Upload all models and assets for arz (20251001)

README.md

@@ -23,14 +23,14 @@ dataset_info:
 metrics:
   - name: best_compression_ratio
     type: compression
-    value: 4.023
+    value: 3.905
   - name: best_isotropy
     type: isotropy
-    value: 0.8067
+    value: 0.7897
   - name: vocabulary_size
     type: vocab
-    value: 1000000
-generated: 2025-12-27
+    value: 0
+generated: 2026-01-03
 ---
 
 # Egyptian Arabic - 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)
 
@@ -68,64 +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** | 3.017x | 2.93 | 0.4317% | 1,742,125 |
-| **16k** | 3.365x | 3.27 | 0.4814% | 1,562,212 |
-| **32k** | 3.703x | 3.60 | 0.5297% | 1,419,540 |
-| **64k** | 4.023x 🏆 | 3.91 | 0.5755% | 1,306,653 |
+| **8k** | 2.876x | 2.88 | 0.8210% | 1,709,035 |
+| **16k** | 3.215x | 3.22 | 0.9180% | 1,528,463 |
+| **32k** | 3.559x | 3.56 | 1.0163% | 1,380,735 |
+| **64k** | 3.905x 🏆 | 3.91 | 1.1149% | 1,258,558 |
 
 ### Tokenization Examples
 
 Below are sample sentences tokenized with each vocabulary size:
 
-**Sample 1:** `دونج سارى فنان من كامبوديا.
-
- حياته 
-دونج سارى من مواليد يوم 1 يناير 1957.
-
- لين...`
+**Sample 1:** `تاملكوت هوا دوار فى المغرب. المكان تاملكوت موجود فى منطقه اداريه اسمها تماسين. س...`
 
 | Vocab | Tokens | Count |
 |-------|--------|-------|
-| 8k | `▁دونج ▁س ارى ▁فنان ▁من ▁كامب ود يا . ▁حياته ... (+28 more)` | 38 |
-| 16k | `▁دونج ▁س ارى ▁فنان ▁من ▁كامب وديا . ▁حياته ▁دونج ... (+26 more)` | 36 |
-| 32k | `▁دونج ▁سارى ▁فنان ▁من ▁كامبوديا . ▁حياته ▁دونج ▁سارى ▁من ... (+22 more)` | 32 |
-| 64k | `▁دونج ▁سارى ▁فنان ▁من ▁كامبوديا . ▁حياته ▁دونج ▁سارى ▁من ... (+22 more)` | 32 |
-
-**Sample 2:** `بيريباروس ( الاسم العلمى: Periparus ) هوا جنس من الطيور بيتبع قرقفيات.
+| 8k | `▁تام لك وت ▁هوا ▁دوار ▁فى ▁المغرب . ▁المك ان ... (+24 more)` | 34 |
+| 16k | `▁تام لك وت ▁هوا ▁دوار ▁فى ▁المغرب . ▁المك ان ... (+24 more)` | 34 |
+| 32k | `▁تام لك وت ▁هوا ▁دوار ▁فى ▁المغرب . ▁المكان ▁تام ... (+23 more)` | 33 |
+| 64k | `▁تام لك وت ▁هوا ▁دوار ▁فى ▁المغرب . ▁المكان ▁تام ... (+23 more)` | 33 |
 
- لينكات ...`
+**Sample 2:** `جيريمى ديفيدسون مخرج افلام من امريكا. حياته جيريمى ديفيدسون من مواليد يوم 24 ديس...`
 
 | Vocab | Tokens | Count |
 |-------|--------|-------|
-| 8k | `▁بير يب اروس ▁( ▁الاسم ▁العلم ى : ▁per ip ... (+36 more)` | 46 |
-| 16k | `▁بير يب اروس ▁( ▁الاسم ▁العلمى : ▁per ip arus ... (+31 more)` | 41 |
-| 32k | `▁بير يب اروس ▁( ▁الاسم ▁العلمى : ▁per ip arus ... (+27 more)` | 37 |
-| 64k | `▁بير يب اروس ▁( ▁الاسم ▁العلمى : ▁per ip arus ... (+26 more)` | 36 |
-
-**Sample 3:** `قلى بيغلو ( بالفارسى قلیبیگلو ) قريه فى ايران.
-
- لينكات برانيه 
+| 8k | `▁جير يمى ▁ديفيد سون ▁مخرج ▁افلام ▁من ▁امريكا . ▁حياته ... (+23 more)` | 33 |
+| 16k | `▁جيريمى ▁ديفيد سون ▁مخرج ▁افلام ▁من ▁امريكا . ▁حياته ▁جيريمى ... (+21 more)` | 31 |
+| 32k | `▁جيريمى ▁ديفيدسون ▁مخرج ▁افلام ▁من ▁امريكا . ▁حياته ▁جيريمى ▁ديفيدسون ... (+19 more)` | 29 |
+| 64k | `▁جيريمى ▁ديفيدسون ▁مخرج ▁افلام ▁من ▁امريكا . ▁حياته ▁جيريمى ▁ديفيدسون ... (+19 more)` | 29 |
 
-
- سبب التسميه 
-...`
+**Sample 3:** `ابهينايا ممثله من الهند. حياتها ابهينايا من مواليد يوم 13 نوفمبر سنة فى كارناتاك...`
 
 | Vocab | Tokens | Count |
 |-------|--------|-------|
-| 8k | `▁ق لى ▁بي غل و ▁( ▁بالفارسى ▁قل ی ب ... (+21 more)` | 31 |
-| 16k | `▁ق لى ▁بي غلو ▁( ▁بالفارسى ▁قل ی ب ی ... (+20 more)` | 30 |
-| 32k | `▁ق لى ▁بي غلو ▁( ▁بالفارسى ▁قل ی بی گ ... (+19 more)` | 29 |
-| 64k | `▁ق لى ▁بي غلو ▁( ▁بالفارسى ▁قل ی بی گ ... (+19 more)` | 29 |
+| 8k | `▁اب ه ينا يا ▁ممثله ▁من ▁الهند . ▁حياتها ▁اب ... (+28 more)` | 38 |
+| 16k | `▁اب ه ينا يا ▁ممثله ▁من ▁الهند . ▁حياتها ▁اب ... (+27 more)` | 37 |
+| 32k | `▁ابه ينا يا ▁ممثله ▁من ▁الهند . ▁حياتها ▁ابه ينا ... (+25 more)` | 35 |
+| 64k | `▁ابه ينا يا ▁ممثله ▁من ▁الهند . ▁حياتها ▁ابه ينا ... (+24 more)` | 34 |
 
 
 ### Key Findings
 
-- **Best Compression:** 64k achieves 4.023x compression
-- **Lowest UNK Rate:** 8k with 0.4317% unknown tokens
+- **Best Compression:** 64k achieves 3.905x compression
+- **Lowest UNK Rate:** 8k with 0.8210% unknown tokens
 - **Trade-off:** Larger vocabularies improve compression but increase model size
 - **Recommendation:** 32k vocabulary provides optimal balance for production use
 
@@ -134,57 +129,89 @@ Below are sample sentences tokenized with each vocabulary size:
 
 ![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** | 6,568 🏆 | 12.68 | 1,353,863 | 27.8% | 64.6% |
-| **2-gram** | 384 🏆 | 8.58 | 15,875 | 59.5% | 97.6% |
-| **3-gram** | 11,727 | 13.52 | 2,306,233 | 23.3% | 59.5% |
-| **3-gram** | 2,432 | 11.25 | 170,638 | 29.7% | 70.0% |
-| **4-gram** | 21,664 | 14.40 | 4,880,347 | 21.3% | 54.9% |
-| **4-gram** | 8,554 | 13.06 | 1,045,558 | 20.0% | 54.3% |
+| N-gram | Variant | Perplexity | Entropy | Unique N-grams | Top-100 Coverage | Top-1000 Coverage |
+|--------|---------|------------|---------|----------------|------------------|-------------------|
+| **2-gram** | Word | 5,793 | 12.50 | 1,073,861 | 30.2% | 66.4% |
+| **2-gram** | Subword | 316 🏆 | 8.30 | 15,451 | 62.6% | 98.6% |
+| **3-gram** | Word | 8,299 | 13.02 | 1,682,809 | 28.5% | 62.7% |
+| **3-gram** | Subword | 2,021 | 10.98 | 129,923 | 30.1% | 74.0% |
+| **4-gram** | Word | 12,842 | 13.65 | 3,054,922 | 27.3% | 59.4% |
+| **4-gram** | Subword | 7,215 | 12.82 | 788,718 | 19.6% | 56.9% |
 
 ### Top 5 N-grams by Size
 
-**2-grams:**
+**2-grams (Word):**
+
+| Rank | N-gram | Count |
+|------|--------|-------|
+| 1 | `لينكات برانيه` | 1,293,684 |
+| 2 | `برانيه مصادر` | 1,167,581 |
+| 3 | `من مواليد` | 829,322 |
+| 4 | `مواليد يوم` | 809,177 |
+| 5 | `الاستوا السماوى` | 668,876 |
+
+**3-grams (Word):**
 
 | Rank | N-gram | Count |
 |------|--------|-------|
-| 1 | `تصنيف :` | 4,787,649 |
-| 2 | `مصادر تصنيف` | 1,597,448 |
-| 3 | `لينكات برانيه` | 1,294,464 |
-| 4 | `برانيه مصادر` | 1,167,374 |
-| 5 | `من مواليد` | 829,312 |
+| 1 | `لينكات برانيه مصادر` | 1,164,952 |
+| 2 | `من مواليد يوم` | 809,029 |
+| 3 | `خط الاستوا السماوى` | 630,228 |
+| 4 | `الدايره الساعيه لجرم` | 445,892 |
+| 5 | `الساعيه لجرم سماوى` | 445,892 |
 
-**3-grams:**
+**4-grams (Word):**
 
 | Rank | N-gram | Count |
 |------|--------|-------|
-| 1 | `مصادر تصنيف :` | 1,597,447 |
-| 2 | `برانيه مصادر تصنيف` | 1,165,016 |
-| 3 | `لينكات برانيه مصادر` | 1,164,743 |
-| 4 | `من مواليد يوم` | 809,006 |
-| 5 | `. المطلع المستقيم` | 668,795 |
+| 1 | `الدايره الساعيه لجرم سماوى` | 445,892 |
+| 2 | `السماوى تكون قيمة بعده` | 445,860 |
+| 3 | `خط الاستوا السماوى تكون` | 445,860 |
+| 4 | `الاستوا السماوى تكون قيمة` | 445,860 |
+| 5 | `لينكات برانيه مصادر من` | 320,727 |
 
-**4-grams:**
+**2-grams (Subword):**
 
 | Rank | N-gram | Count |
 |------|--------|-------|
-| 1 | `برانيه مصادر تصنيف :` | 1,165,016 |
-| 2 | `لينكات برانيه مصادر تصنيف` | 1,162,419 |
-| 3 | `. لينكات برانيه مصادر` | 558,227 |
-| 4 | `الدايره الساعيه لجرم سماوى` | 445,892 |
-| 5 | `خط الاستوا السماوى تكون` | 445,860 |
+| 1 | `_ ا` | 31,144,333 |
+| 2 | `ا ل` | 30,224,243 |
+| 3 | `ه _` | 17,180,633 |
+| 4 | `_ م` | 13,559,836 |
+| 5 | `ى _` | 11,805,719 |
+
+**3-grams (Subword):**
+
+| Rank | N-gram | Count |
+|------|--------|-------|
+| 1 | `_ ا ل` | 25,116,125 |
+| 2 | `ي ه _` | 6,396,587 |
+| 3 | `ه _ ا` | 6,346,797 |
+| 4 | `ا ل م` | 5,946,692 |
+| 5 | `_ م ن` | 4,537,386 |
+
+**4-grams (Subword):**
+
+| Rank | N-gram | Count |
+|------|--------|-------|
+| 1 | `ه _ ا ل` | 5,297,759 |
+| 2 | `_ ا ل م` | 5,200,038 |
+| 3 | `_ ف ى _` | 4,251,301 |
+| 4 | `_ م ن _` | 3,906,606 |
+| 5 | `_ ا ل ا` | 3,578,656 |
 
 
 ### Key Findings
 
-- **Best Perplexity:** 2-gram with 384
+- **Best Perplexity:** 2-gram (subword) with 316
 - **Entropy Trend:** Decreases with larger n-grams (more predictable)
-- **Coverage:** Top-1000 patterns cover ~54% of corpus
+- **Coverage:** Top-1000 patterns cover ~57% of corpus
 - **Recommendation:** 4-gram or 5-gram for best predictive performance
 
 ---
@@ -192,55 +219,86 @@ Below are sample sentences tokenized with each vocabulary size:
 
 ![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** | 0.8482 | 1.800 | 6.43 | 2,177,526 | 15.2% |
-| **1** | 1.2448 | 2.370 | 9.40 | 4,455 | 0.0% |
-| **2** | 0.3675 | 1.290 | 2.04 | 13,972,922 | 63.2% |
-| **2** | 0.9830 | 1.977 | 7.64 | 41,887 | 1.7% |
-| **3** | 0.1212 | 1.088 | 1.34 | 28,492,247 | 87.9% |
-| **3** | 0.9327 | 1.909 | 5.31 | 320,101 | 6.7% |
-| **4** | 0.0720 🏆 | 1.051 | 1.20 | 38,092,155 | 92.8% |
-| **4** | 0.7558 🏆 | 1.689 | 3.71 | 1,700,014 | 24.4% |
+| Context | Variant | Avg Entropy | Perplexity | Branching Factor | Unique Contexts | Predictability |
+|---------|---------|-------------|------------|------------------|-----------------|----------------|
+| **1** | Word | 1.2217 | 2.332 | 9.13 | 1,353,062 | 0.0% |
+| **1** | Subword | 1.0533 | 2.075 | 8.28 | 5,726 | 0.0% |
+| **2** | Word | 0.3648 | 1.288 | 1.91 | 12,336,484 | 63.5% |
+| **2** | Subword | 0.7848 | 1.723 | 5.54 | 47,379 | 21.5% |
+| **3** | Word | 0.1139 | 1.082 | 1.28 | 23,517,673 | 88.6% |
+| **3** | Subword | 0.7673 | 1.702 | 4.73 | 262,420 | 23.3% |
+| **4** | Word | 0.0625 🏆 | 1.044 | 1.17 | 29,894,419 | 93.7% |
+| **4** | Subword | 0.7433 | 1.674 | 3.81 | 1,241,425 | 25.7% |
 
-### 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. `. شوف كمان رواد الشركات فى كليه سانت لورانس و كيكرز . 0 . السرعه الشعاعيه`
-2. `: مهندسين من تشيكوسلوفاكيا . hubble census finds galaxies at the ottoman empire and historical demog...`
-3. `تصنيف : anguilla australis ) و خط الاستوا السماوى تكون قيمة بعده بالموجب ( ) بيتر`
+1. `فى العالم حسب المساحه لستة اكبر بحيرات اوروبا لينكات مصادر من مملكه ايطاليا حياته الرياضيه بيلعب`
+2. `من مواليد يوم 16 يناير سنة فى ذا ماتشيس بتقدم الانواع الفنيه كانت دى لوبو من`
+3. `و بتنقاس بالانزياح الاحمر المطلع المستقيم ممكن يتقاس بقوس دايره الاستواء السماويه من الجرى و نادى`
 
 **Context Size 2:**
 
-1. `تصنيف : لاعب كريكت من دومينيكا . لينكات برانيه مصادر تصنيف : قسيس كاتوليك من مملكة نيديرلاند`
-2. `مصادر تصنيف : مسطحات مائيه فى كندا . جغرافيا نهر وايتاهايا بيصب فى نهر اوبى . اقرا`
-3. `لينكات برانيه مصادر تصنيف : ناس تصنيف : عالم لاهوت من المانيا ) جورج فرانسيس ماكجينيس من`
+1. `لينكات برانيه مصادر عجل ناريه من المانيا حياته اليكساندر انتونوڤيتش ريزونى اليكساندر انستروثير اليكس...`
+2. `برانيه مصادر كوره قدم من الميكسيك حياته اڤير كاباليرو اڤيرالدو فيريرا لاعب كورة قدم من اليابان حياته`
+3. `من مواليد يوم 19 اغسطس لسا عايشين فى استانبول لينكات برانيه مصادر هوكى الجليد من امريكا حياته`
 
 **Context Size 3:**
 
-1. `مصادر تصنيف : موسم رياضى فى كورة قدم اتعمل فى پورتوجال سنة 2007 . لينكات برانيه مصادر تصنيف`
-2. `برانيه مصادر تصنيف : ممثلين من البرازيل تصنيف : متعلمين فى جامعة ڤيرچينيا كومونولث و جامعة ڤيرچينيا ...`
-3. `لينكات برانيه مصادر تصنيف : سياسيين تصنيف : سياسيين تصنيف : سياسيين من تشيكيا تصنيف : ناس لسا`
+1. `لينكات برانيه مصادر سكان سكان فى ايران المكان ادم درهسى عليا adam darrehsi ye olya هيا تجمع سكان`
+2. `من مواليد يوم 7 ديسمبر فى مونتفيدو الحياه الرياضيه بيلعب فى مركز مُدَافِع و لعب مع فريق ريال`
+3. `خط الاستوا السماوى تكون قيمة بعده بالموجب و لو النجم جنوب خط الاستوا السماوى لو كان النجم شمال`
 
 **Context Size 4:**
 
-1. `برانيه مصادر تصنيف : قرى اليمن تصنيف : قرى يافع`
-2. `لينكات برانيه مصادر تصنيف : مغنيين تصنيف : مغنيين امريكان تصنيف : منتجين تيليڤزيون تصنيف : مخرجين تي...`
-3. `. لينكات برانيه مصادر تصنيف : مسطحات مائيه فى فينلاندا`
+1. `الدايره الساعيه لجرم سماوى و الدايره الساعيه لنقطة الاعتدال الربيعى المطلع المستقيم ممكن يتقاس بقوس ...`
+2. `الاستوا السماوى تكون قيمة بعده بالسالب مصادر كوكبه`
+3. `السماوى تكون قيمة بعده بالسالب مصادر 2ماس كوكبه`
+
+
+### Generated Text Samples (Subword-based)
+
+Below are text samples generated from each subword-based Markov chain model:
+
+**Context Size 1:**
+
+1. `_مرا_ارو_لسيه_س_`
+2. `الدريه_اثر_كالال`
+3. `لخطونالمطة_جو_عب`
+
+**Context Size 2:**
+
+1. `_الكريتالسمات_فى_`
+2. `اليه_عاعيه_مطلحجم`
+3. `ه_بقه_ليكا_بقوى_ا`
+
+**Context Size 3:**
+
+1. `_المستقيم_محمد_بيس`
+2. `يه_مصادر_كورة_قدم_`
+3. `ه_العقبت_برات_السم`
+
+**Context Size 4:**
+
+1. `ه_السماوى_مع_فريق_ن`
+2. `_المكافئ_الفلك._الم`
+3. `_فى_باردوه_مصادر_اس`
 
 
 ### Key Findings
 
-- **Best Predictability:** Context-4 with 92.8% predictability
+- **Best Predictability:** Context-4 (word) with 93.7% predictability
 - **Branching Factor:** Decreases with context size (more deterministic)
-- **Memory Trade-off:** Larger contexts require more storage (1,700,014 contexts)
+- **Memory Trade-off:** Larger contexts require more storage (1,241,425 contexts)
 - **Recommendation:** Context-3 or Context-4 for text generation
 
 ---
@@ -256,64 +314,64 @@ Below are text samples generated from each Markov chain model:
 
 | Metric | Value |
 |--------|-------|
-| Vocabulary Size | 1,000,000 |
-| Total Tokens | 132,694,568 |
-| Mean Frequency | 132.69 |
-| Median Frequency | 5 |
-| Frequency Std Dev | 10016.26 |
+| Vocabulary Size | 856,070 |
+| Total Tokens | 116,711,182 |
+| Mean Frequency | 136.33 |
+| Median Frequency | 4 |
+| Frequency Std Dev | 9391.59 |
 
 ### Most Common Words
 
 | Rank | Word | Frequency |
 |------|------|-----------|
-| 1 | تصنيف | 4,791,431 |
-| 2 | فى | 4,424,323 |
-| 3 | من | 3,918,728 |
-| 4 | و | 3,519,108 |
-| 5 | مصادر | 1,613,116 |
-| 6 | لينكات | 1,360,026 |
-| 7 | برانيه | 1,299,627 |
-| 8 | مواليد | 1,130,480 |
-| 9 | هيا | 1,062,827 |
-| 10 | اللى | 967,453 |
+| 1 | فى | 4,414,661 |
+| 2 | من | 3,909,776 |
+| 3 | و | 3,512,508 |
+| 4 | مصادر | 1,612,463 |
+| 5 | لينكات | 1,359,404 |
+| 6 | برانيه | 1,298,834 |
+| 7 | هيا | 1,062,266 |
+| 8 | اللى | 965,103 |
+| 9 | يوم | 853,034 |
+| 10 | مواليد | 836,295 |
 
 ### Least Common Words (from vocabulary)
 
 | Rank | Word | Frequency |
 |------|------|-----------|
-| 1 | 1513504 | 2 |
-| 2 | 412321 | 2 |
-| 3 | j08004409 | 2 |
-| 4 | 1545353 | 2 |
-| 5 | 1073083 | 2 |
-| 6 | 0566528 | 2 |
-| 7 | 2103522 | 2 |
-| 8 | 1579797 | 2 |
-| 9 | 2051684 | 2 |
-| 10 | 600674 | 2 |
+| 1 | algeriens | 2 |
+| 2 | وبتينا | 2 |
+| 3 | روتلُف | 2 |
+| 4 | bouabdellah | 2 |
+| 5 | الخُضرة | 2 |
+| 6 | impressionisms | 2 |
+| 7 | assyriaca | 2 |
+| 8 | جروكبيديا | 2 |
+| 9 | grokipedia | 2 |
+| 10 | grok | 2 |
 
 ### Zipf's Law Analysis
 
 | Metric | Value |
 |--------|-------|
-| Zipf Coefficient | 1.2848 |
-| R² (Goodness of Fit) | 0.995303 |
+| Zipf Coefficient | 1.2602 |
+| R² (Goodness of Fit) | 0.994644 |
 | Adherence Quality | **excellent** |
 
 ### Coverage Analysis
 
 | Top N Words | Coverage |
 |-------------|----------|
-| Top 100 | 44.5% |
-| Top 1,000 | 75.8% |
-| Top 5,000 | 85.7% |
-| Top 10,000 | 88.8% |
+| Top 100 | 46.0% |
+| Top 1,000 | 76.7% |
+| Top 5,000 | 85.9% |
+| Top 10,000 | 89.0% |
 
 ### Key Findings
 
-- **Zipf Compliance:** R²=0.9953 indicates excellent adherence to Zipf's law
-- **High Frequency Dominance:** Top 100 words cover 44.5% of corpus
-- **Long Tail:** 990,000 words needed for remaining 11.2% coverage
+- **Zipf Compliance:** R²=0.9946 indicates excellent adherence to Zipf's law
+- **High Frequency Dominance:** Top 100 words cover 46.0% of corpus
+- **Long Tail:** 846,070 words needed for remaining 11.0% coverage
 
 ---
 ## 5. Word Embeddings Evaluation
@@ -326,24 +384,113 @@ 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** | 573,439 | 32 | 5.170 | 1.489 | 0.8067 🏆 |
-| **mono_64d** | 573,439 | 64 | 5.558 | 1.367 | 0.7701 |
-| **mono_128d** | 573,439 | 128 | 5.982 | 1.280 | 0.7047 |
-| **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 | 0.7897 🏆 | 0.3482 | N/A | N/A |
+| **mono_64d** | 64 | 0.7690 | 0.2976 | N/A | N/A |
+| **mono_128d** | 128 | 0.7177 | 0.2526 | N/A | N/A |
 
 ### Key Findings
 
-- **Best Isotropy:** mono_32d with 0.8067 (more uniform distribution)
-- **Dimension Trade-off:** Higher dimensions capture more semantics but reduce isotropy
-- **Vocabulary Coverage:** All models cover 573,439 words
-- **Recommendation:** 100d for balanced semantic capture and efficiency
+- **Best Isotropy:** mono_32d with 0.7897 (more uniform distribution)
+- **Semantic Density:** Average pairwise similarity of 0.2995. Lower values indicate better semantic separation.
+- **Alignment Quality:** No aligned models evaluated in this run.
+- **Recommendation:** 128d aligned for best cross-lingual performance
+
+---
+## 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 |
+|--------|----------|
+| `-ال` | الخوذ, المندوبين, الدمرداشيه |
+
+#### Productive Suffixes
+| Suffix | Examples |
+|--------|----------|
+| `-ين` | كلوكيرين, بيرجرين, المندوبين |
+| `-ان` | مالڤان, ملازمان, پايرلمان |
+
+### 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 |
+|------|----------|------------------|----------|
+| `العا` | 1.85x | 296 contexts | العام, العاج, العال |
+| `المج` | 1.79x | 267 contexts | المجد, المجر, المجئ |
+| `انزي` | 1.95x | 165 contexts | انزيا, انزيت, انزيد |
+| `الشع` | 2.11x | 103 contexts | الشعب, الشعف, الشعز |
+| `ياته` | 2.11x | 96 contexts | عياته, آياته, حياته |
+| `الاع` | 2.00x | 107 contexts | الاعور, الاعتر, الاعدا |
+| `مستق` | 2.01x | 80 contexts | مستقل, مستقر, مستقله |
+| `الاح` | 1.79x | 110 contexts | الاحد, صالاحى, الاحرش |
+| `لموج` | 2.13x | 48 contexts | لموجة, الموج, الموجب |
+| `لمجر` | 1.85x | 71 contexts | لمجره, المجر, لمجرة |
+| `لساع` | 2.34x | 28 contexts | لساعة, الساعة, لساعات |
+| `مريك` | 1.69x | 102 contexts | لمريك, مريكا, مريكن |
+
+### 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 |
+|--------|--------|-----------|----------|
+| `-ال` | `-ين` | 47 words | الصديقين, الحدوديين |
+| `-ال` | `-ان` | 11 words | الأخوان, الترامان |
+
+### 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 |
+|------|-----------------|------------|------|
+| السريانيين | **`ال-سرياني-ين`** | 6.0 | `سرياني` |
+| كانتيلينين | **`كانتيل-ين-ين`** | 6.0 | `كانتيل` |
+| الجينومية | **`ال-جينومية`** | 4.5 | `جينومية` |
+| البرمجيات | **`ال-برمجيات`** | 4.5 | `برمجيات` |
+| الاستعلامات | **`ال-استعلامات`** | 4.5 | `استعلامات` |
+| بيجلاندسفچوردين | **`بيجلاندسفچورد-ين`** | 4.5 | `بيجلاندسفچورد` |
+| السينابون | **`ال-سينابون`** | 4.5 | `سينابون` |
+| الديمقراطي | **`ال-ديمقراطي`** | 4.5 | `ديمقراطي` |
+| الانبعاثية | **`ال-انبعاثية`** | 4.5 | `انبعاثية` |
+| الميتانيه | **`ال-ميتانيه`** | 4.5 | `ميتانيه` |
+| الطويحينه | **`ال-طويحينه`** | 4.5 | `طويحينه` |
+| الصابونجى | **`ال-صابونجى`** | 4.5 | `صابونجى` |
+| البنغاليه | **`ال-بنغاليه`** | 4.5 | `بنغاليه` |
+| المتحدثون | **`ال-متحدثون`** | 4.5 | `متحدثون` |
+| ستشميدلين | **`ستشميدل-ين`** | 4.5 | `ستشميدل` |
+
+### 6.6 Linguistic Interpretation
+
+> **Automated Insight:**
+The language Egyptian Arabic 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.
 
 ---
-## 6. Summary & Recommendations
+## 7. Summary & Recommendations
 
 ![Performance Dashboard](visualizations/performance_dashboard.png)
 
@@ -351,11 +498,12 @@ Below are text samples generated from each Markov chain model:
 
 | Component | Recommended | Rationale |
 |-----------|-------------|-----------|
-| Tokenizer | **32k BPE** | Best compression (4.02x) with low UNK rate |
-| N-gram | **5-gram** | Lowest perplexity (384) |
-| Markov | **Context-4** | Highest predictability (92.8%) |
+| Tokenizer | **64k BPE** | Best compression (3.91x) |
+| N-gram | **2-gram** | Lowest perplexity (316) |
+| Markov | **Context-4** | Highest predictability (93.7%) |
 | Embeddings | **100d** | Balanced semantic capture and isotropy |
 
+
 ---
 ## Appendix: Metrics Glossary & Interpretation Guide
 
@@ -545,7 +693,8 @@ 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}
 }
@@ -561,7 +710,8 @@ 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 18:23:46*
+*Report Date: 2026-01-03 07:45:31*

models/embeddings/monolingual/arz_128d.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:75a11f48624beba71cf5e2fb4ca6caff536a87ce2cd7ecf4cf2d74b7d061ab24
-size 1623402315
+oid sha256:5bdb36e12bc1a678fd5c157ad32e02341de1eca60c4bc2b5ef93fe49b61bd555
+size 1527330535

models/embeddings/monolingual/arz_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": 573439
+  "vocab_size": 481203
 }

models/embeddings/monolingual/arz_32d.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:20ccadbe8619b844c6eace4a02ce277749c846f73614cb161ecbc7a86c873672
-size 415001163
+oid sha256:30582a755835303bdd9d0ea4ad7243b43b631aa397c3e567b21c8d4a5c4449d3
+size 389766631

models/embeddings/monolingual/arz_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": 573439
+  "vocab_size": 481203
 }

models/embeddings/monolingual/arz_64d.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4801d005678b73c1bb59d155df23fbc73c9f36b3c781e8ec0acef3f3d272ee93
-size 817801547
+oid sha256:5068a8b810efa630cdf0314c353a4495557bcd65811e48356936ecd7a645b35c
+size 768954599

models/embeddings/monolingual/arz_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": 573439
+  "vocab_size": 481203
 }

models/subword_markov/arz_markov_ctx1_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e7f161c6e8fab68e5fbc9600f3451111167fbdb2ccb9ee71b98eeb2ffb7dace1
-size 322433
+oid sha256:93b9ef162389968b455ca3710639cb0faa0e67079e19324a4f3d232dc220101e
+size 347493

models/subword_markov/arz_markov_ctx1_subword_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 1,
   "variant": "subword",
   "language": "arz",
-  "unique_contexts": 4455,
-  "total_transitions": 825886705
+  "unique_contexts": 5726,
+  "total_transitions": 693777470
 }

models/subword_markov/arz_markov_ctx2_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:cabc4c543608790d6c5880f1744ba45b47d2697724261238e8925f80995b0766
-size 2557134
+oid sha256:98db3946a7c34d13376e437f6401ae1d31aa408e141da100b0793536f9682ba1
+size 2267694

models/subword_markov/arz_markov_ctx2_subword_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 2,
   "variant": "subword",
   "language": "arz",
-  "unique_contexts": 41887,
-  "total_transitions": 824256957
+  "unique_contexts": 47379,
+  "total_transitions": 692148775
 }

models/subword_markov/arz_markov_ctx3_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:de040657fc371a339db59ea68152c56a326e76cc8924dac007de0473e86340c4
-size 13516070
+oid sha256:791a2be4e464c30d58ac081746362b42037f50dd8d0c4b552c3b0e68c2518dea
+size 10864076

models/subword_markov/arz_markov_ctx3_subword_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 3,
   "variant": "subword",
   "language": "arz",
-  "unique_contexts": 320101,
-  "total_transitions": 822627209
+  "unique_contexts": 262420,
+  "total_transitions": 690520080
 }

models/subword_markov/arz_markov_ctx4_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a0e33690e76e9f408ce70c7c674c98a698e70a625d4ecd0d337e959c13d77fe6
-size 53877656
+oid sha256:f9412f5ef232c0c613c068829b486f969ed98f144b0956cfba48ca7651d697cc
+size 40934252

models/subword_markov/arz_markov_ctx4_subword_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 4,
   "variant": "subword",
   "language": "arz",
-  "unique_contexts": 1700014,
-  "total_transitions": 820997461
+  "unique_contexts": 1241425,
+  "total_transitions": 688891385
 }

models/subword_ngram/arz_2gram_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:10ff36eb8f35bdaef34f28641466ebac1bd3df6e7831eab6cec38563f7a83d00
-size 224935
+oid sha256:54b1e94823bf6eda78e610a31f50839e3eb6945296345628a51c2775bbc535b5
+size 225336

models/subword_ngram/arz_2gram_subword_metadata.json

@@ -2,6 +2,6 @@
   "n": 2,
   "variant": "subword",
   "language": "arz",
-  "unique_ngrams": 15875,
-  "total_ngrams": 825886705
+  "unique_ngrams": 15451,
+  "total_ngrams": 693777470
 }

models/subword_ngram/arz_3gram_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:eee986ddda6f36ba1d3e23a8759ad8f5509e3c48dc4ac592f689fa826fe34187
-size 2146043
+oid sha256:6241036ff50c386080b4d36854fdc224c1c5099af471b3862e1464cd644b63ae
+size 1697106

models/subword_ngram/arz_3gram_subword_metadata.json

@@ -2,6 +2,6 @@
   "n": 3,
   "variant": "subword",
   "language": "arz",
-  "unique_ngrams": 170638,
-  "total_ngrams": 824256957
+  "unique_ngrams": 129923,
+  "total_ngrams": 692148775
 }

models/subword_ngram/arz_4gram_subword.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2d7383505bdffee6fff1e255e8a66837be06fe62780607b993d354aed68b8d0c
-size 13255914
+oid sha256:c000b6a963ccc11a1bb0feebe6d49196497cbb6667e247ab3df4764a4e91003c
+size 10220298

models/subword_ngram/arz_4gram_subword_metadata.json

@@ -2,6 +2,6 @@
   "n": 4,
   "variant": "subword",
   "language": "arz",
-  "unique_ngrams": 1045558,
-  "total_ngrams": 822627209
+  "unique_ngrams": 788718,
+  "total_ngrams": 690520080
 }

models/tokenizer/arz_tokenizer_16k.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3250e5b1fa849747bd31693274ccdde1fde493080110892d89c056f2ed133c9f
-size 553249
+oid sha256:8182c7627a2b642bc8213e1f478fb8483ae43e1decdfb9d36a8b81c0b1e5db70
+size 553522

models/tokenizer/arz_tokenizer_32k.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1d22c8b625df815da9ab4ad818bdb3e662168bfa181b0747326154e8db1cc2c3
-size 880194
+oid sha256:3f58411587d93e3605bc0bfb493f2adb152e457978754608d4bfa1098bbe3484
+size 874271

models/tokenizer/arz_tokenizer_64k.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e2a8f70071cee82c435cc2eff7086ba3060e694f61e640ad9482184f637090af
-size 1550950
+oid sha256:b7caa9ae0cc952f588395f385251d423792d50e832f759c19f62d2debfa53c97
+size 1535709

models/tokenizer/arz_tokenizer_8k.model

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e3cf01b1aa16a4d259de9989643f646ceecca5967b87fc2b269ffd411f6ec306
-size 394841
+oid sha256:75cf05a2c2e3d6160ef1a57d3e6c1105fcab185e64a3ffd3fdab63dacc685a1f
+size 396360

models/vocabulary/arz_vocabulary.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:08ac9dc5a3750de53ceee73bd67ec52e8f566950d994cd5585cbd78867c9790d
-size 13750687
+oid sha256:f49f18512e3b41813735d519aef4205264437e74599a3808b6c8fbea97f3bb17
+size 12321602

models/vocabulary/arz_vocabulary_metadata.json

@@ -1,16 +1,17 @@
 {
   "language": "arz",
-  "vocabulary_size": 1000000,
+  "vocabulary_size": 856070,
+  "variant": "full",
   "statistics": {
-    "type_token_ratio": 0.016231404296933028,
+    "type_token_ratio": 0.011546453807845636,
     "coverage": {
-      "top_100": 0.44057406359024887,
-      "top_1000": 0.7499824220069986,
-      "top_5000": 0.8482657952087583,
-      "top_10000": 0.8783945758611642
+      "top_100": 0.4579160902506231,
+      "top_1000": 0.7632735433913325,
+      "top_5000": 0.8553371329341142,
+      "top_10000": 0.885855315521865
     },
-    "hapax_count": 918045,
-    "hapax_ratio": 0.42167650913059435,
-    "total_documents": 1629748
+    "hapax_count": 497272,
+    "hapax_ratio": 0.36744001146790684,
+    "total_documents": 1628695
   }
 }

models/word_markov/arz_markov_ctx1_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e5b7e0c4e188921694b306869a2449ce112fb199b792128f1ce21236b921cc84
-size 138534439
+oid sha256:27fac99b38072bd41d19f9620ff6d00511cdf0cb2c23718cc24825c3425b0c81
+size 125535197

models/word_markov/arz_markov_ctx1_word_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 1,
   "variant": "word",
   "language": "arz",
-  "unique_contexts": 2177526,
-  "total_transitions": 156097037
+  "unique_contexts": 1353062,
+  "total_transitions": 115579759
 }

models/word_markov/arz_markov_ctx2_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:40f8b7cfcf0a1b3e4aeec633859a0e58e1f83c410282fc81378b386a72be3a23
-size 440008287
+oid sha256:9256f52202a6653f8df7a39b233dce04b0731a6961f54294e451f0ff631a642f
+size 413546149

models/word_markov/arz_markov_ctx2_word_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 2,
   "variant": "word",
   "language": "arz",
-  "unique_contexts": 13972922,
-  "total_transitions": 154467289
+  "unique_contexts": 12336484,
+  "total_transitions": 113951064
 }

models/word_markov/arz_markov_ctx3_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:10869a5b62f9dc9fced8f99ecf292c2e516552cafb1ebc22439f4c66b29703c6
-size 749801655
+oid sha256:a2fa6b997fa7dcafddfbda3b21ca4c509b51a1df648b8e5d0d2476f263f776b1
+size 672892499

models/word_markov/arz_markov_ctx3_word_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 3,
   "variant": "word",
   "language": "arz",
-  "unique_contexts": 28492247,
-  "total_transitions": 152837542
+  "unique_contexts": 23517673,
+  "total_transitions": 112322369
 }

models/word_markov/arz_markov_ctx4_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1c3be90b2e9c7b80d31d2c18860bc2236215712a3340fa2f9225e2de14a91424
-size 1031396379
+oid sha256:d3b791771441b0e8ec65696094fed0fc807d2e95898fb465fd96d5dd181add4d
+size 913306903

models/word_markov/arz_markov_ctx4_word_metadata.json

@@ -2,6 +2,6 @@
   "context_size": 4,
   "variant": "word",
   "language": "arz",
-  "unique_contexts": 38092155,
-  "total_transitions": 151207798
+  "unique_contexts": 29894419,
+  "total_transitions": 110693674
 }

models/word_ngram/arz_2gram_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:739d5abccec316bf509349d0e17742a1bcf59999ee5f8e810053baf5105c06e5
-size 25088974
+oid sha256:3d603ccb8a9bbb46f85861773239d3031d8c0e830c109fc753e1cadd69e9c1e2
+size 22328144

models/word_ngram/arz_2gram_word_metadata.json

@@ -2,6 +2,6 @@
   "n": 2,
   "variant": "word",
   "language": "arz",
-  "unique_ngrams": 1353863,
-  "total_ngrams": 156097037
+  "unique_ngrams": 1073861,
+  "total_ngrams": 115579759
 }

models/word_ngram/arz_3gram_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f744cb53e652ce297a655902bc6eb5ef9d19d5edfc79585d48fa10ea229f7ccb
-size 49189391
+oid sha256:879ba3166c5b357094ce092093cc1eae670fb25bcde483ad58a4dd24786c83e5
+size 41669189

models/word_ngram/arz_3gram_word_metadata.json

@@ -2,6 +2,6 @@
   "n": 3,
   "variant": "word",
   "language": "arz",
-  "unique_ngrams": 2306233,
-  "total_ngrams": 154467289
+  "unique_ngrams": 1682809,
+  "total_ngrams": 113951064
 }

models/word_ngram/arz_4gram_word.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d78b4518cd7a719707b9135b4062484b3fba60daa9102960b10361304503a812
-size 109747485
+oid sha256:bfe08d57419c8d5a905c1468d1a4d8b376b38c13bc69a838fa0cb4e9e94d0d24
+size 84435807

models/word_ngram/arz_4gram_word_metadata.json

@@ -2,6 +2,6 @@
   "n": 4,
   "variant": "word",
   "language": "arz",
-  "unique_ngrams": 4880347,
-  "total_ngrams": 152837542
+  "unique_ngrams": 3054922,
+  "total_ngrams": 112322369
 }