮅꮧꭶꮃꮻꭲ-ꮝꮧ`** | 1.5 | `ꭰᏸꮅꮧꭶꮃꮻꭲ` |
+| ꭰꮜꮓꮧꮥꮆꮖꮝꮧ | **`ꭰꮜꮓꮧꮥꮆꮖ-ꮝꮧ`** | 1.5 | `ꭰꮜꮓꮧꮥꮆꮖ` |
+
+### 6.6 Linguistic Interpretation
+
+> **Automated Insight:**
+The language CHR appears to be more isolating or has a highly fixed vocabulary. Word-level models perform nearly as well as subword models, indicating fewer productive morphological processes.
+
+---
+## 7. Summary & Recommendations
 
 ![Performance Dashboard](visualizations/performance_dashboard.png)
 
@@ -345,11 +466,12 @@ Below are text samples generated from each Markov chain model:
 
 | Component | Recommended | Rationale |
 |-----------|-------------|-----------|
-| Tokenizer | **32k BPE** | Best compression (3.84x) with low UNK rate |
-| N-gram | **5-gram** | Lowest perplexity (257) |
-| Markov | **Context-4** | Highest predictability (97.8%) |
+| Tokenizer | **32k BPE** | Best compression (3.57x) |
+| N-gram | **2-gram** | Lowest perplexity (160) |
+| Markov | **Context-4** | Highest predictability (98.6%) |
 | Embeddings | **100d** | Balanced semantic capture and isotropy |
 
+
 ---
 ## Appendix: Metrics Glossary & Interpretation Guide
 
@@ -539,7 +661,8 @@ If you use these models in your research, please cite:
   author = {Kamali, Omar},
   title = {Wikilangs: Open NLP Models for Wikipedia Languages},
   year = {2025},
-  publisher = {HuggingFace},
+  doi = {10.5281/zenodo.18073153},
+  publisher = {Zenodo},
   url = {https://huggingface.co/wikilangs}
   institution = {Omneity Labs}
 }
@@ -555,7 +678,8 @@ MIT License - Free for academic and commercial use.
 - 🤗 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: 2025-12-28 22:42:07*
+*Report Date: 2026-01-03 10:09:15*

models/embeddings/monolingual/chr_128d.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3d20dfbda9a0fad185b3fbdecaaf3911d08563e5eec2416cc1d49cabcfd19fbe
-size 1025297761
+oid sha256:b65a441c9c4908f0ffa5c095ba64382b4b323cde80a4f6d979c87e4b93102427
+size 1025246107

models/embeddings/monolingual/chr_128d_metadata.json

@@ -3,11 +3,13 @@
   "dimension": 128,
   "version": "monolingual",
   "training_params": {
-    "dim": 128,
+    "algorithm": "skipgram",
     "min_count": 5,
     "window": 5,
     "negative": 5,
-    "epochs": 5
+    "epochs": 5,
+    "encoding_method": "rope",
+    "dim": 128
   },
-  "vocab_size": 1242
+  "vocab_size": 1193
 }

models/embeddings/monolingual/chr_32d.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c1700517d4093dd3dfbf1011faa55ebb1e1b7cd19fffeb86d693090d2c17c336
-size 256343905
+oid sha256:d9b5ecb349ca3c9d1d8f1f38b34647edbd029de966e73fe6f8043e5b64357e71
+size 256329883

models/embeddings/monolingual/chr_32d_metadata.json

@@ -3,11 +3,13 @@
   "dimension": 32,
   "version": "monolingual",
   "training_params": {
-    "dim": 32,
+    "algorithm": "skipgram",
     "min_count": 5,
     "window": 5,
     "negative": 5,
-    "epochs": 5
+    "epochs": 5,
+    "encoding_method": "rope",
+    "dim": 32
   },
-  "vocab_size": 1242
+  "vocab_size": 1193
 }

models/embeddings/monolingual/chr_64d.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:09d69f872f3cb41d7d61e371667e89c7559d22c55e910dd6ec1cc05d27940abd
-size 512661857
+oid sha256:2ff3baa5798aac84304b756ac363fd816043d3a9775848343c3a39af84cf09bf
+size 512635291

models/embeddings/monolingual/chr_64d_metadata.json

@@ -3,11 +3,13 @@
   "dimension": 64,
   "version": "monolingual",
   "training_params": {
-    "dim": 64,
+    "algorithm": "skipgram",
     "min_count": 5,
     "window": 5,
     "negative": 5,
-    "epochs": 5
+    "epochs": 5,
+    "encoding_method": "rope",
+    "dim": 64
   },
-  "vocab_size": 1242
+  "vocab_size": 1193
 }

models/subword_markov/chr_markov_ctx1_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:94764e1f3d25aa83b86cc0bc93dae20352a14e0147c26b8923882474f061a379
-size 51927
+oid sha256:47a719104c2e085b3f0210500f2e0d35dd3fad4c62155a28ff1885ad952f603e
+size 48128

models/subword_markov/chr_markov_ctx1_subword_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 1,
   "variant": "subword",
   "language": "chr",
-  "unique_contexts": 449,
-  "total_transitions": 304475
+  "unique_contexts": 448,
+  "total_transitions": 244557
 }

models/subword_markov/chr_markov_ctx2_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:832beea5bece4732c22e619cef5564a0512d794894d07a9f5943bd7eea7020fb
-size 242322
+oid sha256:90017c14d3413b7443aa8cfd4e8186bbbfa36eb5c061301510a67e33d59ced11
+size 213598

models/subword_markov/chr_markov_ctx2_subword_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 2,
   "variant": "subword",
   "language": "chr",
-  "unique_contexts": 7851,
-  "total_transitions": 303415
+  "unique_contexts": 7258,
+  "total_transitions": 243650
 }

models/subword_markov/chr_markov_ctx3_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a6e96393ee13e9a83e555b4fc7719ccadbad4a3d5edcae2387524f2c1cb9bced
-size 695224
+oid sha256:3aa3158ad438f31c7b0fe3d234fab09e4e04d062e867a5f9276cd4c9c48b1d0c
+size 611513

models/subword_markov/chr_markov_ctx3_subword_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 3,
   "variant": "subword",
   "language": "chr",
-  "unique_contexts": 38007,
-  "total_transitions": 302355
+  "unique_contexts": 33930,
+  "total_transitions": 242743
 }

models/subword_markov/chr_markov_ctx4_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:91ca7a6f04658e80a3bee03c2eb5c7a6e0575ba5e7b3311b5a5e42c2367c2de4
-size 1281671
+oid sha256:041f8fb81d9f01bcb72c1f16d3266d79dd5c44b4454b909b4095a6dc7c4e41c9
+size 1161581

models/subword_markov/chr_markov_ctx4_subword_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 4,
   "variant": "subword",
   "language": "chr",
-  "unique_contexts": 87252,
-  "total_transitions": 301295
+  "unique_contexts": 79071,
+  "total_transitions": 241836
 }

models/subword_ngram/chr_2gram_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d1b27ecd2e3bf4c97a30d980ece0b0db63017dc2003eabca6ea2fcbd56f7aba2
-size 41814
+oid sha256:8d97d4c240d2e544cee601293886cf200a3ca9744dc6f05309d9f3f6dedb5701
+size 37349

models/subword_ngram/chr_2gram_subword_metadata.json

@@ -2,6 +2,6 @@
   "n": 2,
   "variant": "subword",
   "language": "chr",
-  "unique_ngrams": 3626,
-  "total_ngrams": 304475
+  "unique_ngrams": 3277,
+  "total_ngrams": 244557
 }

models/subword_ngram/chr_3gram_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2491c3a5631b3d1d0f21d6d30f46d2658a5acdb741a45513792750076f809272
-size 171485
+oid sha256:4909748af2830cbec527b301ed9c11dc5fba2c6a9e7f48e1b13a633d5cfdcd87
+size 157760

models/subword_ngram/chr_3gram_subword_metadata.json

@@ -2,6 +2,6 @@
   "n": 3,
   "variant": "subword",
   "language": "chr",
-  "unique_ngrams": 14102,
-  "total_ngrams": 303415
+  "unique_ngrams": 12903,
+  "total_ngrams": 243650
 }

models/subword_ngram/chr_4gram_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1c9d8cf44e962d72f73232a0405cac11c2f2997bf107cc3e7f89f07a266eb115
-size 410498
+oid sha256:a19dcd2eca820072c87f80cca01672f0fafd765e2ca81e6dbfd95249c77828dc
+size 374199

models/subword_ngram/chr_4gram_subword_metadata.json

@@ -2,6 +2,6 @@
   "n": 4,
   "variant": "subword",
   "language": "chr",
-  "unique_ngrams": 31632,
-  "total_ngrams": 302355
+  "unique_ngrams": 28877,
+  "total_ngrams": 242743
 }

models/tokenizer/chr_tokenizer_16k.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4cd7d2cae1a887eaa13ebc4706df2231175ed02c2496506d48c55a25082a93c7
-size 560277
+oid sha256:359423c436a228311e5a63ed11e5f00c42f4e1f43e6dab0726841d8a8cea7899
+size 558319

models/tokenizer/chr_tokenizer_32k.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:590a2d79a920b938da84b347c50ae4007d77893ad6ba128d43467fac61235885
-size 876858
+oid sha256:bd42bf932a4c3f148e083933ebcc3b363a57f53b5b4db22c1e6a60d5017edabe
+size 835851

models/tokenizer/chr_tokenizer_8k.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6b3ffbdb4812b8aa46b6ca458563e45453188e92b8db13d34f690953654b8071
-size 391397
+oid sha256:cb436a7afc74fa9ba90ff21df46b3064327fe8aa4d69c5a5859a405d4b67a6c2
+size 392805

models/vocabulary/chr_vocabulary.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:54b695ca6569f4eb86b2f393922a1789e9b40736180d2d6b5bcee701ba4c0a6d
-size 76700
+oid sha256:8068c404442dc10b84690c7100fc813e925cbe51c1e787ef250cbc0cea84d3d3
+size 73331

models/vocabulary/chr_vocabulary_metadata.json

@@ -1,16 +1,17 @@
 {
   "language": "chr",
-  "vocabulary_size": 4475,
+  "vocabulary_size": 4236,
+  "variant": "full",
   "statistics": {
-    "type_token_ratio": 0.266443314849045,
+    "type_token_ratio": 0.2998054386679336,
     "coverage": {
-      "top_100": 0.3678373382624769,
-      "top_1000": 0.6233441158348737,
-      "top_5000": 0.8298290203327172,
-      "top_10000": 0.9261013555144794
+      "top_100": 0.3170671010361522,
+      "top_1000": 0.590561513053708,
+      "top_5000": 0.8133116148590561,
+      "top_10000": 0.9264286683860459
     },
-    "hapax_count": 9363,
-    "hapax_ratio": 0.6766151177915883,
-    "total_documents": 1060
+    "hapax_count": 9016,
+    "hapax_ratio": 0.6803501358285542,
+    "total_documents": 907
   }
 }

models/word_markov/chr_markov_ctx1_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8918e88174d5169ca34a3d8546e070f207f78c82b6a54021b21f4749b2bcac59
-size 416475
+oid sha256:2e06367eae370c3314f083ee9fc62374591dc05fc32a00400e11de6f6a357de8
+size 388759

models/word_markov/chr_markov_ctx1_word_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 1,
   "variant": "word",
   "language": "chr",
-  "unique_contexts": 13897,
-  "total_transitions": 68467
+  "unique_contexts": 13243,
+  "total_transitions": 43295
 }

models/word_markov/chr_markov_ctx2_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8dfcf759c0a6ff971c3a7533de300aed0be285590a97c16fc1cd0747fefc4a90
-size 763529
+oid sha256:efac8874c4bac9908d723a0cd4c5b8e19f9b13e1c696f3e7076b4376fd315124
+size 694832

models/word_markov/chr_markov_ctx2_word_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 2,
   "variant": "word",
   "language": "chr",
-  "unique_contexts": 36338,
-  "total_transitions": 67407
+  "unique_contexts": 30660,
+  "total_transitions": 42388
 }

models/word_markov/chr_markov_ctx3_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f281bedae46cd1bfd33bbc0700ca416910fa96689b0ac5c6f08a701d13123382
-size 963702
+oid sha256:61472a61178571af03b7dbace2a9fde2feec3747df77700588b54d959fe32645
+size 845527

models/word_markov/chr_markov_ctx3_word_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 3,
   "variant": "word",
   "language": "chr",
-  "unique_contexts": 45090,
-  "total_transitions": 66349
+  "unique_contexts": 35274,
+  "total_transitions": 41481
 }

models/word_markov/chr_markov_ctx4_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2bbf2d1e5afc57aa98fb73deb55260a7990021c69bacc60cfd97aeae6bc20ebf
-size 1109183
+oid sha256:6aa8b7bda0357275e7522519313380f562702793a851b30e31d9572a0788269c
+size 962465

models/word_markov/chr_markov_ctx4_word_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 4,
   "variant": "word",
   "language": "chr",
-  "unique_contexts": 48532,
-  "total_transitions": 65291
+  "unique_contexts": 36782,
+  "total_transitions": 40574
 }

models/word_ngram/chr_2gram_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9b9df0a5e5cf8f1dd0a206e774645f09871457def4961be9f74cad43d9572d79
-size 19190
+oid sha256:d6dcb76c997efc3d0f422d57c9533c80f8b2f241a407922fa917f6630a6cb7a9
+size 12045

models/word_ngram/chr_2gram_word_metadata.json

@@ -2,6 +2,6 @@
   "n": 2,
   "variant": "word",
   "language": "chr",
-  "unique_ngrams": 1020,
-  "total_ngrams": 68467
+  "unique_ngrams": 494,
+  "total_ngrams": 43295
 }

models/word_ngram/chr_3gram_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c6229d1d90e9b36c1f6c18272e1dcff42ed4f2c3ed66c2d7a275f82dfa186aab
-size 26597
+oid sha256:3600f5f22489e28064419bd04a825cda3951b505b66b70ee4817bc8726929f00
+size 16559

models/word_ngram/chr_3gram_word_metadata.json

@@ -2,6 +2,6 @@
   "n": 3,
   "variant": "word",
   "language": "chr",
-  "unique_ngrams": 1275,
-  "total_ngrams": 67407
+  "unique_ngrams": 667,
+  "total_ngrams": 42388
 }

models/word_ngram/chr_4gram_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cf1e68ff7493da7aea3281593ac2e5ae7a6c6111a57fd51b4dcc3cd0e77c021d
-size 56699
+oid sha256:498ec6c6f2cbba90463444020e6eca072fa8d8029af3c11e05672af984d43fc6
+size 34189

models/word_ngram/chr_4gram_word_metadata.json

@@ -2,6 +2,6 @@
   "n": 4,
   "variant": "word",
   "language": "chr",
-  "unique_ngrams": 2691,
-  "total_ngrams": 66349
+  "unique_ngrams": 1308,
+  "total_ngrams": 41481
 }