Upload all models and assets for bpy (latest)
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- .gitattributes +1 -0
- README.md +179 -147
- models/embeddings/aligned/bpy_128d.bin +3 -0
- models/embeddings/aligned/bpy_128d.meta.json +1 -0
- models/embeddings/aligned/bpy_128d.projection.npy +3 -0
- models/embeddings/aligned/bpy_128d_metadata.json +8 -0
- models/embeddings/aligned/bpy_32d.bin +3 -0
- models/embeddings/aligned/bpy_32d.meta.json +1 -0
- models/embeddings/aligned/bpy_32d.projection.npy +3 -0
- models/embeddings/aligned/bpy_32d_metadata.json +8 -0
- models/embeddings/aligned/bpy_64d.bin +3 -0
- models/embeddings/aligned/bpy_64d.meta.json +1 -0
- models/embeddings/aligned/bpy_64d.projection.npy +3 -0
- models/embeddings/aligned/bpy_64d_metadata.json +8 -0
- models/embeddings/monolingual/bpy_128d.bin +2 -2
- models/embeddings/monolingual/bpy_128d_metadata.json +1 -1
- models/embeddings/monolingual/bpy_32d.bin +2 -2
- models/embeddings/monolingual/bpy_32d_metadata.json +1 -1
- models/embeddings/monolingual/bpy_64d.bin +2 -2
- models/embeddings/monolingual/bpy_64d_metadata.json +1 -1
- models/subword_markov/bpy_markov_ctx1_subword.parquet +2 -2
- models/subword_markov/bpy_markov_ctx1_subword_metadata.json +2 -2
- models/subword_markov/bpy_markov_ctx2_subword.parquet +2 -2
- models/subword_markov/bpy_markov_ctx2_subword_metadata.json +2 -2
- models/subword_markov/bpy_markov_ctx3_subword.parquet +2 -2
- models/subword_markov/bpy_markov_ctx3_subword_metadata.json +2 -2
- models/subword_markov/bpy_markov_ctx4_subword.parquet +2 -2
- models/subword_markov/bpy_markov_ctx4_subword_metadata.json +2 -2
- models/subword_ngram/bpy_2gram_subword.parquet +2 -2
- models/subword_ngram/bpy_2gram_subword_metadata.json +2 -2
- models/subword_ngram/bpy_3gram_subword.parquet +2 -2
- models/subword_ngram/bpy_3gram_subword_metadata.json +2 -2
- models/subword_ngram/bpy_4gram_subword.parquet +2 -2
- models/subword_ngram/bpy_4gram_subword_metadata.json +2 -2
- models/subword_ngram/bpy_5gram_subword.parquet +3 -0
- models/subword_ngram/bpy_5gram_subword_metadata.json +7 -0
- models/tokenizer/bpy_tokenizer_16k.model +2 -2
- models/tokenizer/bpy_tokenizer_16k.vocab +0 -0
- models/tokenizer/bpy_tokenizer_32k.model +2 -2
- models/tokenizer/bpy_tokenizer_32k.vocab +0 -0
- models/tokenizer/bpy_tokenizer_64k.model +2 -2
- models/tokenizer/bpy_tokenizer_64k.vocab +0 -0
- models/tokenizer/bpy_tokenizer_8k.model +2 -2
- models/tokenizer/bpy_tokenizer_8k.vocab +0 -0
- models/vocabulary/bpy_vocabulary.parquet +2 -2
- models/vocabulary/bpy_vocabulary_metadata.json +9 -9
- models/word_markov/bpy_markov_ctx1_word.parquet +2 -2
- models/word_markov/bpy_markov_ctx1_word_metadata.json +2 -2
- models/word_markov/bpy_markov_ctx2_word.parquet +2 -2
- models/word_markov/bpy_markov_ctx2_word_metadata.json +2 -2
.gitattributes
CHANGED
|
@@ -39,3 +39,4 @@ visualizations/position_encoding_comparison.png filter=lfs diff=lfs merge=lfs -t
|
|
| 39 |
visualizations/tsne_sentences.png filter=lfs diff=lfs merge=lfs -text
|
| 40 |
visualizations/tsne_words.png filter=lfs diff=lfs merge=lfs -text
|
| 41 |
visualizations/zipf_law.png filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 39 |
visualizations/tsne_sentences.png filter=lfs diff=lfs merge=lfs -text
|
| 40 |
visualizations/tsne_words.png filter=lfs diff=lfs merge=lfs -text
|
| 41 |
visualizations/zipf_law.png filter=lfs diff=lfs merge=lfs -text
|
| 42 |
+
visualizations/embedding_tsne_multilingual.png filter=lfs diff=lfs merge=lfs -text
|
README.md
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
---
|
| 2 |
language: bpy
|
| 3 |
-
language_name:
|
| 4 |
language_family: indoaryan_eastern
|
| 5 |
tags:
|
| 6 |
- wikilangs
|
|
@@ -10,11 +10,21 @@ tags:
|
|
| 10 |
- n-gram
|
| 11 |
- markov
|
| 12 |
- wikipedia
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
- monolingual
|
| 14 |
- family-indoaryan_eastern
|
| 15 |
license: mit
|
| 16 |
library_name: wikilangs
|
| 17 |
-
pipeline_tag:
|
| 18 |
datasets:
|
| 19 |
- omarkamali/wikipedia-monthly
|
| 20 |
dataset_info:
|
|
@@ -23,20 +33,20 @@ dataset_info:
|
|
| 23 |
metrics:
|
| 24 |
- name: best_compression_ratio
|
| 25 |
type: compression
|
| 26 |
-
value: 4.
|
| 27 |
- name: best_isotropy
|
| 28 |
type: isotropy
|
| 29 |
-
value: 0.
|
| 30 |
- name: vocabulary_size
|
| 31 |
type: vocab
|
| 32 |
value: 0
|
| 33 |
generated: 2026-01-03
|
| 34 |
---
|
| 35 |
|
| 36 |
-
#
|
| 37 |
## Comprehensive Research Report & Full Ablation Study
|
| 38 |
|
| 39 |
-
This repository contains NLP models trained and evaluated by Wikilangs, specifically on **
|
| 40 |
We analyze tokenizers, n-gram models, Markov chains, vocabulary statistics, and word embeddings.
|
| 41 |
|
| 42 |
## 📋 Repository Contents
|
|
@@ -60,7 +70,7 @@ We analyze tokenizers, n-gram models, Markov chains, vocabulary statistics, and
|
|
| 60 |
- [3. Markov Chain Evaluation](#3-markov-chain-evaluation)
|
| 61 |
- [4. Vocabulary Analysis](#4-vocabulary-analysis)
|
| 62 |
- [5. Word Embeddings Evaluation](#5-word-embeddings-evaluation)
|
| 63 |
-
- [6. Morphological Analysis (Experimental)](#6
|
| 64 |
- [7. Summary & Recommendations](#7-summary--recommendations)
|
| 65 |
- [Metrics Glossary](#appendix-metrics-glossary--interpretation-guide)
|
| 66 |
- [Visualizations Index](#visualizations-index)
|
|
@@ -80,47 +90,47 @@ We analyze tokenizers, n-gram models, Markov chains, vocabulary statistics, and
|
|
| 80 |
|
| 81 |
| Vocab Size | Compression | Avg Token Len | UNK Rate | Total Tokens |
|
| 82 |
|------------|-------------|---------------|----------|--------------|
|
| 83 |
-
| **8k** | 4.
|
| 84 |
-
| **16k** | 4.662x | 4.67 | 0.2469% | 96,
|
| 85 |
-
| **32k** | 4.
|
| 86 |
-
| **64k** | 4.
|
| 87 |
|
| 88 |
### Tokenization Examples
|
| 89 |
|
| 90 |
Below are sample sentences tokenized with each vocabulary size:
|
| 91 |
|
| 92 |
-
**Sample 1:**
|
| 93 |
|
| 94 |
| Vocab | Tokens | Count |
|
| 95 |
|-------|--------|-------|
|
| 96 |
-
| 8k |
|
| 97 |
-
| 16k |
|
| 98 |
-
| 32k |
|
| 99 |
-
| 64k |
|
| 100 |
|
| 101 |
-
**Sample 2:**
|
| 102 |
|
| 103 |
| Vocab | Tokens | Count |
|
| 104 |
|-------|--------|-------|
|
| 105 |
-
| 8k |
|
| 106 |
-
| 16k |
|
| 107 |
-
| 32k |
|
| 108 |
-
| 64k |
|
| 109 |
|
| 110 |
-
**Sample 3:**
|
| 111 |
|
| 112 |
| Vocab | Tokens | Count |
|
| 113 |
|-------|--------|-------|
|
| 114 |
-
| 8k |
|
| 115 |
-
| 16k |
|
| 116 |
-
| 32k |
|
| 117 |
-
| 64k |
|
| 118 |
|
| 119 |
|
| 120 |
### Key Findings
|
| 121 |
|
| 122 |
-
- **Best Compression:** 64k achieves 4.
|
| 123 |
-
- **Lowest UNK Rate:** 8k with 0.
|
| 124 |
- **Trade-off:** Larger vocabularies improve compression but increase model size
|
| 125 |
- **Recommendation:** 32k vocabulary provides optimal balance for production use
|
| 126 |
|
|
@@ -137,12 +147,14 @@ Below are sample sentences tokenized with each vocabulary size:
|
|
| 137 |
|
| 138 |
| N-gram | Variant | Perplexity | Entropy | Unique N-grams | Top-100 Coverage | Top-1000 Coverage |
|
| 139 |
|--------|---------|------------|---------|----------------|------------------|-------------------|
|
| 140 |
-
| **2-gram** | Word |
|
| 141 |
-
| **2-gram** | Subword | 598 🏆 | 9.
|
| 142 |
-
| **3-gram** | Word | 1,
|
| 143 |
-
| **3-gram** | Subword | 1,
|
| 144 |
-
| **4-gram** | Word | 2,
|
| 145 |
-
| **4-gram** | Subword | 3,
|
|
|
|
|
|
|
| 146 |
|
| 147 |
### Top 5 N-grams by Size
|
| 148 |
|
|
@@ -151,8 +163,8 @@ Below are sample sentences tokenized with each vocabulary size:
|
|
| 151 |
| Rank | N-gram | Count |
|
| 152 |
|------|--------|-------|
|
| 153 |
| 1 | `সাক্ষরতার হারহান` | 26,823 |
|
| 154 |
-
| 2 | `অতার মা` | 20,
|
| 155 |
-
| 3 | `জনসংখ্যার উপাত্ত` | 19,
|
| 156 |
| 4 | `জনসংখ্যা ইলাতাই` | 19,552 |
|
| 157 |
| 5 | `লোক গননা` | 19,533 |
|
| 158 |
|
|
@@ -176,42 +188,62 @@ Below are sample sentences tokenized with each vocabulary size:
|
|
| 176 |
| 4 | `অক্ষাংশ বারো দ্রাঘিমাংশ ইলতাই` | 9,366 |
|
| 177 |
| 5 | `মাপাহানর অক্ষাংশ বারো দ্রাঘিমাংশ` | 9,315 |
|
| 178 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 179 |
**2-grams (Subword):**
|
| 180 |
|
| 181 |
| Rank | N-gram | Count |
|
| 182 |
|------|--------|-------|
|
| 183 |
-
| 1 | `র _` | 407,
|
| 184 |
-
| 2 | `। _` | 163,
|
| 185 |
-
| 3 | `হা ন` | 154,
|
| 186 |
-
| 4 | `ন _` | 147,
|
| 187 |
-
| 5 | `_ মা` | 138,
|
| 188 |
|
| 189 |
**3-grams (Subword):**
|
| 190 |
|
| 191 |
| Rank | N-gram | Count |
|
| 192 |
|------|--------|-------|
|
| 193 |
-
| 1 | `র _ মা` | 95,
|
| 194 |
-
| 2 | `হা ন _` | 94,
|
| 195 |
-
| 3 | `_ বা রো` | 68,
|
| 196 |
-
| 4 | `বা রো _` | 68,
|
| 197 |
-
| 5 | `_ ই উ` | 64,
|
| 198 |
|
| 199 |
**4-grams (Subword):**
|
| 200 |
|
| 201 |
| Rank | N-gram | Count |
|
| 202 |
|------|--------|-------|
|
| 203 |
-
| 1 | `_ বা রো _` | 68,
|
| 204 |
-
| 2 | `_ ই উ নি` | 64,
|
| 205 |
-
| 3 | `ই উ নি য়` | 55,
|
| 206 |
-
| 4 | `উ নি য় ন` | 55,
|
| 207 |
-
| 5 | `জ ন সং খ্যা` | 44,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 208 |
|
| 209 |
|
| 210 |
### Key Findings
|
| 211 |
|
| 212 |
- **Best Perplexity:** 2-gram (subword) with 598
|
| 213 |
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
|
| 214 |
-
- **Coverage:** Top-1000 patterns cover ~
|
| 215 |
- **Recommendation:** 4-gram or 5-gram for best predictive performance
|
| 216 |
|
| 217 |
---
|
|
@@ -227,14 +259,14 @@ Below are sample sentences tokenized with each vocabulary size:
|
|
| 227 |
|
| 228 |
| Context | Variant | Avg Entropy | Perplexity | Branching Factor | Unique Contexts | Predictability |
|
| 229 |
|---------|---------|-------------|------------|------------------|-----------------|----------------|
|
| 230 |
-
| **1** | Word | 0.
|
| 231 |
-
| **1** | Subword | 1.
|
| 232 |
-
| **2** | Word | 0.1820 | 1.134 | 1.54 | 262,
|
| 233 |
-
| **2** | Subword | 0.
|
| 234 |
-
| **3** | Word | 0.0756 | 1.054 | 1.27 |
|
| 235 |
-
| **3** | Subword | 0.
|
| 236 |
-
| **4** | Word | 0.
|
| 237 |
-
| **4** | Subword | 0.
|
| 238 |
|
| 239 |
### Generated Text Samples (Word-based)
|
| 240 |
|
|
@@ -242,27 +274,27 @@ Below are text samples generated from each word-based Markov chain model:
|
|
| 242 |
|
| 243 |
**Context Size 1:**
|
| 244 |
|
| 245 |
-
1. `বারো
|
| 246 |
-
2. `ইউনিয়ন এগত
|
| 247 |
-
3. `উপাত্ত শহর এহার
|
| 248 |
|
| 249 |
**Context Size 2:**
|
| 250 |
|
| 251 |
-
1. `সাক্ষরতার হারহান
|
| 252 |
-
2. `অতার মা
|
| 253 |
-
3. `জনসংখ্যার উপাত্ত
|
| 254 |
|
| 255 |
**Context Size 3:**
|
| 256 |
|
| 257 |
-
1. `মানুলেহা লোক গননা অনুসারে
|
| 258 |
-
2. `মারির মানুলেহা লোক গননা অনুসারে
|
| 259 |
-
3. `অতার মা মুনি
|
| 260 |
|
| 261 |
**Context Size 4:**
|
| 262 |
|
| 263 |
-
1. `মারির মানুলেহা লোক গননা অনুসারে
|
| 264 |
-
2. `গ অতার মা মুনি ৫২ বারো জিলা বেয়াপা
|
| 265 |
-
3. `মানুলেহা লোক গননা অনুসারে
|
| 266 |
|
| 267 |
|
| 268 |
### Generated Text Samples (Subword-based)
|
|
@@ -271,34 +303,34 @@ Below are text samples generated from each subword-based Markov chain model:
|
|
| 271 |
|
| 272 |
**Context Size 1:**
|
| 273 |
|
| 274 |
-
1. `_
|
| 275 |
-
2.
|
| 276 |
-
3. `নর_
|
| 277 |
|
| 278 |
**Context Size 2:**
|
| 279 |
|
| 280 |
-
1. `র_
|
| 281 |
-
2. `।_
|
| 282 |
-
3.
|
| 283 |
|
| 284 |
**Context Size 3:**
|
| 285 |
|
| 286 |
-
1. `র_
|
| 287 |
-
2. `হান_
|
| 288 |
-
3. `_বারো_
|
| 289 |
|
| 290 |
**Context Size 4:**
|
| 291 |
|
| 292 |
-
1. `_বারো_
|
| 293 |
2. `_ইউনিট_আসে।_চৌদ্দাহান_মুঙেদে:`
|
| 294 |
-
3.
|
| 295 |
|
| 296 |
|
| 297 |
### Key Findings
|
| 298 |
|
| 299 |
- **Best Predictability:** Context-4 (word) with 95.1% predictability
|
| 300 |
- **Branching Factor:** Decreases with context size (more deterministic)
|
| 301 |
-
- **Memory Trade-off:** Larger contexts require more storage (
|
| 302 |
- **Recommendation:** Context-3 or Context-4 for text generation
|
| 303 |
|
| 304 |
---
|
|
@@ -314,26 +346,26 @@ Below are text samples generated from each subword-based Markov chain model:
|
|
| 314 |
|
| 315 |
| Metric | Value |
|
| 316 |
|--------|-------|
|
| 317 |
-
| Vocabulary Size |
|
| 318 |
-
| Total Tokens | 2,
|
| 319 |
-
| Mean Frequency | 61.
|
| 320 |
| Median Frequency | 3 |
|
| 321 |
-
| Frequency Std Dev |
|
| 322 |
|
| 323 |
### Most Common Words
|
| 324 |
|
| 325 |
| Rank | Word | Frequency |
|
| 326 |
|------|------|-----------|
|
| 327 |
-
| 1 | বারো | 68,
|
| 328 |
-
| 2 | ইউনিয়ন | 42,
|
| 329 |
-
| 3 | উপাত্ত | 36,
|
| 330 |
| 4 | হারহান | 31,910 |
|
| 331 |
-
| 5 | মা | 31,
|
| 332 |
-
| 6 | মানু | 30,
|
| 333 |
| 7 | সাক্ষরতার | 26,839 |
|
| 334 |
-
| 8 | গ | 26,
|
| 335 |
-
| 9 | অতার | 25,
|
| 336 |
-
| 10 | জনসংখ্যার | 24,
|
| 337 |
|
| 338 |
### Least Common Words (from vocabulary)
|
| 339 |
|
|
@@ -354,8 +386,8 @@ Below are text samples generated from each subword-based Markov chain model:
|
|
| 354 |
|
| 355 |
| Metric | Value |
|
| 356 |
|--------|-------|
|
| 357 |
-
| Zipf Coefficient | 1.
|
| 358 |
-
| R² (Goodness of Fit) | 0.
|
| 359 |
| Adherence Quality | **excellent** |
|
| 360 |
|
| 361 |
### Coverage Analysis
|
|
@@ -371,7 +403,7 @@ Below are text samples generated from each subword-based Markov chain model:
|
|
| 371 |
|
| 372 |
- **Zipf Compliance:** R²=0.9803 indicates excellent adherence to Zipf's law
|
| 373 |
- **High Frequency Dominance:** Top 100 words cover 62.6% of corpus
|
| 374 |
-
- **Long Tail:**
|
| 375 |
|
| 376 |
---
|
| 377 |
## 5. Word Embeddings Evaluation
|
|
@@ -387,37 +419,40 @@ Below are text samples generated from each subword-based Markov chain model:
|
|
| 387 |
|
| 388 |
### 5.1 Cross-Lingual Alignment
|
| 389 |
|
| 390 |
-
|
|
|
|
|
|
|
| 391 |
|
| 392 |
|
| 393 |
### 5.2 Model Comparison
|
| 394 |
|
| 395 |
| Model | Dimension | Isotropy | Semantic Density | Alignment R@1 | Alignment R@10 |
|
| 396 |
|-------|-----------|----------|------------------|---------------|----------------|
|
| 397 |
-
| **mono_32d** | 32 | 0.
|
| 398 |
-
| **mono_64d** | 64 | 0.
|
| 399 |
-
| **mono_128d** | 128 | 0.
|
|
|
|
|
|
|
|
|
|
| 400 |
|
| 401 |
### Key Findings
|
| 402 |
|
| 403 |
-
- **Best Isotropy:** mono_32d with 0.
|
| 404 |
-
- **Semantic Density:** Average pairwise similarity of 0.
|
| 405 |
-
- **Alignment Quality:**
|
| 406 |
- **Recommendation:** 128d aligned for best cross-lingual performance
|
| 407 |
|
| 408 |
---
|
| 409 |
## 6. Morphological Analysis (Experimental)
|
| 410 |
|
| 411 |
-
> ⚠️ **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.
|
| 412 |
-
|
| 413 |
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.
|
| 414 |
|
| 415 |
### 6.1 Productivity & Complexity
|
| 416 |
|
| 417 |
| Metric | Value | Interpretation | Recommendation |
|
| 418 |
|--------|-------|----------------|----------------|
|
| 419 |
-
| Productivity Index | **
|
| 420 |
-
| Idiomaticity Gap | **-
|
| 421 |
|
| 422 |
### 6.2 Affix Inventory (Productive Units)
|
| 423 |
|
|
@@ -426,20 +461,19 @@ These are the most productive prefixes and suffixes identified by sampling the v
|
|
| 426 |
#### Productive Prefixes
|
| 427 |
| Prefix | Examples |
|
| 428 |
|--------|----------|
|
| 429 |
-
| `-কা` |
|
| 430 |
-
| `-মা` | মাকৌপিন, মাঝরদিয়া, মার্চ |
|
| 431 |
|
| 432 |
#### Productive Suffixes
|
| 433 |
| Suffix | Examples |
|
| 434 |
|--------|----------|
|
| 435 |
-
| `-া` |
|
| 436 |
-
| `-র` |
|
| 437 |
-
| `-়া` |
|
| 438 |
-
|
|
| 439 |
-
|
|
| 440 |
-
|
|
| 441 |
-
|
|
| 442 |
-
|
|
| 443 |
|
| 444 |
### 6.3 Bound Stems (Lexical Roots)
|
| 445 |
|
|
@@ -454,16 +488,14 @@ This table shows which prefixes and suffixes most frequently co-occur on the sam
|
|
| 454 |
|
| 455 |
| Prefix | Suffix | Frequency | Examples |
|
| 456 |
|--------|--------|-----------|----------|
|
| 457 |
-
| `-কা` | `-া` |
|
| 458 |
-
|
|
| 459 |
-
| `-কা` |
|
| 460 |
-
|
|
| 461 |
-
| `-কা` |
|
| 462 |
-
| `-কা` |
|
| 463 |
-
|
|
| 464 |
-
| `-কা` |
|
| 465 |
-
| `-কা` | `-িয়া` | 10 words | কামারিয়া, কালেডোনিয়া |
|
| 466 |
-
| `-মা` | `-য়া` | 7 words | মাইসাটুয়া, মাছুয়া |
|
| 467 |
|
| 468 |
### 6.5 Recursive Morpheme Segmentation
|
| 469 |
|
|
@@ -471,26 +503,26 @@ Using **Recursive Hierarchical Substitutability**, we decompose complex words in
|
|
| 471 |
|
| 472 |
| Word | Suggested Split | Confidence | Stem |
|
| 473 |
|------|-----------------|------------|------|
|
| 474 |
-
|
|
| 475 |
-
|
|
| 476 |
-
|
|
| 477 |
-
|
|
| 478 |
-
|
|
| 479 |
-
|
|
| 480 |
-
|
|
| 481 |
-
|
|
| 482 |
-
|
|
| 483 |
-
|
|
| 484 |
-
|
|
| 485 |
-
|
|
| 486 |
-
|
|
| 487 |
| কাউন্দিয়া | **`কা-উন্দ-িয়া`** | 3.0 | `উন্দ` |
|
| 488 |
-
|
|
| 489 |
|
| 490 |
### 6.6 Linguistic Interpretation
|
| 491 |
|
| 492 |
> **Automated Insight:**
|
| 493 |
-
The language
|
| 494 |
|
| 495 |
---
|
| 496 |
## 7. Summary & Recommendations
|
|
@@ -501,7 +533,7 @@ The language BPY appears to be more isolating or has a highly fixed vocabulary.
|
|
| 501 |
|
| 502 |
| Component | Recommended | Rationale |
|
| 503 |
|-----------|-------------|-----------|
|
| 504 |
-
| Tokenizer | **64k BPE** | Best compression (4.
|
| 505 |
| N-gram | **2-gram** | Lowest perplexity (598) |
|
| 506 |
| Markov | **Context-4** | Highest predictability (95.1%) |
|
| 507 |
| Embeddings | **100d** | Balanced semantic capture and isotropy |
|
|
@@ -717,4 +749,4 @@ MIT License - Free for academic and commercial use.
|
|
| 717 |
---
|
| 718 |
*Generated by Wikilangs Models Pipeline*
|
| 719 |
|
| 720 |
-
*Report Date: 2026-01-03
|
|
|
|
| 1 |
---
|
| 2 |
language: bpy
|
| 3 |
+
language_name: Bishnupriya
|
| 4 |
language_family: indoaryan_eastern
|
| 5 |
tags:
|
| 6 |
- wikilangs
|
|
|
|
| 10 |
- n-gram
|
| 11 |
- markov
|
| 12 |
- wikipedia
|
| 13 |
+
- feature-extraction
|
| 14 |
+
- sentence-similarity
|
| 15 |
+
- tokenization
|
| 16 |
+
- n-grams
|
| 17 |
+
- markov-chain
|
| 18 |
+
- text-mining
|
| 19 |
+
- fasttext
|
| 20 |
+
- babelvec
|
| 21 |
+
- vocabulous
|
| 22 |
+
- vocabulary
|
| 23 |
- monolingual
|
| 24 |
- family-indoaryan_eastern
|
| 25 |
license: mit
|
| 26 |
library_name: wikilangs
|
| 27 |
+
pipeline_tag: text-generation
|
| 28 |
datasets:
|
| 29 |
- omarkamali/wikipedia-monthly
|
| 30 |
dataset_info:
|
|
|
|
| 33 |
metrics:
|
| 34 |
- name: best_compression_ratio
|
| 35 |
type: compression
|
| 36 |
+
value: 4.935
|
| 37 |
- name: best_isotropy
|
| 38 |
type: isotropy
|
| 39 |
+
value: 0.6926
|
| 40 |
- name: vocabulary_size
|
| 41 |
type: vocab
|
| 42 |
value: 0
|
| 43 |
generated: 2026-01-03
|
| 44 |
---
|
| 45 |
|
| 46 |
+
# Bishnupriya - Wikilangs Models
|
| 47 |
## Comprehensive Research Report & Full Ablation Study
|
| 48 |
|
| 49 |
+
This repository contains NLP models trained and evaluated by Wikilangs, specifically on **Bishnupriya** Wikipedia data.
|
| 50 |
We analyze tokenizers, n-gram models, Markov chains, vocabulary statistics, and word embeddings.
|
| 51 |
|
| 52 |
## 📋 Repository Contents
|
|
|
|
| 70 |
- [3. Markov Chain Evaluation](#3-markov-chain-evaluation)
|
| 71 |
- [4. Vocabulary Analysis](#4-vocabulary-analysis)
|
| 72 |
- [5. Word Embeddings Evaluation](#5-word-embeddings-evaluation)
|
| 73 |
+
- [6. Morphological Analysis (Experimental)](#6--morphological-analysis-experimental)
|
| 74 |
- [7. Summary & Recommendations](#7-summary--recommendations)
|
| 75 |
- [Metrics Glossary](#appendix-metrics-glossary--interpretation-guide)
|
| 76 |
- [Visualizations Index](#visualizations-index)
|
|
|
|
| 90 |
|
| 91 |
| Vocab Size | Compression | Avg Token Len | UNK Rate | Total Tokens |
|
| 92 |
|------------|-------------|---------------|----------|--------------|
|
| 93 |
+
| **8k** | 4.501x | 4.51 | 0.2384% | 99,847 |
|
| 94 |
+
| **16k** | 4.662x | 4.67 | 0.2469% | 96,404 |
|
| 95 |
+
| **32k** | 4.818x | 4.83 | 0.2551% | 93,284 |
|
| 96 |
+
| **64k** | 4.935x 🏆 | 4.95 | 0.2614% | 91,058 |
|
| 97 |
|
| 98 |
### Tokenization Examples
|
| 99 |
|
| 100 |
Below are sample sentences tokenized with each vocabulary size:
|
| 101 |
|
| 102 |
+
**Sample 1:** `ইথাক বিষ্ণুপ্রিয়া মণিপুরী ঠারর অনিয়মিত পত্রিকা আহান, যেহান সংগ্রাম সিংহর সম্পা...`
|
| 103 |
|
| 104 |
| Vocab | Tokens | Count |
|
| 105 |
|-------|--------|-------|
|
| 106 |
+
| 8k | `▁ই থ াক ▁বিষ্ণুপ্রিয়া ▁মণিপুরী ▁ঠারর ▁অ নি য় মি ... (+21 more)` | 31 |
|
| 107 |
+
| 16k | `▁ই থ াক ▁বিষ্ণুপ্রিয়া ▁মণিপুরী ▁ঠারর ▁অ নি য় মিত ... (+18 more)` | 28 |
|
| 108 |
+
| 32k | `▁ই থ াক ▁বিষ্ণুপ্রিয়া ▁মণিপুরী ▁ঠারর ▁অনি য়মিত ▁পত্রিকা ▁আহান ... (+13 more)` | 23 |
|
| 109 |
+
| 64k | `▁ইথাক ▁বিষ্ণুপ্রিয়া ▁মণিপুরী ▁ঠারর ▁অনিয়মিত ▁পত্রিকা ▁আহান , ▁যেহান ▁সংগ্রাম ... (+8 more)` | 18 |
|
| 110 |
|
| 111 |
+
**Sample 2:** `.এমও(.mo) এগ মাকাউর নাঙে লেপকরিসি চিঙপা ডমেইনগ (ccTLD)। মিলাপ আইএএনএ-র মাকাউর তথ...`
|
| 112 |
|
| 113 |
| Vocab | Tokens | Count |
|
| 114 |
|-------|--------|-------|
|
| 115 |
+
| 8k | `▁. এম ও (. mo ) ▁এগ ▁মাকা উর ▁নাঙে ... (+23 more)` | 33 |
|
| 116 |
+
| 16k | `▁. এম ও (. mo ) ▁এগ ▁মাকা উর ▁নাঙে ... (+23 more)` | 33 |
|
| 117 |
+
| 32k | `▁. এম ও (. mo ) ▁এগ ▁মাকাউর ▁নাঙে ▁লেপকরিসি ... (+21 more)` | 31 |
|
| 118 |
+
| 64k | `▁. এম ও (. mo ) ▁এগ ▁মাকাউর ▁নাঙে ▁লেপকরিসি ... (+21 more)` | 31 |
|
| 119 |
|
| 120 |
+
**Sample 3:** `বাংলাদেশর স্থানীয় সরকারর সিজিলে আসেতাই জিলা পরিষদ সিটি কর্পোরেশন (৬গ) থানা বারো...`
|
| 121 |
|
| 122 |
| Vocab | Tokens | Count |
|
| 123 |
|-------|--------|-------|
|
| 124 |
+
| 8k | `▁বাংলাদেশর ▁স্ থান ীয় ▁সরকারর ▁সিজিল ে ▁আসেতাই ▁জিলা ▁পরিষ ... (+21 more)` | 31 |
|
| 125 |
+
| 16k | `▁বাংলাদেশর ▁স্থানীয় ▁সরকারর ▁সিজিল ে ▁আসেতাই ▁জিলা ▁পরিষদ ▁সিটি ▁কর্পোরেশন ... (+15 more)` | 25 |
|
| 126 |
+
| 32k | `▁বাংলাদেশর ▁স্থানীয় ▁সরকারর ▁সিজিল ে ▁আসেতাই ▁জিলা ▁পরিষদ ▁সিটি ▁কর্পোরেশন ... (+15 more)` | 25 |
|
| 127 |
+
| 64k | `▁বাংলাদেশর ▁স্থানীয় ▁সরকারর ▁সিজিল ে ▁আসেতাই ▁জিলা ▁পরিষদ ▁সিটি ▁কর্পোরেশন ... (+15 more)` | 25 |
|
| 128 |
|
| 129 |
|
| 130 |
### Key Findings
|
| 131 |
|
| 132 |
+
- **Best Compression:** 64k achieves 4.935x compression
|
| 133 |
+
- **Lowest UNK Rate:** 8k with 0.2384% unknown tokens
|
| 134 |
- **Trade-off:** Larger vocabularies improve compression but increase model size
|
| 135 |
- **Recommendation:** 32k vocabulary provides optimal balance for production use
|
| 136 |
|
|
|
|
| 147 |
|
| 148 |
| N-gram | Variant | Perplexity | Entropy | Unique N-grams | Top-100 Coverage | Top-1000 Coverage |
|
| 149 |
|--------|---------|------------|---------|----------------|------------------|-------------------|
|
| 150 |
+
| **2-gram** | Word | 917 | 9.84 | 15,091 | 44.2% | 86.3% |
|
| 151 |
+
| **2-gram** | Subword | 598 🏆 | 9.22 | 14,901 | 51.1% | 92.9% |
|
| 152 |
+
| **3-gram** | Word | 1,565 | 10.61 | 31,633 | 38.0% | 79.5% |
|
| 153 |
+
| **3-gram** | Subword | 1,912 | 10.90 | 68,690 | 32.6% | 79.7% |
|
| 154 |
+
| **4-gram** | Word | 2,617 | 11.35 | 60,965 | 35.0% | 72.0% |
|
| 155 |
+
| **4-gram** | Subword | 3,535 | 11.79 | 166,549 | 26.1% | 72.8% |
|
| 156 |
+
| **5-gram** | Word | 3,304 | 11.69 | 65,705 | 33.6% | 68.3% |
|
| 157 |
+
| **5-gram** | Subword | 4,752 | 12.21 | 229,112 | 22.8% | 68.8% |
|
| 158 |
|
| 159 |
### Top 5 N-grams by Size
|
| 160 |
|
|
|
|
| 163 |
| Rank | N-gram | Count |
|
| 164 |
|------|--------|-------|
|
| 165 |
| 1 | `সাক্ষরতার হারহান` | 26,823 |
|
| 166 |
+
| 2 | `অতার মা` | 20,497 |
|
| 167 |
+
| 3 | `জনসংখ্যার উপাত্ত` | 19,704 |
|
| 168 |
| 4 | `জনসংখ্যা ইলাতাই` | 19,552 |
|
| 169 |
| 5 | `লোক গননা` | 19,533 |
|
| 170 |
|
|
|
|
| 188 |
| 4 | `অক্ষাংশ বারো দ্রাঘিমাংশ ইলতাই` | 9,366 |
|
| 189 |
| 5 | `মাপাহানর অক্ষাংশ বারো দ্রাঘিমাংশ` | 9,315 |
|
| 190 |
|
| 191 |
+
**5-grams (Word):**
|
| 192 |
+
|
| 193 |
+
| Rank | N-gram | Count |
|
| 194 |
+
|------|--------|-------|
|
| 195 |
+
| 1 | `মারির মানুলেহা লোক গননা অনুসারে` | 14,180 |
|
| 196 |
+
| 2 | `মাপাহানর অক্ষাংশ বারো দ্রাঘিমাংশ ইলতাই` | 9,315 |
|
| 197 |
+
| 3 | `এহার মাপাহানর অক্ষাংশ বারো দ্রাঘিমাংশ` | 9,310 |
|
| 198 |
+
| 4 | `এহানর গড় উচ হান ইলতাই` | 6,096 |
|
| 199 |
+
| 5 | `মান্নাহাত্ত এহানর গড় উচ হান` | 6,096 |
|
| 200 |
+
|
| 201 |
**2-grams (Subword):**
|
| 202 |
|
| 203 |
| Rank | N-gram | Count |
|
| 204 |
|------|--------|-------|
|
| 205 |
+
| 1 | `র _` | 407,202 |
|
| 206 |
+
| 2 | `। _` | 163,086 |
|
| 207 |
+
| 3 | `হা ন` | 154,676 |
|
| 208 |
+
| 4 | `ন _` | 147,838 |
|
| 209 |
+
| 5 | `_ মা` | 138,460 |
|
| 210 |
|
| 211 |
**3-grams (Subword):**
|
| 212 |
|
| 213 |
| Rank | N-gram | Count |
|
| 214 |
|------|--------|-------|
|
| 215 |
+
| 1 | `র _ মা` | 95,254 |
|
| 216 |
+
| 2 | `হা ন _` | 94,536 |
|
| 217 |
+
| 3 | `_ বা রো` | 68,915 |
|
| 218 |
+
| 4 | `বা রো _` | 68,891 |
|
| 219 |
+
| 5 | `_ ই উ` | 64,643 |
|
| 220 |
|
| 221 |
**4-grams (Subword):**
|
| 222 |
|
| 223 |
| Rank | N-gram | Count |
|
| 224 |
|------|--------|-------|
|
| 225 |
+
| 1 | `_ বা রো _` | 68,886 |
|
| 226 |
+
| 2 | `_ ই উ নি` | 64,359 |
|
| 227 |
+
| 3 | `ই উ নি য়` | 55,648 |
|
| 228 |
+
| 4 | `উ নি য় ন` | 55,615 |
|
| 229 |
+
| 5 | `জ ন সং খ্যা` | 44,873 |
|
| 230 |
+
|
| 231 |
+
**5-grams (Subword):**
|
| 232 |
+
|
| 233 |
+
| Rank | N-gram | Count |
|
| 234 |
+
|------|--------|-------|
|
| 235 |
+
| 1 | `_ ই উ নি য়` | 55,620 |
|
| 236 |
+
| 2 | `ই উ নি য় ন` | 55,614 |
|
| 237 |
+
| 3 | `_ জ ন সং খ্যা` | 44,868 |
|
| 238 |
+
| 4 | `_ উ পা ত্ত _` | 36,516 |
|
| 239 |
+
| 5 | `_ পৌ র স ভা` | 34,339 |
|
| 240 |
|
| 241 |
|
| 242 |
### Key Findings
|
| 243 |
|
| 244 |
- **Best Perplexity:** 2-gram (subword) with 598
|
| 245 |
- **Entropy Trend:** Decreases with larger n-grams (more predictable)
|
| 246 |
+
- **Coverage:** Top-1000 patterns cover ~69% of corpus
|
| 247 |
- **Recommendation:** 4-gram or 5-gram for best predictive performance
|
| 248 |
|
| 249 |
---
|
|
|
|
| 259 |
|
| 260 |
| Context | Variant | Avg Entropy | Perplexity | Branching Factor | Unique Contexts | Predictability |
|
| 261 |
|---------|---------|-------------|------------|------------------|-----------------|----------------|
|
| 262 |
+
| **1** | Word | 0.7841 | 1.722 | 4.39 | 60,191 | 21.6% |
|
| 263 |
+
| **1** | Subword | 1.0505 | 2.071 | 11.75 | 3,037 | 0.0% |
|
| 264 |
+
| **2** | Word | 0.1820 | 1.134 | 1.54 | 262,172 | 81.8% |
|
| 265 |
+
| **2** | Subword | 0.6365 | 1.555 | 3.68 | 35,639 | 36.4% |
|
| 266 |
+
| **3** | Word | 0.0756 | 1.054 | 1.27 | 399,673 | 92.4% |
|
| 267 |
+
| **3** | Subword | 0.4888 | 1.403 | 2.43 | 130,940 | 51.1% |
|
| 268 |
+
| **4** | Word | 0.0494 🏆 | 1.035 | 1.19 | 504,719 | 95.1% |
|
| 269 |
+
| **4** | Subword | 0.3613 | 1.285 | 1.77 | 317,931 | 63.9% |
|
| 270 |
|
| 271 |
### Generated Text Samples (Word-based)
|
| 272 |
|
|
|
|
| 274 |
|
| 275 |
**Context Size 1:**
|
| 276 |
|
| 277 |
+
1. `বারো জিলা বেয়াপা ১৫ ৪৪ ৮২৮ মিটার ফুট জনসংখ্যার উপাত্ত পৌরসভা আহান ভৌগলিক উপাত্ত ব্রাজিলর ঔয়াংমুঙ`
|
| 278 |
+
2. `ইউনিয়ন এগত গাঙ বারো ফুংগালাইরু বুলিয়া কিত্তাও নেই অহাত্তবারো এহার আয়তন লয়াহান ৪১৬ গ অতার মা`
|
| 279 |
+
3. `উপাত্ত শহর এহার আয়তন লয়াহান ৩৫৪ গ অতার মা হারহান ৫৯ ৫ অহানাত্ত এস নইচত জনসংখ্যার`
|
| 280 |
|
| 281 |
**Context Size 2:**
|
| 282 |
|
| 283 |
+
1. `সাক্ষরতার হারহান ৫৯ ৫ অহানাত্ত গঞ্জাম এহানর সাক্ষরতার হারহান ৭২ মুনির মা সাক্ষরতার হারহান ৬৫ বারো হু...`
|
| 284 |
+
2. `অতার মা মুনি ৫০ বারো জিলা বেয়াপা এরে পৌরসভার মানু শহরেদে বারো ১১ ৭৩৬গ গাঙেদে থাইতারা হারি`
|
| 285 |
+
3. `জনসংখ্যার উপাত্ত ভারতর মারির মানুলেহা লোক গননা অনুসারে আলসটের কাউন্টি ইংরেজি oglethorpe county এহান ...`
|
| 286 |
|
| 287 |
**Context Size 3:**
|
| 288 |
|
| 289 |
+
1. `মানুলেহা লোক গননা অনুসারে বার্বোসা পৌরসভাহানর জনসংখ্যা ইলাতাই ১০ ৪২৫ গ অতার মা মুনি ৫০ বারো জিলা বেয...`
|
| 290 |
+
2. `মারির মানুলেহা লোক গননা অনুসারে পালেসটিনা ডে গোয়াস পর্তুগীজ santa bárbara de goiás এহান ব্রাজিলর হম...`
|
| 291 |
+
3. `অতার মা মুনি ৫১ বারো জেলা বেয়াপা ৪৯ এহানাত সাক্ষরতার হারহান ৭৩ বারো জেলার মা হারহান ৬৮ আস্তা`
|
| 292 |
|
| 293 |
**Context Size 4:**
|
| 294 |
|
| 295 |
+
1. `মারির মানুলেহা লোক গননা অনুসারে টের্রেবোন পারিশ র জনসংখ্যা ইলাতাই ৮৭ ৯০৪ গ ৩২ ৭৩২গ ঘরর ইউনিট আসে হার...`
|
| 296 |
+
2. `গ অতার মা মুনি ৫২ বারো জিলা বেয়াপা এরে পৌরসভার মানু ৪২৩গ শহরেদে বারো গাঙেদে থাইতারা হারি বর্গ কিলোম...`
|
| 297 |
+
3. `মানুলেহা লোক গননা অনুসারে ক্লাবেরাস কাউন্টি র জনসংখ্যা ইলাতাই ১৮ ৫৬১ গ ঘরর ইউনিট আসে চৌদ্দাহান মুঙেদ...`
|
| 298 |
|
| 299 |
|
| 300 |
### Generated Text Samples (Subword-based)
|
|
|
|
| 303 |
|
| 304 |
**Context Size 1:**
|
| 305 |
|
| 306 |
+
1. `_মারিসিতার_ঔয়াঙেদে:_মুনিয়`
|
| 307 |
+
2. `রসভা_সাক্ষর_শহর_পৌর_ই`
|
| 308 |
+
3. `নর_অক্টোবসভার_হানিয়ন।_`
|
| 309 |
|
| 310 |
**Context Size 2:**
|
| 311 |
|
| 312 |
+
1. `র_সাই_৬৬%।_ঔয়াঙেদে_থা_`
|
| 313 |
+
2. `।_অনুসারে_৩১তম_বিয়া_জিলা`
|
| 314 |
+
3. `হান_এহান_ইউনিয়নর_সান্টা_`
|
| 315 |
|
| 316 |
**Context Size 3:**
|
| 317 |
|
| 318 |
+
1. `র_মা_সাক্ষরতার_হারহান_৫৯.`
|
| 319 |
+
2. `হান_৭৯%,_অতার_হারহান_(`
|
| 320 |
+
3. `_বারো_গাঙেদে_থাইতারা।_হারি_ব`
|
| 321 |
|
| 322 |
**Context Size 4:**
|
| 323 |
|
| 324 |
+
1. `_বারো_জিলা/বেয়াপা_(১৫-৪৪_ব`
|
| 325 |
2. `_ইউনিট_আসে।_চৌদ্দাহান_মুঙেদে:`
|
| 326 |
+
3. `ইউনিয়ন_আগ।_ভৌগলিক_উপাত্ত_`
|
| 327 |
|
| 328 |
|
| 329 |
### Key Findings
|
| 330 |
|
| 331 |
- **Best Predictability:** Context-4 (word) with 95.1% predictability
|
| 332 |
- **Branching Factor:** Decreases with context size (more deterministic)
|
| 333 |
+
- **Memory Trade-off:** Larger contexts require more storage (317,931 contexts)
|
| 334 |
- **Recommendation:** Context-3 or Context-4 for text generation
|
| 335 |
|
| 336 |
---
|
|
|
|
| 346 |
|
| 347 |
| Metric | Value |
|
| 348 |
|--------|-------|
|
| 349 |
+
| Vocabulary Size | 32,965 |
|
| 350 |
+
| Total Tokens | 2,030,616 |
|
| 351 |
+
| Mean Frequency | 61.60 |
|
| 352 |
| Median Frequency | 3 |
|
| 353 |
+
| Frequency Std Dev | 897.18 |
|
| 354 |
|
| 355 |
### Most Common Words
|
| 356 |
|
| 357 |
| Rank | Word | Frequency |
|
| 358 |
|------|------|-----------|
|
| 359 |
+
| 1 | বারো | 68,888 |
|
| 360 |
+
| 2 | ইউনিয়ন | 42,535 |
|
| 361 |
+
| 3 | উপাত্ত | 36,516 |
|
| 362 |
| 4 | হারহান | 31,910 |
|
| 363 |
+
| 5 | মা | 31,022 |
|
| 364 |
+
| 6 | মানু | 30,460 |
|
| 365 |
| 7 | সাক্ষরতার | 26,839 |
|
| 366 |
+
| 8 | গ | 26,421 |
|
| 367 |
+
| 9 | অতার | 25,584 |
|
| 368 |
+
| 10 | জনসংখ্যার | 24,823 |
|
| 369 |
|
| 370 |
### Least Common Words (from vocabulary)
|
| 371 |
|
|
|
|
| 386 |
|
| 387 |
| Metric | Value |
|
| 388 |
|--------|-------|
|
| 389 |
+
| Zipf Coefficient | 1.3137 |
|
| 390 |
+
| R² (Goodness of Fit) | 0.980288 |
|
| 391 |
| Adherence Quality | **excellent** |
|
| 392 |
|
| 393 |
### Coverage Analysis
|
|
|
|
| 403 |
|
| 404 |
- **Zipf Compliance:** R²=0.9803 indicates excellent adherence to Zipf's law
|
| 405 |
- **High Frequency Dominance:** Top 100 words cover 62.6% of corpus
|
| 406 |
+
- **Long Tail:** 22,965 words needed for remaining 3.2% coverage
|
| 407 |
|
| 408 |
---
|
| 409 |
## 5. Word Embeddings Evaluation
|
|
|
|
| 419 |
|
| 420 |
### 5.1 Cross-Lingual Alignment
|
| 421 |
|
| 422 |
+

|
| 423 |
+
|
| 424 |
+

|
| 425 |
|
| 426 |
|
| 427 |
### 5.2 Model Comparison
|
| 428 |
|
| 429 |
| Model | Dimension | Isotropy | Semantic Density | Alignment R@1 | Alignment R@10 |
|
| 430 |
|-------|-----------|----------|------------------|---------------|----------------|
|
| 431 |
+
| **mono_32d** | 32 | 0.6926 🏆 | 0.3671 | N/A | N/A |
|
| 432 |
+
| **mono_64d** | 64 | 0.5161 | 0.3444 | N/A | N/A |
|
| 433 |
+
| **mono_128d** | 128 | 0.2440 | 0.3266 | N/A | N/A |
|
| 434 |
+
| **aligned_32d** | 32 | 0.6926 | 0.3703 | 0.0100 | 0.0740 |
|
| 435 |
+
| **aligned_64d** | 64 | 0.5161 | 0.3426 | 0.0240 | 0.1200 |
|
| 436 |
+
| **aligned_128d** | 128 | 0.2440 | 0.3276 | 0.0380 | 0.1340 |
|
| 437 |
|
| 438 |
### Key Findings
|
| 439 |
|
| 440 |
+
- **Best Isotropy:** mono_32d with 0.6926 (more uniform distribution)
|
| 441 |
+
- **Semantic Density:** Average pairwise similarity of 0.3465. Lower values indicate better semantic separation.
|
| 442 |
+
- **Alignment Quality:** Aligned models achieve up to 3.8% R@1 in cross-lingual retrieval.
|
| 443 |
- **Recommendation:** 128d aligned for best cross-lingual performance
|
| 444 |
|
| 445 |
---
|
| 446 |
## 6. Morphological Analysis (Experimental)
|
| 447 |
|
|
|
|
|
|
|
| 448 |
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.
|
| 449 |
|
| 450 |
### 6.1 Productivity & Complexity
|
| 451 |
|
| 452 |
| Metric | Value | Interpretation | Recommendation |
|
| 453 |
|--------|-------|----------------|----------------|
|
| 454 |
+
| Productivity Index | **5.000** | High morphological productivity | Reliable analysis |
|
| 455 |
+
| Idiomaticity Gap | **-0.006** | Low formulaic content | - |
|
| 456 |
|
| 457 |
### 6.2 Affix Inventory (Productive Units)
|
| 458 |
|
|
|
|
| 461 |
#### Productive Prefixes
|
| 462 |
| Prefix | Examples |
|
| 463 |
|--------|----------|
|
| 464 |
+
| `-কা` | কানেডো, কাইতলী, কানিনা |
|
|
|
|
| 465 |
|
| 466 |
#### Productive Suffixes
|
| 467 |
| Suffix | Examples |
|
| 468 |
|--------|----------|
|
| 469 |
+
| `-া` | বারুইয়া, খানা, বুললা |
|
| 470 |
+
| `-র` | ০০০র, চাটমোহর, ফুর |
|
| 471 |
+
| `-়া` | বারুইয়া, বেলেয়া, ভরাপাড়া |
|
| 472 |
+
| `-য়া` | বারুইয়া, বেলেয়া, বড়হাতিয়া |
|
| 473 |
+
| `-ুর` | ফুর, গোপালপুর, সরদারপুর |
|
| 474 |
+
| `-পুর` | গোপালপুর, সরদারপুর, কুতবউল্লাপুর |
|
| 475 |
+
| `-িয়া` | বড়হাতিয়া, বাসুন্দিয়া, ঘাটলোদিয়া |
|
| 476 |
+
| `-রা` | ভাদ্রা, ভাটারা, মোরেইরা |
|
| 477 |
|
| 478 |
### 6.3 Bound Stems (Lexical Roots)
|
| 479 |
|
|
|
|
| 488 |
|
| 489 |
| Prefix | Suffix | Frequency | Examples |
|
| 490 |
|--------|--------|-----------|----------|
|
| 491 |
+
| `-কা` | `-া` | 44 words | কারোবা, কাটাৱাবা |
|
| 492 |
+
| `-কা` | `-র` | 41 words | কামর, কান্নানুর |
|
| 493 |
+
| `-কা` | `-ুর` | 15 words | কান্নানুর, কাজীপুর |
|
| 494 |
+
| `-কা` | `-়া` | 15 words | কাদিরপাড়া, কালকরিয়া |
|
| 495 |
+
| `-কা` | `-য়া` | 10 words | কালকরিয়া, কালাবাড়িয়া |
|
| 496 |
+
| `-কা` | `-িয়া` | 10 words | কালকরিয়া, কালাবাড়িয়া |
|
| 497 |
+
| `-কা` | `-পুর` | 5 words | কাজীপুর, কালিদাসপুর |
|
| 498 |
+
| `-কা` | `-রা` | 5 words | কাংরা, কাকৈরগরা |
|
|
|
|
|
|
|
| 499 |
|
| 500 |
### 6.5 Recursive Morpheme Segmentation
|
| 501 |
|
|
|
|
| 503 |
|
| 504 |
| Word | Suggested Split | Confidence | Stem |
|
| 505 |
|------|-----------------|------------|------|
|
| 506 |
+
| জাঙ্গালিয়া | **`জাঙ্গাল-িয়া`** | 4.5 | `জাঙ্গাল` |
|
| 507 |
+
| মাখদুমপুর | **`মাখদুম-পুর`** | 4.5 | `মাখদুম` |
|
| 508 |
+
| স্লোভাকিয়া | **`স্লোভাক-িয়া`** | 4.5 | `স্লোভাক` |
|
| 509 |
+
| বাল্লাপুর | **`বাল্লা-পুর`** | 4.5 | `বাল্লা` |
|
| 510 |
+
| ওসমানীয়া | **`ওসমানী-য়া`** | 4.5 | `ওসমানী` |
|
| 511 |
+
| কাসকালহেইরা | **`কা-সকালহেই-রা`** | 3.0 | `সকালহেই` |
|
| 512 |
+
| কারুপ্পুর | **`কা-রুপ্-পুর`** | 3.0 | `রুপ্` |
|
| 513 |
+
| বাহাদুরপুর | **`বাহাদ-ুর-পুর`** | 3.0 | `বাহাদ` |
|
| 514 |
+
| কাফেলান্ডিয়া | **`কা-ফেলান্ড-িয়া`** | 3.0 | `ফেলান্ড` |
|
| 515 |
+
| ইটাকোয়াটিয়ারা | **`ইটাকোয়াট-িয়া-রা`** | 3.0 | `ইটাকোয়াট` |
|
| 516 |
+
| পীরযাত্রাপুর | **`পীরযাত্-রা-পুর`** | 3.0 | `পীরযাত্` |
|
| 517 |
+
| কাসসিলান্ডিয়া | **`কা-সসিলান্ড-িয়া`** | 3.0 | `সসিলান্ড` |
|
| 518 |
+
| কাশালিয়া | **`কা-শালি-য়া`** | 3.0 | `শালি` |
|
| 519 |
| কাউন্দিয়া | **`কা-উন্দ-িয়া`** | 3.0 | `উন্দ` |
|
| 520 |
+
| কান্নানুর | **`কা-ন্নান-ুর`** | 3.0 | `ন্নান` |
|
| 521 |
|
| 522 |
### 6.6 Linguistic Interpretation
|
| 523 |
|
| 524 |
> **Automated Insight:**
|
| 525 |
+
The language Bishnupriya shows high morphological productivity. The subword models are significantly more efficient than word models, suggesting a rich system of affixation or compounding.
|
| 526 |
|
| 527 |
---
|
| 528 |
## 7. Summary & Recommendations
|
|
|
|
| 533 |
|
| 534 |
| Component | Recommended | Rationale |
|
| 535 |
|-----------|-------------|-----------|
|
| 536 |
+
| Tokenizer | **64k BPE** | Best compression (4.94x) |
|
| 537 |
| N-gram | **2-gram** | Lowest perplexity (598) |
|
| 538 |
| Markov | **Context-4** | Highest predictability (95.1%) |
|
| 539 |
| Embeddings | **100d** | Balanced semantic capture and isotropy |
|
|
|
|
| 749 |
---
|
| 750 |
*Generated by Wikilangs Models Pipeline*
|
| 751 |
|
| 752 |
+
*Report Date: 2026-01-03 19:21:34*
|
models/embeddings/aligned/bpy_128d.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:76e9bbacd4e64b12a81c21302b0ecd60d8b4224be68548745100c183b9fa9949
|
| 3 |
+
size 1035025249
|
models/embeddings/aligned/bpy_128d.meta.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"lang": "bpy", "dim": 128, "max_seq_len": 512, "is_aligned": true}
|
models/embeddings/aligned/bpy_128d.projection.npy
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8a1939689895cfb9681333ac640214fc5e7c31f1be2e3e9164d9a1a409c1412d
|
| 3 |
+
size 65664
|
models/embeddings/aligned/bpy_128d_metadata.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"language": "bpy",
|
| 3 |
+
"dimension": 128,
|
| 4 |
+
"version": "aligned",
|
| 5 |
+
"hub_language": "en",
|
| 6 |
+
"seed_vocab_size": 500,
|
| 7 |
+
"vocab_size": 10494
|
| 8 |
+
}
|
models/embeddings/aligned/bpy_32d.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:02880e8504bfc113d468394f8e5015bccd24be57edc220d8fb13185a9544f068
|
| 3 |
+
size 258965857
|
models/embeddings/aligned/bpy_32d.meta.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"lang": "bpy", "dim": 32, "max_seq_len": 512, "is_aligned": true}
|
models/embeddings/aligned/bpy_32d.projection.npy
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0f324ddb00dc3315e2b2bcffd069f69f667e874b156f3c537df44d4b77e29980
|
| 3 |
+
size 4224
|
models/embeddings/aligned/bpy_32d_metadata.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"language": "bpy",
|
| 3 |
+
"dimension": 32,
|
| 4 |
+
"version": "aligned",
|
| 5 |
+
"hub_language": "en",
|
| 6 |
+
"seed_vocab_size": 500,
|
| 7 |
+
"vocab_size": 10494
|
| 8 |
+
}
|
models/embeddings/aligned/bpy_64d.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e591b38e456d9d0604df1b9813aa0a5a1fee9aa692017ca6dd07928aceef7408
|
| 3 |
+
size 517652321
|
models/embeddings/aligned/bpy_64d.meta.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"lang": "bpy", "dim": 64, "max_seq_len": 512, "is_aligned": true}
|
models/embeddings/aligned/bpy_64d.projection.npy
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:72376aea9901daf5e408ac767e8585277a5eff77b1e6f5e989f7e53e6295cc29
|
| 3 |
+
size 16512
|
models/embeddings/aligned/bpy_64d_metadata.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"language": "bpy",
|
| 3 |
+
"dimension": 64,
|
| 4 |
+
"version": "aligned",
|
| 5 |
+
"hub_language": "en",
|
| 6 |
+
"seed_vocab_size": 500,
|
| 7 |
+
"vocab_size": 10494
|
| 8 |
+
}
|
models/embeddings/monolingual/bpy_128d.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:76e9bbacd4e64b12a81c21302b0ecd60d8b4224be68548745100c183b9fa9949
|
| 3 |
+
size 1035025249
|
models/embeddings/monolingual/bpy_128d_metadata.json
CHANGED
|
@@ -11,5 +11,5 @@
|
|
| 11 |
"encoding_method": "rope",
|
| 12 |
"dim": 128
|
| 13 |
},
|
| 14 |
-
"vocab_size":
|
| 15 |
}
|
|
|
|
| 11 |
"encoding_method": "rope",
|
| 12 |
"dim": 128
|
| 13 |
},
|
| 14 |
+
"vocab_size": 10494
|
| 15 |
}
|
models/embeddings/monolingual/bpy_32d.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:02880e8504bfc113d468394f8e5015bccd24be57edc220d8fb13185a9544f068
|
| 3 |
+
size 258965857
|
models/embeddings/monolingual/bpy_32d_metadata.json
CHANGED
|
@@ -11,5 +11,5 @@
|
|
| 11 |
"encoding_method": "rope",
|
| 12 |
"dim": 32
|
| 13 |
},
|
| 14 |
-
"vocab_size":
|
| 15 |
}
|
|
|
|
| 11 |
"encoding_method": "rope",
|
| 12 |
"dim": 32
|
| 13 |
},
|
| 14 |
+
"vocab_size": 10494
|
| 15 |
}
|
models/embeddings/monolingual/bpy_64d.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e591b38e456d9d0604df1b9813aa0a5a1fee9aa692017ca6dd07928aceef7408
|
| 3 |
+
size 517652321
|
models/embeddings/monolingual/bpy_64d_metadata.json
CHANGED
|
@@ -11,5 +11,5 @@
|
|
| 11 |
"encoding_method": "rope",
|
| 12 |
"dim": 64
|
| 13 |
},
|
| 14 |
-
"vocab_size":
|
| 15 |
}
|
|
|
|
| 11 |
"encoding_method": "rope",
|
| 12 |
"dim": 64
|
| 13 |
},
|
| 14 |
+
"vocab_size": 10494
|
| 15 |
}
|
models/subword_markov/bpy_markov_ctx1_subword.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:53ecf1194c0f35f0c1834e508bfd482989df9dca9eb28a0397d1e35c6f032909
|
| 3 |
+
size 258220
|
models/subword_markov/bpy_markov_ctx1_subword_metadata.json
CHANGED
|
@@ -2,6 +2,6 @@
|
|
| 2 |
"context_size": 1,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
-
"unique_contexts":
|
| 6 |
-
"total_transitions":
|
| 7 |
}
|
|
|
|
| 2 |
"context_size": 1,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
+
"unique_contexts": 3037,
|
| 6 |
+
"total_transitions": 9205778
|
| 7 |
}
|
models/subword_markov/bpy_markov_ctx2_subword.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:056fd69d66d4953342ad146096e757ba181309056cd0b68b2fbf373c8aecd979
|
| 3 |
+
size 1146436
|
models/subword_markov/bpy_markov_ctx2_subword_metadata.json
CHANGED
|
@@ -2,6 +2,6 @@
|
|
| 2 |
"context_size": 2,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
-
"unique_contexts":
|
| 6 |
-
"total_transitions":
|
| 7 |
}
|
|
|
|
| 2 |
"context_size": 2,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
+
"unique_contexts": 35639,
|
| 6 |
+
"total_transitions": 9180784
|
| 7 |
}
|
models/subword_markov/bpy_markov_ctx3_subword.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e798ff5c5367436ac31ec0922ef037b674b21f0dcff52db03d171df06cf91639
|
| 3 |
+
size 3028709
|
models/subword_markov/bpy_markov_ctx3_subword_metadata.json
CHANGED
|
@@ -2,6 +2,6 @@
|
|
| 2 |
"context_size": 3,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
-
"unique_contexts":
|
| 6 |
-
"total_transitions":
|
| 7 |
}
|
|
|
|
| 2 |
"context_size": 3,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
+
"unique_contexts": 130940,
|
| 6 |
+
"total_transitions": 9155790
|
| 7 |
}
|
models/subword_markov/bpy_markov_ctx4_subword.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2d6f613e1d9a83dada92d90ee93bbe0ae61671608ccce0772e2deaf7ef8fe367
|
| 3 |
+
size 6221781
|
models/subword_markov/bpy_markov_ctx4_subword_metadata.json
CHANGED
|
@@ -2,6 +2,6 @@
|
|
| 2 |
"context_size": 4,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
-
"unique_contexts":
|
| 6 |
-
"total_transitions":
|
| 7 |
}
|
|
|
|
| 2 |
"context_size": 4,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
+
"unique_contexts": 317931,
|
| 6 |
+
"total_transitions": 9130796
|
| 7 |
}
|
models/subword_ngram/bpy_2gram_subword.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2509a5963d933a51f76a6c9fc15157d1957dbefba178052f701c15189186fa7c
|
| 3 |
+
size 222766
|
models/subword_ngram/bpy_2gram_subword_metadata.json
CHANGED
|
@@ -2,6 +2,6 @@
|
|
| 2 |
"n": 2,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
-
"unique_ngrams":
|
| 6 |
-
"total_ngrams":
|
| 7 |
}
|
|
|
|
| 2 |
"n": 2,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
+
"unique_ngrams": 14901,
|
| 6 |
+
"total_ngrams": 9205778
|
| 7 |
}
|
models/subword_ngram/bpy_3gram_subword.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b4990caea4db5781e6603a792bfda6c333ae440a00091149cbee821ef7ec9063
|
| 3 |
+
size 992322
|
models/subword_ngram/bpy_3gram_subword_metadata.json
CHANGED
|
@@ -2,6 +2,6 @@
|
|
| 2 |
"n": 3,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
-
"unique_ngrams":
|
| 6 |
-
"total_ngrams":
|
| 7 |
}
|
|
|
|
| 2 |
"n": 3,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
+
"unique_ngrams": 68690,
|
| 6 |
+
"total_ngrams": 9180784
|
| 7 |
}
|
models/subword_ngram/bpy_4gram_subword.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:53e624fc4370cde44f4fc4d0c61a17e1b39044a88584a0c86f2a866bfbd8e9f3
|
| 3 |
+
size 2371272
|
models/subword_ngram/bpy_4gram_subword_metadata.json
CHANGED
|
@@ -2,6 +2,6 @@
|
|
| 2 |
"n": 4,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
-
"unique_ngrams":
|
| 6 |
-
"total_ngrams":
|
| 7 |
}
|
|
|
|
| 2 |
"n": 4,
|
| 3 |
"variant": "subword",
|
| 4 |
"language": "bpy",
|
| 5 |
+
"unique_ngrams": 166549,
|
| 6 |
+
"total_ngrams": 9155790
|
| 7 |
}
|
models/subword_ngram/bpy_5gram_subword.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3321c1aa0abf6e27986f25a3d3271867383c94b3976d1fc15af6698d2ddcdde4
|
| 3 |
+
size 3422631
|
models/subword_ngram/bpy_5gram_subword_metadata.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"n": 5,
|
| 3 |
+
"variant": "subword",
|
| 4 |
+
"language": "bpy",
|
| 5 |
+
"unique_ngrams": 229112,
|
| 6 |
+
"total_ngrams": 9130796
|
| 7 |
+
}
|
models/tokenizer/bpy_tokenizer_16k.model
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:96e3775ba39ed0bb24dd1fb3df3532777c30978b300d76e45145f9add115d314
|
| 3 |
+
size 606654
|
models/tokenizer/bpy_tokenizer_16k.vocab
CHANGED
|
The diff for this file is too large to render.
See raw diff
|
|
|
models/tokenizer/bpy_tokenizer_32k.model
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:008aabca8ea673ff2fbdee64241ebe3436f14afaa7c7c3e1d4f98400b23f0b71
|
| 3 |
+
size 1016768
|
models/tokenizer/bpy_tokenizer_32k.vocab
CHANGED
|
The diff for this file is too large to render.
See raw diff
|
|
|
models/tokenizer/bpy_tokenizer_64k.model
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:85f38e1304284a1aa434bf7cde178755d2289a27e2e8be35038d16b032b66053
|
| 3 |
+
size 1838893
|
models/tokenizer/bpy_tokenizer_64k.vocab
CHANGED
|
The diff for this file is too large to render.
See raw diff
|
|
|
models/tokenizer/bpy_tokenizer_8k.model
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:17af654476a7dac69ac244796ac2bd513f5a17692c05b58f1c4ae1fbe1a9a84d
|
| 3 |
+
size 420681
|
models/tokenizer/bpy_tokenizer_8k.vocab
CHANGED
|
The diff for this file is too large to render.
See raw diff
|
|
|
models/vocabulary/bpy_vocabulary.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b803f73ad6996d5e65aa6c22d4bc7b19f0ae1347de6a595a7408b8ce705589ae
|
| 3 |
+
size 614630
|
models/vocabulary/bpy_vocabulary_metadata.json
CHANGED
|
@@ -1,17 +1,17 @@
|
|
| 1 |
{
|
| 2 |
"language": "bpy",
|
| 3 |
-
"vocabulary_size":
|
| 4 |
"variant": "full",
|
| 5 |
"statistics": {
|
| 6 |
-
"type_token_ratio": 0.
|
| 7 |
"coverage": {
|
| 8 |
-
"top_100": 0.
|
| 9 |
-
"top_1000": 0.
|
| 10 |
-
"top_5000": 0.
|
| 11 |
-
"top_10000": 0.
|
| 12 |
},
|
| 13 |
-
"hapax_count":
|
| 14 |
-
"hapax_ratio": 0.
|
| 15 |
-
"total_documents":
|
| 16 |
}
|
| 17 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"language": "bpy",
|
| 3 |
+
"vocabulary_size": 32965,
|
| 4 |
"variant": "full",
|
| 5 |
"statistics": {
|
| 6 |
+
"type_token_ratio": 0.029294385591006916,
|
| 7 |
"coverage": {
|
| 8 |
+
"top_100": 0.6175529182866142,
|
| 9 |
+
"top_1000": 0.8867885654419936,
|
| 10 |
+
"top_5000": 0.937852810848923,
|
| 11 |
+
"top_10000": 0.9553426562620722
|
| 12 |
},
|
| 13 |
+
"hapax_count": 27321,
|
| 14 |
+
"hapax_ratio": 0.453189795309027,
|
| 15 |
+
"total_documents": 24994
|
| 16 |
}
|
| 17 |
}
|
models/word_markov/bpy_markov_ctx1_word.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e98776e04a3ecce6bef6e46a01ebfd3a3bd9fef503a13515304de7c5eb01ee20
|
| 3 |
+
size 2569061
|
models/word_markov/bpy_markov_ctx1_word_metadata.json
CHANGED
|
@@ -2,6 +2,6 @@
|
|
| 2 |
"context_size": 1,
|
| 3 |
"variant": "word",
|
| 4 |
"language": "bpy",
|
| 5 |
-
"unique_contexts":
|
| 6 |
-
"total_transitions":
|
| 7 |
}
|
|
|
|
| 2 |
"context_size": 1,
|
| 3 |
"variant": "word",
|
| 4 |
"language": "bpy",
|
| 5 |
+
"unique_contexts": 60191,
|
| 6 |
+
"total_transitions": 2032943
|
| 7 |
}
|
models/word_markov/bpy_markov_ctx2_word.parquet
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b67b084187c0f2a2c5f084edcd37766e7eef3083ae9ce4f03eb617ae6281f156
|
| 3 |
+
size 6129526
|
models/word_markov/bpy_markov_ctx2_word_metadata.json
CHANGED
|
@@ -2,6 +2,6 @@
|
|
| 2 |
"context_size": 2,
|
| 3 |
"variant": "word",
|
| 4 |
"language": "bpy",
|
| 5 |
-
"unique_contexts":
|
| 6 |
-
"total_transitions":
|
| 7 |
}
|
|
|
|
| 2 |
"context_size": 2,
|
| 3 |
"variant": "word",
|
| 4 |
"language": "bpy",
|
| 5 |
+
"unique_contexts": 262172,
|
| 6 |
+
"total_transitions": 2007949
|
| 7 |
}
|