File size: 15,131 Bytes
0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 0ee6c80 41f5161 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 |
{
"cells": [
{
"cell_type": "markdown",
"metadata": {
"id": "zejqYqv5XXKN"
},
"source": [
"# 🧪 Day 01 – Sentiment-Analysis & Zero-Shot Classification with Hugging Face 🤗\n",
"\n",
"This notebook contains all the code experiments for Day 1 of my [30 Days of GenAI](https://huggingface.co/Musno/30-days-of-genai) challenge.\n",
"\n",
"_For detailed commentary and discoveries, see 👉 [Day 1 Log](https://huggingface.co/Musno/30-days-of-genai/blob/main/logs/day1.md)_\n",
"\n",
"---\n",
"\n",
"## 📌 What’s Covered Today\n",
"\n",
"- Exploring the `Sentiment-analysis-classification` pipeline\n",
" - same phrase has **different confidence scores**\n",
"- Exploring the `zero-shot-classification` pipeline\n",
"- Comparing model behavior across:\n",
" - Arabic input + Arabic labels\n",
" - Arabic input + English labels\n",
" - English input + Arabic labels\n",
" - Mixed language labels\n",
"- Observing language bias and label ordering\n",
"\n",
"---"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"id": "ZjmHyfTZIfdb"
},
"outputs": [],
"source": [
"from transformers import pipeline"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "cyyf-sdqmxkS"
},
"source": [
"### ✍️ Language & Sentiment:\n",
" This highlights how these models:\n",
"- Are heavily biased toward English\n",
"- Struggle with Arabic dialects (like Egyptian Arabic)\n",
"- Might not have seen enough emotionally expressive Arabic data during training"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "qpSZ2SiUKGgV",
"outputId": "406d85ad-c86e-4be2-c575-1d842668069e"
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"No model was supplied, defaulted to distilbert/distilbert-base-uncased-finetuned-sst-2-english and revision 714eb0f (https://huggingface.co/distilbert/distilbert-base-uncased-finetuned-sst-2-english).\n",
"Using a pipeline without specifying a model name and revision in production is not recommended.\n",
"Device set to use cpu\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"[{'label': 'POSITIVE', 'score': 0.9998656511306763}]\n",
"[{'label': 'POSITIVE', 'score': 0.5509597659111023}]\n",
"[{'label': 'NEGATIVE', 'score': 0.5022428631782532}]\n",
"[{'label': 'POSITIVE', 'score': 0.9394443035125732}]\n"
]
}
],
"source": [
"classifier = pipeline(\"sentiment-analysis\")\n",
"\n",
"english = classifier(\"I love you\")\n",
"arabic = classifier(\"أنا بحبك\")\n",
"arabic_dialect = classifier(\"انابحبك اوي\")\n",
"french = classifier(\"je t'aime\")\n",
"\n",
"print(english)\n",
"print(arabic)\n",
"print(arabic_dialect)\n",
"print(french)"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "ZMOgRJC2nVp0"
},
"source": [
"If we use a specific model would that give us better result? Probably yes but we will figure this out later during our exploration journey."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "Lqc1ANwwYwZJ"
},
"source": [
"###🧪 Test 1: Arabic Input + Arabic Labels\n",
"\n",
"Testing how the model handles Arabic input when all labels are also Arabic.\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "vKptToyTAdyv",
"outputId": "b5e59b45-5f1c-4323-af34-d048efa44e4f"
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"No model was supplied, defaulted to facebook/bart-large-mnli and revision d7645e1 (https://huggingface.co/facebook/bart-large-mnli).\n",
"Using a pipeline without specifying a model name and revision in production is not recommended.\n",
"Device set to use cpu\n"
]
},
{
"data": {
"text/plain": [
"{'sequence': 'أنا أحب تعلم الذكاء الاصطناعي',\n",
" 'labels': ['تعليم', 'طعام', 'رياضة'],\n",
" 'scores': [0.754145085811615, 0.20169343054294586, 0.04416144639253616]}"
]
},
"execution_count": 91,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"classifier = pipeline(\"zero-shot-classification\")\n",
"classifier(\n",
" \"أنا أحب تعلم الذكاء الاصطناعي\",\n",
" candidate_labels=[\"تعليم\", \"رياضة\", \"طعام\"]\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "aPTGodEkY-Ff"
},
"source": [
"Results looks incorrect because of RTL Arabic writing. Because Arabic is right-to-left, the order of the printed labels may be visually reversed. The actual top label is تعليم (education) with the highest confidence. So the model results are correct and to confirm that see the following cell output"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "bfYwofzkZzq8",
"outputId": "87c817ca-c39e-42fb-8066-60bf036f84dc"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"تعليم: 0.754\n",
"طعام: 0.202\n",
"رياضة: 0.044\n"
]
}
],
"source": [
"output = classifier(\n",
" \"أنا أحب تعلم الذكاء الاصطناعي\",\n",
" candidate_labels=[\"تعليم\", \"رياضة\", \"طعام\"]\n",
")\n",
"for label, score in zip(output['labels'], output['scores']):\n",
" print(f\"{label}: {score:.3f}\")"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "qzbuleH_aA90"
},
"source": [
"###🧪 Test 2: Arabic Input + English Labels\n",
"Same sentence as above, but now labels are in English. Checking how this affects accuracy and confidence.\n"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "omQ5DCJMD8zp",
"outputId": "2e1c2cac-393c-4579-ca03-dff1f97d8fa2"
},
"outputs": [
{
"data": {
"text/plain": [
"{'sequence': 'أنا أحب تعلم الذكاء الاصطناعي',\n",
" 'labels': ['education', 'sports', 'politics'],\n",
" 'scores': [0.49976322054862976, 0.28726592659950256, 0.21297085285186768]}"
]
},
"execution_count": 94,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"classifier(\n",
" \"أنا أحب تعلم الذكاء الاصطناعي\",\n",
" candidate_labels=[\"education\", \"sports\", \"politics\"]\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "RSjDItt5a7t_"
},
"source": [
"This result is less accurate (lower confidence), but easier to interpret — no RTL formatting confusion."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "ajm4l1RAbEXc"
},
"source": [
"###🧪 Test 3: English Input and Labels = Most Accurate (as expected)\n",
"\n",
"When both the text and labels are in English, the model performs better:"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "q4HHJjMHEY03",
"outputId": "0e47a0c2-ef5b-4f3c-c8ce-834efdf7405f"
},
"outputs": [
{
"data": {
"text/plain": [
"{'sequence': 'I love learning AI',\n",
" 'labels': ['education', 'sports', 'food'],\n",
" 'scores': [0.7564858198165894, 0.12874628603458405, 0.11476800590753555]}"
]
},
"execution_count": 95,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"classifier(\n",
" \"I love learning AI\",\n",
" candidate_labels=[\"education\", \"sports\", \"food\"]\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "e6uUPRXWbZFe"
},
"source": [
"###🧪 Test 4: Arabic Labels with English Input = Inaccurate & Low Confidence"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "Bo3CpyhbEdbc",
"outputId": "60fe6502-acae-4e85-dd75-49ae02970f9f"
},
"outputs": [
{
"data": {
"text/plain": [
"{'sequence': 'I love learning AI',\n",
" 'labels': ['طعام', 'تعليم', 'رياضة'],\n",
" 'scores': [0.37267985939979553, 0.33104342222213745, 0.296276718378067]}"
]
},
"execution_count": 96,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"classifier(\n",
" \"I love learning AI\",\n",
" candidate_labels=[\"طعام\", \"تعليم\", \"رياضة\"]\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "pS4oPUzybkH6"
},
"source": [
"The output is really confusing here because (تعليم) is in the middle that means the model didn't pick the correct word. To confirm that let's try the formatting output."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "bGhzDDv4b6rD",
"outputId": "aa7f0788-075f-4491-de29-fe70a772e41b"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"طعام: 0.373\n",
"تعليم: 0.331\n",
"رياضة: 0.296\n"
]
}
],
"source": [
"output = classifier(\"I love learning AI\", candidate_labels=[\"طعام\", \"تعليم\", \"رياضة\"])\n",
"for label, score in zip(output['labels'], output['scores']):\n",
" print(f\"{label}: {score:.3f}\")"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "XS3XnbRvcCLC"
},
"source": [
"It picked food (طعام) I dunno why, if I find out later I will update this."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "oAo5KT8zcdZw"
},
"source": [
"###🧪 Test 5: Mixed Labels with Arabic Input = A Funny Twist\n",
"\n",
"Using a mix of Arabic and English labels:"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "pomecKh8FQCY",
"outputId": "ae18001c-da1d-4bba-edd9-982201fb06ad"
},
"outputs": [
{
"data": {
"text/plain": [
"{'sequence': 'أنا أحب تعلم الذكاء الاصطناعي',\n",
" 'labels': ['طعام', 'رياضة', 'education'],\n",
" 'scores': [0.7335377335548401, 0.16061054170131683, 0.10585174709558487]}"
]
},
"execution_count": 105,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"classifier(\n",
" \"أنا أحب تعلم الذكاء الاصطناعي\",\n",
" candidate_labels=[\"education\", \"رياضة\", \"طعام\"]\n",
")"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "yKv2-idigfII"
},
"source": [
"The output resembles using Arabic labels; let's use formatting for clarification."
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "_8ez4mgkcwpR",
"outputId": "20721ea1-bd2e-4775-f591-713b2e3dda9b"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"1. طعام: 0.734\n",
"2. رياضة: 0.161\n",
"3. education: 0.106\n"
]
}
],
"source": [
"output = classifier(\"أنا أحب تعلم الذكاء الاصطناعي\", candidate_labels=[\"education\", \"رياضة\", \"طعام\"])\n",
"sorted_results = sorted(zip(output['labels'], output['scores']), key=lambda x: x[1], reverse=True)\n",
"for i, (label, score) in enumerate(sorted_results, 1):\n",
" print(f\"{i}. {label}: {score:.3f}\")"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "MPYWjgNxhCEn"
},
"source": [
"I didn't expect that tbh 🧐. I'm sure the model got the correct result, but the results will always look confusing in this case specially if you format the output. Why? I'm not sure, but I guess education wasn't counted because it's not Arabic word, and it started counting from the next Arabic word (طعام). So our model knows the right answer but it doesn't know how to represent it in the correct way. I wonder how does it work with other languages 👀"
]
}
],
"metadata": {
"colab": {
"provenance": []
},
"kernelspec": {
"display_name": "Python 3",
"name": "python3"
},
"language_info": {
"name": "python"
}
},
"nbformat": 4,
"nbformat_minor": 0
}
|