Upload all models and assets for ace (20251001)

README.md

@@ -23,14 +23,14 @@ dataset_info:
 metrics:
   - name: best_compression_ratio
     type: compression
-    value: 4.814
+    value: 4.925
   - name: best_isotropy
     type: isotropy
-    value: 0.5452
+    value: 0.5172
   - name: vocabulary_size
     type: vocab
-    value: 16834
-generated: 2025-12-27
+    value: 0
+generated: 2026-01-03
 ---
 
 # ACE - 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)
 
@@ -68,55 +70,57 @@ 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** | 3.989x | 3.94 | 0.2625% | 138,651 |
-| **16k** | 4.326x | 4.27 | 0.2847% | 127,870 |
-| **32k** | 4.588x | 4.53 | 0.3019% | 120,551 |
-| **64k** | 4.814x 🏆 | 4.76 | 0.3168% | 114,908 |
+| **8k** | 4.119x | 4.13 | 0.2682% | 125,632 |
+| **16k** | 4.488x | 4.50 | 0.2923% | 115,301 |
+| **32k** | 4.727x | 4.74 | 0.3079% | 109,452 |
+| **64k** | 4.925x 🏆 | 4.93 | 0.3208% | 105,066 |
 
 ### Tokenization Examples
 
 Below are sample sentences tokenized with each vocabulary size:
 
-**Sample 1:** `Jeremiah "Jerry" O'Connell (1974 – ) nakeuh sidroe aktor asay Amirika Syarikat.
-...`
+**Sample 1:** `Mukim Sepakat nakeuh saboh mukim di keucamatan Lawe Sigala-Gala Kabupatèn Acèh T...`
 
 | Vocab | Tokens | Count |
 |-------|--------|-------|
-| 8k | `▁jer em iah ▁" j er ry " ▁o ' ... (+20 more)` | 30 |
-| 16k | `▁jer em iah ▁" j er ry " ▁o ' ... (+19 more)` | 29 |
-| 32k | `▁jer em iah ▁" jerry " ▁o ' con nell ... (+17 more)` | 27 |
-| 64k | `▁jeremiah ▁" jerry " ▁o ' connell ▁( 1 9 ... (+14 more)` | 24 |
+| 8k | `▁mukim ▁sepakat ▁nakeuh ▁saboh ▁mukim ▁di ▁keucamatan ▁lawe ▁sigala - ... (+12 more)` | 22 |
+| 16k | `▁mukim ▁sepakat ▁nakeuh ▁saboh ▁mukim ▁di ▁keucamatan ▁lawe ▁sigala - ... (+12 more)` | 22 |
+| 32k | `▁mukim ▁sepakat ▁nakeuh ▁saboh ▁mukim ▁di ▁keucamatan ▁lawe ▁sigala - ... (+12 more)` | 22 |
+| 64k | `▁mukim ▁sepakat ▁nakeuh ▁saboh ▁mukim ▁di ▁keucamatan ▁lawe ▁sigala - ... (+12 more)` | 22 |
 
-**Sample 2:** `Darul Aman nakeuh saboh gampông nyang na lam keucamatan Permata, Kabupaten Bener...`
+**Sample 2:** `Propinsi Nakhon Ratchasima nakeuh saboh propinsi di timu baroh Muangthai. Nang n...`
 
 | Vocab | Tokens | Count |
 |-------|--------|-------|
-| 8k | `▁darul ▁aman ▁nakeuh ▁saboh ▁gampông ▁nyang ▁na ▁lam ▁keucamatan ▁permata ... (+10 more)` | 20 |
-| 16k | `▁darul ▁aman ▁nakeuh ▁saboh ▁gampông ▁nyang ▁na ▁lam ▁keucamatan ▁permata ... (+10 more)` | 20 |
-| 32k | `▁darul ▁aman ▁nakeuh ▁saboh ▁gampông ▁nyang ▁na ▁lam ▁keucamatan ▁permata ... (+10 more)` | 20 |
-| 64k | `▁darul ▁aman ▁nakeuh ▁saboh ▁gampông ▁nyang ▁na ▁lam ▁keucamatan ▁permata ... (+10 more)` | 20 |
-
-**Sample 3:** `Nè 
+| 8k | `▁propinsi ▁nakhon ▁ratch asi ma ▁nakeuh ▁saboh ▁propinsi ▁di ▁timu ... (+11 more)` | 21 |
+| 16k | `▁propinsi ▁nakhon ▁ratchasima ▁nakeuh ▁saboh ▁propinsi ▁di ▁timu ▁baroh ▁muangthai ... (+7 more)` | 17 |
+| 32k | `▁propinsi ▁nakhon ▁ratchasima ▁nakeuh ▁saboh ▁propinsi ▁di ▁timu ▁baroh ▁muangthai ... (+7 more)` | 17 |
+| 64k | `▁propinsi ▁nakhon ▁ratchasima ▁nakeuh ▁saboh ▁propinsi ▁di ▁timu ▁baroh ▁muangthai ... (+7 more)` | 17 |
 
-Kawan:Gampông di Subulussalam
-Kawan:Longkib, Subulussalam`
+**Sample 3:** `Kandang nakeuh gampông di Keucamatan Samalanga, Kabupatèn Bireuen, Acèh. Lumbôi ...`
 
 | Vocab | Tokens | Count |
 |-------|--------|-------|
-| 8k | `▁nè ▁kawan : gampông ▁di ▁subulussalam ▁kawan : longkib , ... (+1 more)` | 11 |
-| 16k | `▁nè ▁kawan : gampông ▁di ▁subulussalam ▁kawan : longkib , ... (+1 more)` | 11 |
-| 32k | `▁nè ▁kawan : gampông ▁di ▁subulussalam ▁kawan : longkib , ... (+1 more)` | 11 |
-| 64k | `▁nè ▁kawan : gampông ▁di ▁subulussalam ▁kawan : longkib , ... (+1 more)` | 11 |
+| 8k | `▁kandang ▁nakeuh ▁gampông ▁di ▁keucamatan ▁samalanga , ▁kabupatèn ▁bireuen , ... (+13 more)` | 23 |
+| 16k | `▁kandang ▁nakeuh ▁gampông ▁di ▁keucamatan ▁samalanga , ▁kabupatèn ▁bireuen , ... (+13 more)` | 23 |
+| 32k | `▁kandang ▁nakeuh ▁gampông ▁di ▁keucamatan ▁samalanga , ▁kabupatèn ▁bireuen , ... (+13 more)` | 23 |
+| 64k | `▁kandang ▁nakeuh ▁gampông ▁di ▁keucamatan ▁samalanga , ▁kabupatèn ▁bireuen , ... (+13 more)` | 23 |
 
 
 ### Key Findings
 
-- **Best Compression:** 64k achieves 4.814x compression
-- **Lowest UNK Rate:** 8k with 0.2625% unknown tokens
+- **Best Compression:** 64k achieves 4.925x compression
+- **Lowest UNK Rate:** 8k with 0.2682% unknown tokens
 - **Trade-off:** Larger vocabularies improve compression but increase model size
 - **Recommendation:** 32k vocabulary provides optimal balance for production use
 
@@ -125,57 +129,89 @@ Kawan:Longkib, Subulussalam`
 
 ![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** | 936 🏆 | 9.87 | 9,992 | 52.8% | 80.2% |
-| **2-gram** | 261 🏆 | 8.03 | 2,689 | 68.6% | 99.2% |
-| **3-gram** | 1,100 | 10.10 | 14,167 | 50.6% | 79.4% |
-| **3-gram** | 1,398 | 10.45 | 17,865 | 36.2% | 82.1% |
-| **4-gram** | 1,474 | 10.53 | 24,432 | 49.2% | 75.0% |
-| **4-gram** | 4,053 | 11.98 | 71,115 | 25.9% | 65.3% |
+| N-gram | Variant | Perplexity | Entropy | Unique N-grams | Top-100 Coverage | Top-1000 Coverage |
+|--------|---------|------------|---------|----------------|------------------|-------------------|
+| **2-gram** | Word | 637 | 9.32 | 7,009 | 62.6% | 83.4% |
+| **2-gram** | Subword | 224 🏆 | 7.80 | 2,204 | 71.8% | 99.5% |
+| **3-gram** | Word | 577 | 9.17 | 8,214 | 65.4% | 85.5% |
+| **3-gram** | Subword | 1,194 | 10.22 | 14,605 | 37.9% | 84.9% |
+| **4-gram** | Word | 673 | 9.39 | 12,805 | 64.5% | 83.7% |
+| **4-gram** | Subword | 3,551 | 11.79 | 59,251 | 26.2% | 67.5% |
 
 ### Top 5 N-grams by Size
 
-**2-grams:**
+**2-grams (Word):**
 
 | Rank | N-gram | Count |
 |------|--------|-------|
-| 1 | `kawan :` | 14,756 |
-| 2 | `bak laman` | 7,389 |
-| 3 | `gunong nyoe` | 7,388 |
-| 4 | `gampông di` | 6,982 |
-| 5 | `. nè` | 6,946 |
+| 1 | `bak laman` | 7,389 |
+| 2 | `gunong nyoe` | 7,388 |
+| 3 | `nyoe bak` | 5,543 |
+| 4 | `nakeuh saboh` | 5,045 |
+| 5 | `di acèh` | 4,748 |
 
-**3-grams:**
+**3-grams (Word):**
 
 | Rank | N-gram | Count |
 |------|--------|-------|
 | 1 | `gunong nyoe bak` | 5,541 |
-| 2 | `nè kawan :` | 4,508 |
-| 3 | `. nè kawan` | 4,441 |
-| 4 | `nyoe bak laman` | 3,694 |
-| 5 | `, acèh .` | 3,576 |
+| 2 | `nyoe bak laman` | 3,694 |
+| 3 | `lumbôi gampông nyoe` | 3,567 |
+| 4 | `acèh lumbôi gampông` | 3,564 |
+| 5 | `nyoe lam data` | 3,499 |
 
-**4-grams:**
+**4-grams (Word):**
 
 | Rank | N-gram | Count |
 |------|--------|-------|
-| 1 | `. nè kawan :` | 4,441 |
-| 2 | `gunong nyoe bak laman` | 3,694 |
-| 3 | `. lumbôi gampông nyoe` | 3,567 |
-| 4 | `acèh . lumbôi gampông` | 3,564 |
-| 5 | `, acèh . lumbôi` | 3,564 |
+| 1 | `gunong nyoe bak laman` | 3,694 |
+| 2 | `acèh lumbôi gampông nyoe` | 3,564 |
+| 3 | `nyoe lam data peumeurèntah` | 3,499 |
+| 4 | `gampông nyoe lam data` | 3,499 |
+| 5 | `lam data peumeurèntah nakeuh` | 3,499 |
+
+**2-grams (Subword):**
+
+| Rank | N-gram | Count |
+|------|--------|-------|
+| 1 | `e u` | 117,818 |
+| 2 | `_ n` | 79,411 |
+| 3 | `a n` | 69,436 |
+| 4 | `h _` | 68,029 |
+| 5 | `n g` | 67,573 |
+
+**3-grams (Subword):**
+
+| Rank | N-gram | Count |
+|------|--------|-------|
+| 1 | `n g _` | 44,439 |
+| 2 | `_ n a` | 31,640 |
+| 3 | `_ b a` | 30,463 |
+| 4 | `k e u` | 30,322 |
+| 5 | `_ n y` | 26,537 |
+
+**4-grams (Subword):**
+
+| Rank | N-gram | Count |
+|------|--------|-------|
+| 1 | `e u h _` | 23,348 |
+| 2 | `b a k _` | 23,260 |
+| 3 | `_ d i _` | 21,144 |
+| 4 | `k e u h` | 21,117 |
+| 5 | `a k e u` | 20,691 |
 
 
 ### Key Findings
 
-- **Best Perplexity:** 2-gram with 261
+- **Best Perplexity:** 2-gram (subword) with 224
 - **Entropy Trend:** Decreases with larger n-grams (more predictable)
-- **Coverage:** Top-1000 patterns cover ~65% of corpus
+- **Coverage:** Top-1000 patterns cover ~68% of corpus
 - **Recommendation:** 4-gram or 5-gram for best predictive performance
 
 ---
@@ -183,55 +219,86 @@ Kawan:Longkib, Subulussalam`
 
 ![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** | 0.7141 | 1.640 | 4.47 | 38,800 | 28.6% |
-| **1** | 1.0130 | 2.018 | 7.11 | 1,076 | 0.0% |
-| **2** | 0.2550 | 1.193 | 1.56 | 172,855 | 74.5% |
-| **2** | 0.8567 | 1.811 | 4.89 | 7,655 | 14.3% |
-| **3** | 0.0849 | 1.061 | 1.15 | 269,187 | 91.5% |
-| **3** | 0.7773 | 1.714 | 3.54 | 37,427 | 22.3% |
-| **4** | 0.0363 🏆 | 1.025 | 1.07 | 309,167 | 96.4% |
-| **4** | 0.5505 🏆 | 1.465 | 2.29 | 132,546 | 44.9% |
+| Context | Variant | Avg Entropy | Perplexity | Branching Factor | Unique Contexts | Predictability |
+|---------|---------|-------------|------------|------------------|-----------------|----------------|
+| **1** | Word | 0.7515 | 1.684 | 4.35 | 36,025 | 24.8% |
+| **1** | Subword | 0.8633 | 1.819 | 5.38 | 1,269 | 13.7% |
+| **2** | Word | 0.2148 | 1.161 | 1.44 | 155,224 | 78.5% |
+| **2** | Subword | 0.7739 | 1.710 | 4.50 | 6,822 | 22.6% |
+| **3** | Word | 0.0655 | 1.046 | 1.11 | 221,018 | 93.4% |
+| **3** | Subword | 0.7559 | 1.689 | 3.54 | 30,615 | 24.4% |
+| **4** | Word | 0.0242 🏆 | 1.017 | 1.04 | 242,720 | 97.6% |
+| **4** | Subword | 0.5660 | 1.480 | 2.36 | 108,223 | 43.4% |
 
-### Generated Text Samples
+### Generated Text Samples (Word-based)
 
-Below are text samples generated from each Markov chain model:
+Below are text samples generated from each word-based Markov chain model:
 
 **Context Size 1:**
 
-1. `. kawan : kalimantan seulatan china ) dan la ' el republik indonesia . bab manga`
-2. `, bah seulamat nibak watèe jitamong lam data cuaca daerah gunong nyoe geuneuk teuka nibak limong`
-3. `di kabupaten acèh . kawan : lawe sumur nakeuh saboh nanggroe nyang geuseumurat lé shogakkukan .`
+1. `di pidie acèh timu acèh indonesia the colour of life seuneubeuet bak saboh spèsiès nibak takson`
+2. `nakeuh gunong nyoe geupeuteubiet bak wikidata data peumeurèntah nakeuh gunong di teungoh ngon geukeu...`
+3. `bak wikidata data matauroe teubiet teunom di ateuh babah la ôt peunawôt luwa data gunong nyoe`
 
 **Context Size 2:**
 
-1. `kawan : langsa lama langsa timur ( langsa timur ( langsa timur ) nakeuh sidroe aktor asay`
-2. `bak laman nasa data matauroe teubiet & teunom di da ' irah bak laman nasa data matauroe`
-3. `gunong nyoe bak wikidata data cuaca daerah gunong nyoe bak wikidata data cuaca daerah gunong nyoe ba...`
+1. `bak laman sunrisesunset com di acèh seulatan acèh lumbôi gampông nyoe lam data peumeurèntah nakeuh n...`
+2. `gunong nyoe bak laman geonames data gunong nyoe bak laman sunrisesunset com di acèh nakeuh gampông d...`
+3. `nyoe bak wikidata data cuaca daerah gunong nyoe nakeuh kagoshima banda`
 
 **Context Size 3:**
 
-1. `gunong nyoe bak laman geonames data gunong nyoe bak laman geonames data gunong nyoe bak wikidata dat...`
-2. `nè kawan : gampông di pidie jaya kawan : meurah dua , pidie jaya . ngon keudèe panté`
-3. `. nè kawan : gampông di simeulue kawan : teupah teungöh , kabupatèn simeulue , acèh . lumbôi`
+1. `gunong nyoe bak laman geonames data gunong nyoe bak wikidata data cuaca daerah gunong nyoe bak wikid...`
+2. `nyoe bak laman geonames data gunong nyoe bak laman geonames data gunong nyoe bak wikidata data cuaca...`
+3. `lumbôi gampông nyoe lam data peumeurèntah nakeuh nè di acèh rayek kawan ingin jaya acèh rayek nibak ...`
 
 **Context Size 4:**
 
-1. `. nè kawan : gampông di acèh rayek kawan : darussalam , acèh rayek . meunurôt riwayat , kuta`
-2. `gunong nyoe bak laman nasa data matauroe teubiet & teunom di da ' irah bak laman sunrisesunset . com`
-3. `. lumbôi gampông nyoe lam data peumeurèntah nakeuh 11 . 11 . 03 . 06 . 2040 . nè`
+1. `gunong nyoe bak laman nasa data matauroe teubiet teunom di da irah bak laman sunrisesunset com di ac...`
+2. `acèh lumbôi gampông nyoe lam data peumeurèntah nakeuh nè di acèh rayek acèh acèh rayek`
+3. `nyoe lam data peumeurèntah nakeuh nè di bireuen bireuen`
+
+
+### Generated Text Samples (Subword-based)
+
+Below are text samples generated from each subword-based Markov chain model:
+
+**Context Size 1:**
+
+1. `_da_geriè_kahara`
+2. `ata_jeetabam_lab`
+3. `ng_ngeung_teukeu`
+
+**Context Size 2:**
+
+1. `euna_preunomyza_d`
+2. `_nya_-_diet_lis_a`
+3. `h_nak_lam_diversi`
+
+**Context Size 3:**
+
+1. `ng_udeh_nyoe_lam_d`
+2. `_nakeuh_spèsi_acèh`
+3. `_bagiang_bak_lagèe`
+
+**Context Size 4:**
+
+1. `euh_tarèh_seuë_deun`
+2. `bak_encyclopedia_of`
+3. `_di_surat_lé_gosho_`
 
 
 ### Key Findings
 
-- **Best Predictability:** Context-4 with 96.4% predictability
+- **Best Predictability:** Context-4 (word) with 97.6% predictability
 - **Branching Factor:** Decreases with context size (more deterministic)
-- **Memory Trade-off:** Larger contexts require more storage (132,546 contexts)
+- **Memory Trade-off:** Larger contexts require more storage (108,223 contexts)
 - **Recommendation:** Context-3 or Context-4 for text generation
 
 ---
@@ -247,64 +314,64 @@ Below are text samples generated from each Markov chain model:
 
 | Metric | Value |
 |--------|-------|
-| Vocabulary Size | 16,834 |
-| Total Tokens | 576,109 |
-| Mean Frequency | 34.22 |
+| Vocabulary Size | 15,502 |
+| Total Tokens | 515,006 |
+| Mean Frequency | 33.22 |
 | Median Frequency | 3 |
-| Frequency Std Dev | 429.12 |
+| Frequency Std Dev | 415.97 |
 
 ### Most Common Words
 
 | Rank | Word | Frequency |
 |------|------|-----------|
-| 1 | di | 21,413 |
-| 2 | nakeuh | 20,825 |
-| 3 | bak | 18,295 |
-| 4 | acèh | 17,701 |
-| 5 | kawan | 14,998 |
-| 6 | nyoe | 13,193 |
-| 7 | gampông | 12,105 |
-| 8 | gunong | 11,874 |
-| 9 | data | 11,091 |
-| 10 | nyang | 9,069 |
+| 1 | di | 21,196 |
+| 2 | nakeuh | 20,604 |
+| 3 | bak | 18,159 |
+| 4 | acèh | 17,511 |
+| 5 | nyoe | 13,184 |
+| 6 | data | 11,090 |
+| 7 | gunong | 10,023 |
+| 8 | nyang | 9,025 |
+| 9 | gampông | 8,794 |
+| 10 | lam | 7,941 |
 
 ### Least Common Words (from vocabulary)
 
 | Rank | Word | Frequency |
 |------|------|-----------|
-| 1 | own | 2 |
-| 2 | became | 2 |
-| 3 | influence | 2 |
-| 4 | across | 2 |
-| 5 | represent | 2 |
-| 6 | raising | 2 |
-| 7 | ceremony | 2 |
-| 8 | flown | 2 |
-| 9 | reconstructions | 2 |
-| 10 | jawatimu | 2 |
+| 1 | saûdep | 2 |
+| 2 | teuleungah | 2 |
+| 3 | mutuskeun | 2 |
+| 4 | ekshumasi | 2 |
+| 5 | teukeuh | 2 |
+| 6 | dilegalisasikan | 2 |
+| 7 | jendela | 2 |
+| 8 | prosès | 2 |
+| 9 | piazza | 2 |
+| 10 | fontana | 2 |
 
 ### Zipf's Law Analysis
 
 | Metric | Value |
 |--------|-------|
-| Zipf Coefficient | 1.1968 |
-| R² (Goodness of Fit) | 0.996896 |
+| Zipf Coefficient | 1.1704 |
+| R² (Goodness of Fit) | 0.995382 |
 | Adherence Quality | **excellent** |
 
 ### Coverage Analysis
 
 | Top N Words | Coverage |
 |-------------|----------|
-| Top 100 | 61.4% |
-| Top 1,000 | 83.7% |
-| Top 5,000 | 93.9% |
-| Top 10,000 | 97.4% |
+| Top 100 | 63.2% |
+| Top 1,000 | 84.2% |
+| Top 5,000 | 94.2% |
+| Top 10,000 | 97.8% |
 
 ### Key Findings
 
-- **Zipf Compliance:** R²=0.9969 indicates excellent adherence to Zipf's law
-- **High Frequency Dominance:** Top 100 words cover 61.4% of corpus
-- **Long Tail:** 6,834 words needed for remaining 2.6% coverage
+- **Zipf Compliance:** R²=0.9954 indicates excellent adherence to Zipf's law
+- **High Frequency Dominance:** Top 100 words cover 63.2% of corpus
+- **Long Tail:** 5,502 words needed for remaining 2.2% coverage
 
 ---
 ## 5. Word Embeddings Evaluation
@@ -317,24 +384,125 @@ 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** | 6,710 | 32 | 3.746 | 0.875 | 0.5452 🏆 |
-| **mono_64d** | 6,710 | 64 | 3.801 | 0.865 | 0.1802 |
-| **mono_128d** | 6,710 | 128 | 3.849 | 0.857 | 0.0333 |
-| **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 | 0.5172 🏆 | 0.4104 | N/A | N/A |
+| **mono_64d** | 64 | 0.1209 | 0.4362 | N/A | N/A |
+| **mono_128d** | 128 | 0.0271 | 0.4092 | N/A | N/A |
 
 ### Key Findings
 
-- **Best Isotropy:** mono_32d with 0.5452 (more uniform distribution)
-- **Dimension Trade-off:** Higher dimensions capture more semantics but reduce isotropy
-- **Vocabulary Coverage:** All models cover 6,710 words
-- **Recommendation:** 100d for balanced semantic capture and efficiency
+- **Best Isotropy:** mono_32d with 0.5172 (more uniform distribution)
+- **Semantic Density:** Average pairwise similarity of 0.4186. 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 |
+|--------|----------|
+| `-me` | meulagu, meukeunong, meulabôh |
+| `-ge` | geumeuhoi, geupasoe, geupeuresmi |
+| `-geu` | geumeuhoi, geupasoe, geupeuresmi |
+| `-meu` | meulagu, meukeunong, meulabôh |
+| `-pe` | peunuman, peureudee, peumurah |
+
+#### Productive Suffixes
+| Suffix | Examples |
+|--------|----------|
+| `-ng` | meukeunong, gelampang, seberang |
+| `-an` | jonathan, peunuman, kyrgyzstan |
+| `-ah` | bawah, geupeujeulah, jumlah |
+
+### 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 |
+|------|----------|------------------|----------|
+| `eung` | 1.41x | 64 contexts | reung, meung, jeung |
+| `uneu` | 1.70x | 28 contexts | runeu, uneun, seuneu |
+| `euen` | 1.54x | 38 contexts | eueng, meuen, leuen |
+| `euna` | 1.36x | 59 contexts | peuna, beuna, keuna |
+| `ubeu` | 1.47x | 22 contexts | ubeut, neubeu, ubeuet |
+| `umeu` | 1.44x | 23 contexts | jumeu, geumeu, jeumeu |
+| `meur` | 1.63x | 15 contexts | meuri, meurô, meurôn |
+| `anga` | 1.36x | 23 contexts | panga, manga, langa |
+| `teun` | 1.32x | 25 contexts | uteun, ateung, teunga |
+| `neub` | 1.57x | 14 contexts | neuba, neubeu, neubôk |
+| `eube` | 1.48x | 16 contexts | leube, teubee, leubeh |
+| `eune` | 1.63x | 12 contexts | seuneu, geuneu, keuneu |
+
+### 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 |
+|--------|--------|-----------|----------|
+| `-ge` | `-ng` | 56 words | geupeutrang, geudông |
+| `-pe` | `-an` | 51 words | penyiaran, permukaan |
+| `-me` | `-ng` | 40 words | meulinteueng, meuhubông |
+| `-pe` | `-ng` | 22 words | perang, peukeumang |
+| `-pe` | `-ah` | 18 words | peujeunajah, peuleumah |
+| `-ge` | `-ah` | 17 words | geupeuglah, geupeuluwah |
+| `-me` | `-ah` | 16 words | meujumeulah, meurah |
+| `-me` | `-an` | 10 words | meridian, meukeujadian |
+| `-ge` | `-an` | 6 words | geurakan, geuritan |
+
+### 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 |
+|------|-----------------|------------|------|
+| geumeudong | **`geu-meu-dong`** | 6.0 | `dong` |
+| geumeututô | **`geu-meu-tutô`** | 6.0 | `tutô` |
+| meubileueng | **`meu-bileue-ng`** | 6.0 | `bileue` |
+| geulumbang | **`geu-lumba-ng`** | 6.0 | `lumba` |
+| geumeupakat | **`geu-meu-pakat`** | 6.0 | `pakat` |
+| geumeuniaga | **`geu-meu-niaga`** | 6.0 | `niaga` |
+| geumeuturi | **`geu-meu-turi`** | 6.0 | `turi` |
+| geuseubarô | **`geu-seubarô`** | 4.5 | `seubarô` |
+| geudapeuta | **`geu-dapeuta`** | 4.5 | `dapeuta` |
+| meusampoe | **`meu-sampoe`** | 4.5 | `sampoe` |
+| geubayeuë | **`geu-bayeuë`** | 4.5 | `bayeuë` |
+| meulingka | **`meu-lingka`** | 4.5 | `lingka` |
+| meusiyasat | **`meu-siyasat`** | 4.5 | `siyasat` |
+| meulaksana | **`meu-laksana`** | 4.5 | `laksana` |
+| geubayeue | **`geu-bayeue`** | 4.5 | `bayeue` |
+
+### 6.6 Linguistic Interpretation
+
+> **Automated Insight:**
+The language ACE 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)
 
@@ -342,11 +510,12 @@ Below are text samples generated from each Markov chain model:
 
 | Component | Recommended | Rationale |
 |-----------|-------------|-----------|
-| Tokenizer | **32k BPE** | Best compression (4.81x) with low UNK rate |
-| N-gram | **5-gram** | Lowest perplexity (261) |
-| Markov | **Context-4** | Highest predictability (96.4%) |
+| Tokenizer | **64k BPE** | Best compression (4.92x) |
+| N-gram | **2-gram** | Lowest perplexity (224) |
+| Markov | **Context-4** | Highest predictability (97.6%) |
 | Embeddings | **100d** | Balanced semantic capture and isotropy |
 
+
 ---
 ## Appendix: Metrics Glossary & Interpretation Guide
 
@@ -536,7 +705,8 @@ 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}
 }
@@ -552,7 +722,8 @@ 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: 2025-12-27 04:33:03*
+*Report Date: 2026-01-03 05:05:30*

models/embeddings/monolingual/ace_128d.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2046e3b06e8436b76db682bbcbcff52a79ae03beb79e02a5c143feb4aa30e845
-size 1030981956
+oid sha256:0cf768b8cc26ee028b74fbc795a0084120d5ec9fa9ee7265d5277f3228676402
+size 1030413699

models/embeddings/monolingual/ace_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": 6710
+  "vocab_size": 6165
 }

models/embeddings/monolingual/ace_32d.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:865bc53096b61029d31129feb2421f97605a4e1b1fc71a16f09ae2a11377dc30
-size 257828676
+oid sha256:b69885b365da465ab4bdfb040ee28e5b1f011ab235bd18a0e87c25eae3925f8f
+size 257678979

models/embeddings/monolingual/ace_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": 6710
+  "vocab_size": 6165
 }

models/embeddings/monolingual/ace_64d.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e407ad194bf72cb228e2d6744a4203d66d913abff54ac372bb7ee406098236c3
-size 515546436
+oid sha256:79de1c214df5334f0fa005b795ccfce46a4ac070f1ea7654cfaab5c2dc34dda5
+size 515257219

models/embeddings/monolingual/ace_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": 6710
+  "vocab_size": 6165
 }

models/subword_markov/ace_markov_ctx1_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:994945c1c8459c3679e3e00e0210f36d27f7b92323ac2b7120e6d2efd5d2d126
-size 61873
+oid sha256:0f3a9abf57afb060a211a69e96c23c22c115a944e460bb459448be1adea274b0
+size 59763

models/subword_markov/ace_markov_ctx1_subword_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 1,
   "variant": "subword",
   "language": "ace",
-  "unique_contexts": 1076,
-  "total_transitions": 3908719
+  "unique_contexts": 1269,
+  "total_transitions": 3470466
 }

models/subword_markov/ace_markov_ctx2_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c7276dd9bca94f9f03917ff721ff8bf88482f94093dfa6019b62501feb464811
-size 312446
+oid sha256:d76a3d707c4ae33b25f22e9cbdcb78745638dfcdc99ed80df41d977cbb81cca0
+size 268452

models/subword_markov/ace_markov_ctx2_subword_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 2,
   "variant": "subword",
   "language": "ace",
-  "unique_contexts": 7655,
-  "total_transitions": 3895602
+  "unique_contexts": 6822,
+  "total_transitions": 3457577
 }

models/subword_markov/ace_markov_ctx3_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fa959fe08fc66a84cabcff2aea19b15147f3ed325699ecddb1034ee1e56ba332
-size 1020456
+oid sha256:652d2af547c89ae398ecf44a87cac49a050a3836bbeb328bda5d380d30c7c8c0
+size 900905

models/subword_markov/ace_markov_ctx3_subword_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 3,
   "variant": "subword",
   "language": "ace",
-  "unique_contexts": 37427,
-  "total_transitions": 3882485
+  "unique_contexts": 30615,
+  "total_transitions": 3444688
 }

models/subword_markov/ace_markov_ctx4_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0ae4f621898a6636e9b6c7126e341387c206ed707f6ed0fba958ebfe9fc9fbb4
-size 2470198
+oid sha256:32adbfca2120175fa43f1b0bd2cedc5d7bd19099677319e6128cab29e66007d4
+size 2085174

models/subword_markov/ace_markov_ctx4_subword_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 4,
   "variant": "subword",
   "language": "ace",
-  "unique_contexts": 132546,
-  "total_transitions": 3869368
+  "unique_contexts": 108223,
+  "total_transitions": 3431799
 }

models/subword_ngram/ace_2gram_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4a5e9b80103fed58ac45a1a1dc5018d7bf762a47692e47d327a50b683698f3ad
-size 36523
+oid sha256:0e6d83ed77f7cb491e9f8059b986f03ae648a2fb694c94ba29ac349e3055f671
+size 30955

models/subword_ngram/ace_2gram_subword_metadata.json

@@ -2,6 +2,6 @@
   "n": 2,
   "variant": "subword",
   "language": "ace",
-  "unique_ngrams": 2689,
-  "total_ngrams": 3908719
+  "unique_ngrams": 2204,
+  "total_ngrams": 3470466
 }

models/subword_ngram/ace_3gram_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8d4bacbf3031b3897258cf39672bf0992a78d116a540da2c29da63b6fe9f5442
-size 218763
+oid sha256:5f58aa5cf358032d5b4d7a52f7a9b81c4a9bcf9dda5c142a65fdae14ae7135f4
+size 178456

models/subword_ngram/ace_3gram_subword_metadata.json

@@ -2,6 +2,6 @@
   "n": 3,
   "variant": "subword",
   "language": "ace",
-  "unique_ngrams": 17865,
-  "total_ngrams": 3895602
+  "unique_ngrams": 14605,
+  "total_ngrams": 3457577
 }

models/subword_ngram/ace_4gram_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1535b59d3ffbe8075678699c37413a180fe8e0b10d9a35c6873de3dbef8e23f7
-size 844170
+oid sha256:529d84e3eda9c7669ad3732639f36554c4e26803ae7ddfb8b4ca0fbcc326fbe7
+size 709589

models/subword_ngram/ace_4gram_subword_metadata.json

@@ -2,6 +2,6 @@
   "n": 4,
   "variant": "subword",
   "language": "ace",
-  "unique_ngrams": 71115,
-  "total_ngrams": 3882485
+  "unique_ngrams": 59251,
+  "total_ngrams": 3444688
 }

models/tokenizer/ace_tokenizer_16k.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e0486832283fb4af24e9a049fd7267139c87b145310f599caa110fb2824272a2
-size 502107
+oid sha256:772eccdeefe417a39cb6d9efddc039fe1535ce5b6b9728ed7a7a9f133127fb6c
+size 503799

models/tokenizer/ace_tokenizer_32k.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:68ba228d17ef3b883e0874b7c05972fa9a2da79bae088ee77a799ec8ba684117
-size 780590
+oid sha256:1fc76a32a006f73f304c48abbc3874f1bbba66904ea507a5989c19d84064eb2b
+size 784975

models/tokenizer/ace_tokenizer_64k.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6eaefce9e3832566d4b03402b226a5055d116d560a4b6f5c9d79e2f9014b5be2
-size 1365161
+oid sha256:638ca8a1f7bfb0862f2a4e2f334f6a0b19cd0775fcc9ea20c4d6f11edfca90d1
+size 1328218

models/tokenizer/ace_tokenizer_8k.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8afabfa554db88a64b6659881e7d87e20f7717006505bdc6bbb199898a6e53a2
-size 371036
+oid sha256:4f2cf6a3125500fbc0a6568259881ab1bdc31c6fd6b01f32fd845acf44e03d4a
+size 371130

models/vocabulary/ace_vocabulary.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0cdcdc6259363ef9c7e4d76eeee62ba0c6f4da7b3a86d0e338193a83e61e901a
-size 279594
+oid sha256:2ba6b44c1dbf28322c8da89370cfc09afb1c3724b0f3302b03929700e5dc3332
+size 251932

models/vocabulary/ace_vocabulary_metadata.json

@@ -1,16 +1,17 @@
 {
   "language": "ace",
-  "vocabulary_size": 16834,
+  "vocabulary_size": 15502,
+  "variant": "full",
   "statistics": {
-    "type_token_ratio": 0.06473562926222352,
+    "type_token_ratio": 0.06761988315009426,
     "coverage": {
-      "top_100": 0.591821547661651,
-      "top_1000": 0.8065606887117759,
-      "top_5000": 0.9046783703966313,
-      "top_10000": 0.9388346884375219
+      "top_100": 0.6077445728258638,
+      "top_1000": 0.8095215873667706,
+      "top_5000": 0.9057864969294234,
+      "top_10000": 0.9405017452821384
     },
-    "hapax_count": 21877,
-    "hapax_ratio": 0.5651365245020795,
-    "total_documents": 13117
+    "hapax_count": 20724,
+    "hapax_ratio": 0.5720753050295369,
+    "total_documents": 12889
   }
 }

models/word_markov/ace_markov_ctx1_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:064d6047862cf68da2b906472e39980d8a5b9be018d078f381c09a061281db10
-size 1341704
+oid sha256:f453896eb7f6ff049a8bcdf5e17901d47f42c1057f32a9f36b112d6d94fd70e3
+size 1218673

models/word_markov/ace_markov_ctx1_word_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 1,
   "variant": "word",
   "language": "ace",
-  "unique_contexts": 38800,
-  "total_transitions": 712935
+  "unique_contexts": 36025,
+  "total_transitions": 522841
 }

models/word_markov/ace_markov_ctx2_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:80655ab7d24dd0562f6dd00bf8454b23ac16d8f6145ab9327656208ee794c040
-size 2960527
+oid sha256:5f80eb40706cfeda9139c3371c96005f0fea0733f1ee4c3a2e5eecc815929a6a
+size 2625141

models/word_markov/ace_markov_ctx2_word_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 2,
   "variant": "word",
   "language": "ace",
-  "unique_contexts": 172855,
-  "total_transitions": 699818
+  "unique_contexts": 155224,
+  "total_transitions": 509952
 }

models/word_markov/ace_markov_ctx3_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c713caa0041630e45e31c03c7212c262abebdd3fc3bd44d99b0c957bc7aa8187
-size 4205022
+oid sha256:6af7f232b9a0d1fc8f7b6bc13752d652f933bd8f13f2d7ddea846f7703f26706
+size 3563465

models/word_markov/ace_markov_ctx3_word_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 3,
   "variant": "word",
   "language": "ace",
-  "unique_contexts": 269187,
-  "total_transitions": 686702
+  "unique_contexts": 221018,
+  "total_transitions": 497063
 }

models/word_markov/ace_markov_ctx4_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:26e6fd733dee534e51fd8ce9dac5ca06b9de38d8fcabd817bd83cb1b54510e03
-size 4916810
+oid sha256:09c8a9f9ca1bfa2918c4e7bf8d8469899049c695aaa95d251654d321f8974772
+size 4102971

models/word_markov/ace_markov_ctx4_word_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 4,
   "variant": "word",
   "language": "ace",
-  "unique_contexts": 309167,
-  "total_transitions": 673586
+  "unique_contexts": 242720,
+  "total_transitions": 484174
 }

models/word_ngram/ace_2gram_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:86dcd174e63bb4d8107dc7bd51314f22ac64b7d467c5f885bf2addb1e9905a3f
-size 136829
+oid sha256:7b153c8f5fd67123746ac3b5aa53610477c2eb5d60b0e38122617038ce36c8dc
+size 98124

models/word_ngram/ace_2gram_word_metadata.json

@@ -2,6 +2,6 @@
   "n": 2,
   "variant": "word",
   "language": "ace",
-  "unique_ngrams": 9992,
-  "total_ngrams": 712935
+  "unique_ngrams": 7009,
+  "total_ngrams": 522841
 }

models/word_ngram/ace_3gram_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:97bce04690e426c1a9f5c64e762c9e3a260299f30e22a6cee8f161a141a80787
-size 209506
+oid sha256:9443d067941cb6c518bc87ed1b98c16dd4ecef5d8326cb06122d012ffb5bb1b3
+size 127588

models/word_ngram/ace_3gram_word_metadata.json

@@ -2,6 +2,6 @@
   "n": 3,
   "variant": "word",
   "language": "ace",
-  "unique_ngrams": 14167,
-  "total_ngrams": 699818
+  "unique_ngrams": 8214,
+  "total_ngrams": 509952
 }

models/word_ngram/ace_4gram_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:31a6dcf0f123c74ce564b8bc022d3cfcb9f30465f679ad41e38919ddffdd61c9
-size 368875
+oid sha256:27a652701956d4deaac999e0b5e9993649004a051bfd158f4efd609b2f52f048
+size 209749

models/word_ngram/ace_4gram_word_metadata.json

@@ -2,6 +2,6 @@
   "n": 4,
   "variant": "word",
   "language": "ace",
-  "unique_ngrams": 24432,
-  "total_ngrams": 686702
+  "unique_ngrams": 12805,
+  "total_ngrams": 497063
 }