File size: 17,440 Bytes
9a199b4 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 |
# from fastapi import FastAPI, File, UploadFile
# from fastapi.responses import JSONResponse
# import uvicorn
# import tempfile
# import nemo.collections.asr as nemo_asr
# import re
# import os
# import librosa
# import soundfile as sf
# # ===== Arabic number mapping (expanded) =====
# arabic_numbers = {
# "صفر": "0", "زيرو": "0", "٠": "0", "زيو": "0", "زير": "0",
# "واحد": "1", "واحدة": "1", "١": "1",
# "اتنين": "2", "اثنين": "2", "اثنان": "2", "٢": "2",
# "تلاتة": "3", "ثلاثة": "3", "٣": "3","ثلاث": "3","تلات": "3",
# "اربعة": "4", "أربعة": "4", "٤": "4",
# "خمسة": "5", "٥": "5","خمسه": "5",
# "ستة": "6", "٦": "6",
# "سبعة": "7", "٧": "7","سبعه": "7",
# "تمانية": "8", "ثمانية": "8", "٨": "8",
# "تسعة": "9", "٩": "9",
# "عشرة": "10", "١٠": "10","عشره": "10",
# "حداشر": "11", "احد عشر": "11", "احداشر": "11",
# "اتناشر": "12", "اثنا عشر": "12",
# "تلتاشر": "13", "ثلاثة عشر": "13",
# "اربعتاشر": "14", "أربعة عشر": "14",
# "خمستاشر": "15", "خمسة عشر": "15",
# "ستاشر": "16", "ستة عشر": "16",
# "سبعتاشر": "17", "سبعة عشر": "17",
# "طمنتاشر": "18", "ثمانية عشر": "18",
# "تسعتاشر": "19", "تسعة عشر": "19",
# "عشرين": "20", "٢٠": "20",
# "تلاتين": "30", "ثلاثين": "30", "٣٠": "30",
# "اربعين": "40", "أربعين": "40", "٤٠": "40",
# "خمسين": "50", "٥٠": "50",
# "ستين": "60", "٦٠": "60",
# "سبعين": "70", "٧٠": "70",
# "تمانين": "80", "ثمانين": "80", "٨٠": "80",
# "تسعين": "90", "٩٠": "90",
# "مية": "100", "مائة": "100", "مئة": "100", "١٠٠": "100",
# "ميتين": "200", "مائتين": "200",
# "تلاتمية": "300", "ثلاثمائة": "300",
# "اربعمية": "400", "أربعمائة": "400",
# "خمسمية": "500", "خمسمائة": "500",
# "ستمية": "600", "ستمائة": "600",
# "سبعمية": "700", "سبعمائة": "700",
# "تمانمية": "800", "ثمانمائة": "800",
# "تسعمية": "900", "تسعمائة": "900",
# "ألف": "1000", "الف": "1000", "١٠٠٠": "1000",
# "ألفين": "2000", "الفين": "2000",
# "تلات تلاف": "3000", "ثلاثة آلاف": "3000",
# "اربعة آلاف": "4000", "أربعة آلاف": "4000",
# "خمسة آلاف": "5000",
# "ستة آلاف": "6000",
# "سبعة آلاف": "7000",
# "تمانية آلاف": "8000", "ثمانية آلاف": "8000",
# "تسعة آلاف": "9000",
# "عشرة آلاف": "10000",
# "مية ألف": "100000", "مائة ألف": "100000",
# "مليون": "1000000", "ملايين": "1000000",
# "مليار": "1000000000"
# }
# # ===== Helpers =====
# def normalize_arabic(text: str) -> str:
# diacritics = re.compile(r'[\u0617-\u061A\u064B-\u0652]')
# text = re.sub(diacritics, '', text)
# text = re.sub(r'[إأآا]', 'ا', text)
# text = re.sub(r'ى', 'ي', text)
# text = re.sub(r'ؤ', 'و', text)
# text = re.sub(r'ئ', 'ي', text)
# text = re.sub(r'ة', 'ه', text)
# return text
# def replace_arabic_numbers(text: str) -> str:
# for word, digit in arabic_numbers.items():
# text = re.sub(fr"(?:^|\s){word}(?:$|\s)", f" {digit} ", text)
# return " ".join(text.split())
# def join_digit_sequences(text: str) -> str:
# tokens = text.split()
# out, buffer = [], []
# for tok in tokens:
# if tok.isdigit() and len(tok) == 1:
# buffer.append(tok)
# else:
# if buffer:
# out.append("".join(buffer))
# buffer = []
# out.append(tok)
# if buffer:
# out.append("".join(buffer))
# return " ".join(out)
# def ensure_16k_wav(input_path, output_path):
# y, sr = librosa.load(input_path, sr=16000, mono=True)
# sf.write(output_path, y, 16000)
# # ===== FastAPI app =====
# app = FastAPI(title="Arabic ASR API", description="ASR API with NeMo and Arabic digit conversion")
# @app.on_event("startup")
# def load_model():
# global asr_model
# model_path = "C:/Users/thegh/Python_Projects/Expertflow/UnderProgress/Peter_Projects/NP_Detection_Nvidia_conformer/asr-egyptian-nemo-v2.0.nemo"
# asr_model = nemo_asr.models.EncDecCTCModel.restore_from(model_path)
# @app.post("/transcribe")
# async def transcribe_audio(file: UploadFile = File(...)):
# with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
# tmp.write(await file.read())
# tmp_path = tmp.name
# # Resample to 16kHz
# resampled_path = tmp_path.replace(".wav", "_16k.wav")
# ensure_16k_wav(tmp_path, resampled_path)
# try:
# result = asr_model.transcribe([resampled_path])
# raw_text = result[0].text
# raw_text = normalize_arabic(raw_text)
# cleaned_text = replace_arabic_numbers(raw_text)
# cleaned_text = join_digit_sequences(cleaned_text)
# return JSONResponse(content={"transcription": cleaned_text})
# finally:
# os.remove(tmp_path)
# if os.path.exists(resampled_path):
# os.remove(resampled_path)
# @app.post("/transcribe-bytes")
# async def transcribe_audio_bytes(audio_bytes: bytes = File(...)):
# with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
# tmp.write(audio_bytes)
# tmp_path = tmp.name
# resampled_path = tmp_path.replace(".wav", "_16k.wav")
# ensure_16k_wav(tmp_path, resampled_path)
# try:
# result = asr_model.transcribe([resampled_path])
# raw_text = result[0].text
# raw_text = normalize_arabic(raw_text)
# cleaned_text = replace_arabic_numbers(raw_text)
# cleaned_text = join_digit_sequences(cleaned_text)
# return JSONResponse(content={"transcription": cleaned_text})
# finally:
# os.remove(tmp_path)
# if os.path.exists(resampled_path):
# os.remove(resampled_path)
# if __name__ == "__main__":
# uvicorn.run(app, host="0.0.0.0", port=8000, reload=True)
from fastapi import FastAPI, File, UploadFile
from fastapi.responses import JSONResponse
import uvicorn
import tempfile
import nemo.collections.asr as nemo_asr
import re
import os
import librosa
import soundfile as sf
# ===== Arabic + English number mapping (expanded) =====
arabic_numbers = {
# Arabic digits
"صفر": "0", "زيرو": "0", "٠": "0", "زيو": "0", "زير": "0","زينوا": "0",
"واحد": "1", "واحدة": "1", "١": "1",
"اتنين": "2", "اثنين": "2", "اثنان": "2", "٢": "2",
"تلاتة": "3", "ثلاثة": "3", "٣": "3", "ثلاث": "3", "تلات": "3",
"اربعة": "4", "أربعة": "4", "٤": "4",
"خمسة": "5", "٥": "5", "خمسه": "5",
"ستة": "6", "٦": "6",
"سبعة": "7", "٧": "7", "سبعه": "7",
"تمانية": "8", "ثمانية": "8", "٨": "8",
"تسعة": "9", "٩": "9",
"عشرة": "10", "١٠": "10", "عشره": "10",
"حداشر": "11", "احد عشر": "11", "احداشر": "11",
"اتناشر": "12", "اثنا عشر": "12",
"تلتاشر": "13", "ثلاثة عشر": "13",
"اربعتاشر": "14", "أربعة عشر": "14",
"خمستاشر": "15", "خمسة عشر": "15",
"ستاشر": "16", "ستة عشر": "16",
"سبعتاشر": "17", "سبعة عشر": "17",
"طمنتاشر": "18", "ثمانية عشر": "18",
"تسعتاشر": "19", "تسعة عشر": "19",
"عشرين": "20", "٢٠": "20",
"تلاتين": "30", "ثلاثين": "30", "٣٠": "30",
"اربعين": "40", "أربعين": "40", "٤٠": "40",
"خمسين": "50", "٥٠": "50",
"ستين": "60", "٦٠": "60",
"سبعين": "70", "٧٠": "70",
"تمانين": "80", "ثمانين": "80", "٨٠": "80",
"تسعين": "90", "٩٠": "90",
"مية": "100", "مائة": "100", "مئة": "100", "١٠٠": "100",
"ميتين": "200", "مائتين": "200",
"تلاتمية": "300", "ثلاثمائة": "300",
"اربعمية": "400", "أربعمائة": "400",
"خمسمية": "500", "خمسمائة": "500",
"ستمية": "600", "ستمائة": "600",
"سبعمية": "700", "سبعمائة": "700",
"تمانمية": "800", "ثمانمائة": "800",
"تسعمية": "900", "تسعمائة": "900",
"ألف": "1000", "الف": "1000", "١٠٠٠": "1000",
"ألفين": "2000", "الفين": "2000",
"تلات تلاف": "3000", "ثلاثة آلاف": "3000",
"اربعة آلاف": "4000", "أربعة آلاف": "4000",
"خمسة آلاف": "5000",
"ستة آلاف": "6000",
"سبعة آلاف": "7000",
"تمانية آلاف": "8000", "ثمانية آلاف": "8000",
"تسعة آلاف": "9000",
"عشرة آلاف": "10000",
"مية ألف": "100000", "مائة ألف": "100000",
"مليون": "1000000", "ملايين": "1000000",
"مليار": "1000000000",
# English digits
"zero": "0", "one": "1", "two": "2", "to": "2", "too": "2", "three": "3",
"four": "4", "for": "4", "five": "5", "six": "6", "seven": "7",
"eight": "8", "nine": "9", "ten": "10",
"eleven": "11", "twelve": "12", "thirteen": "13", "fourteen": "14",
"fifteen": "15", "sixteen": "16", "seventeen": "17",
"eighteen": "18", "nineteen": "19", "twenty": "20",
"thirty": "30", "forty": "40", "fifty": "50",
"sixty": "60", "seventy": "70", "eighty": "80", "ninety": "90",
"hundred": "100", "thousand": "1000", "million": "1000000",
# Arabic variants
"تلاته": "3", "اربعه": "4", "سته": "6",
"تمانيه": "8", "ثماني": "80", "تسعه": "9",
"واحده": "1", "عشره": "10",
"حداشر": "11", "اتناشر": "12",
"تلاته عشر": "13", "اربعه عشر": "14",
"خمسه عشر": "15", "سته عشر": "16",
"سبعه عشر": "17", "ثمانيه عشر": "18",
"تسعه عشر": "19",
# English tricky forms
"oh": "0",
"double zero": "00",
"double one": "11",
"double two": "22",
"double three": "33",
"double four": "44",
"double five": "55",
"double six": "66",
"double seven": "77",
"double eight": "88",
"double nine": "99",
"for": "4", "to": "2", "too": "2",
"nite": "9", "fiv": "5",
# 🔹 Repeated Digits
"واحد واحد": "11",
"اثنين اثنين": "22",
"اتنين اتنين": "22",
"ثلاثة ثلاثة": "33",
"تلاتة تلاتة": "33",
"أربعة أربعة": "44",
"اربعة اربعة": "44",
"خمسة خمسة": "55",
"ستة ستة": "66",
"سبعة سبعة": "77",
"ثمانية ثمانية": "88",
"تمانية تمانية": "88",
"تسعة تسعة": "99",
# 🔹 Hundreds
"مئة": "100",
"مية": "100",
"مئتين": "200",
"ميتين": "200",
"ثلاثمية": "300",
"تلتمية": "300",
"أربعمية": "400",
"اربعمية": "400",
"خمسمية": "500",
"ستمية": "600",
"سبعمية": "700",
"تمانمية": "800",
"تسعمية": "900",
# 🔹 Teens
"أحد عشر": "11",
"حداشر": "11",
"اثنا عشر": "12",
"اتناشر": "12",
"ثلاثة عشر": "13",
"تلاتاشر": "13",
"أربعة عشر": "14",
"اربعتاشر": "14",
"خمسة عشر": "15",
"خمسطاشر": "15",
"ستة عشر": "16",
"ستاشر": "16",
"سبعة عشر": "17",
"سبعتاشر": "17",
"ثمانية عشر": "18",
"طمنتاشر": "18",
"تسعة عشر": "19",
"تسعتاشر": "19",
# 🔹 Tens
"عشرين": "20",
"تلاتين": "30",
"ثلاثين": "30",
"أربعين": "40",
"اربعين": "40",
"خمسين": "50",
"ستين": "60",
"سبعين": "70",
"ثمانين": "80",
"تمانين": "80",
"ثامنين": "80",
"تسعين": "90",
# 🔹 Mixed Word + Digits
"خمسة صفر": "50",
"ثلاثة صفر صفر": "300",
"تلاتة صفر صفر": "300",
"واحد صفر": "10",
"واحد صفر واحد": "101",
"واحد اثنين": "12",
"واحد اتنين": "12",
"واحد ثلاثة": "13",
"واحد تلاتة": "13",
"واحد خمسة": "15",
"عشرة عشرة": "1010",
# Zero
"صفر": "0",
"زيرو": "0",
"زيو": "0",
"٠": "0",
# One
"واحد": "1",
"واحدة": "1",
"١": "1",
# Two
"اتنين": "2",
"اتنين": "2",
"اثنين": "2",
"اثنان": "2",
"٢": "2",
# Three
"تلاتة": "3",
"تلات": "3",
"ثلاثة": "3",
"ثلاث": "3",
"٣": "3",
# Four
"اربعة": "4",
"أربعة": "4",
"٤": "4",
# Five
"خمسة": "5",
"خمسه": "5",
"٥": "5",
# Six
"ستة": "6",
"ست": "6",
"٦": "6",
# Seven
"سبعة": "7",
"سبعه": "7",
"٧": "7",
# Eight
"تمانية": "8",
"تمنية": "8",
"ثمانية": "8",
"ثماني": "8",
"ثمان": "8",
"ثمانيّة": "8",
"ثماني": "8",
"٨": "8",
# Nine
"تسعة": "9",
"تسعه": "9",
"٩": "9"
}
# ===== Helpers =====
def normalize_arabic(text: str) -> str:
diacritics = re.compile(r'[\u0617-\u061A\u064B-\u0652]')
text = re.sub(diacritics, '', text)
text = re.sub(r'[إأآا]', 'ا', text)
text = re.sub(r'ى', 'ي', text)
text = re.sub(r'ؤ', 'و', text)
text = re.sub(r'ئ', 'ي', text)
text = re.sub(r'ة', 'ه', text)
return text
def replace_arabic_numbers(text: str) -> str:
for word, digit in arabic_numbers.items():
text = re.sub(fr"(?:^|\s){word}(?:$|\s)", f" {digit} ", text, flags=re.IGNORECASE)
return " ".join(text.split())
def join_digit_sequences(text: str) -> str:
tokens = text.split()
out, buffer = [], []
for tok in tokens:
if tok.isdigit() and len(tok) == 1:
buffer.append(tok)
else:
if buffer:
out.append("".join(buffer)) # join sequences like 8 5 -> 85
buffer = []
out.append(tok)
if buffer:
out.append("".join(buffer))
return " ".join(out)
def ensure_16k_wav(input_path, output_path):
y, sr = librosa.load(input_path, sr=16000, mono=True)
sf.write(output_path, y, 16000)
# ===== FastAPI app =====
app = FastAPI(title="Arabic ASR API", description="ASR API with NeMo and Arabic/English digit conversion")
@app.on_event("startup")
def load_model():
global asr_model
model_path = "C:/Users/thegh/Python_Projects/Expertflow/UnderProgress/Peter_Projects/NP_Detection_Nvidia_conformer/stt_ar_fastconformer_hybrid_large_pc_v1.0.nemo"
asr_model = nemo_asr.models.EncDecCTCModel.restore_from(model_path)
@app.post("/transcribe")
async def transcribe_audio(file: UploadFile = File(...)):
with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
tmp.write(await file.read())
tmp_path = tmp.name
resampled_path = tmp_path.replace(".wav", "_16k.wav")
ensure_16k_wav(tmp_path, resampled_path)
try:
result = asr_model.transcribe([resampled_path])
raw_text = result[0].text
print(raw_text)
raw_text = normalize_arabic(raw_text)
cleaned_text = replace_arabic_numbers(raw_text)
cleaned_text = join_digit_sequences(cleaned_text)
return JSONResponse(content={"transcription": cleaned_text})
finally:
os.remove(tmp_path)
if os.path.exists(resampled_path):
os.remove(resampled_path)
@app.post("/transcribe-bytes")
async def transcribe_audio_bytes(audio_bytes: bytes = File(...)):
with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
tmp.write(audio_bytes)
tmp_path = tmp.name
resampled_path = tmp_path.replace(".wav", "_16k.wav")
ensure_16k_wav(tmp_path, resampled_path)
try:
result = asr_model.transcribe([resampled_path])
raw_text = result[0].text
raw_text = normalize_arabic(raw_text)
cleaned_text = replace_arabic_numbers(raw_text)
cleaned_text = join_digit_sequences(cleaned_text)
return JSONResponse(content={"transcription": cleaned_text})
finally:
os.remove(tmp_path)
if os.path.exists(resampled_path):
os.remove(resampled_path)
if __name__ == "__main__":
uvicorn.run(app, host="0.0.0.0", port=8000, reload=True)
|