Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -27,7 +27,7 @@ from tqdm import tqdm
|
|
| 27 |
random.seed(42)
|
| 28 |
DEFAULT_LANG = "ar"
|
| 29 |
DEFAULT_NUM_QUESTIONS = 8
|
| 30 |
-
DEFAULT_TROCR_MODEL = "microsoft/trocr-base-printed"
|
| 31 |
DEFAULT_TROCR_ZOOM = 2.8
|
| 32 |
|
| 33 |
# كاش بسيط للـ OCR pipeline (تحميل كسول)
|
|
@@ -42,7 +42,7 @@ def _get_ocr_pipeline(model_id: str):
|
|
| 42 |
return _OCR_PIPE[model_id]
|
| 43 |
|
| 44 |
# =========================
|
| 45 |
-
# 2) استخراج النص من PDF
|
| 46 |
# =========================
|
| 47 |
def extract_text_with_pypdf(pdf_path: str) -> str:
|
| 48 |
reader = PdfReader(pdf_path)
|
|
@@ -88,7 +88,8 @@ def is_extraction_good(text: str, min_chars: int = 250, min_alpha_ratio: float =
|
|
| 88 |
return ratio >= min_alpha_ratio
|
| 89 |
|
| 90 |
def save_text(text: str, out_path: str) -> None:
|
| 91 |
-
os.makedirs(os.path.dirname(out_path) or ".", exist_ok=True
|
|
|
|
| 92 |
with open(out_path, "w", encoding="utf-8") as f:
|
| 93 |
f.write(text)
|
| 94 |
|
|
@@ -103,7 +104,6 @@ def pdf_to_txt(pdf_path: str, out_txt_path: str = None,
|
|
| 103 |
method = "embedded (pypdf)"
|
| 104 |
else:
|
| 105 |
if not ocr_model:
|
| 106 |
-
# وضع تجريبي بلا OCR
|
| 107 |
final_text = embedded_text
|
| 108 |
method = "embedded (pypdf: weak)"
|
| 109 |
else:
|
|
@@ -125,12 +125,9 @@ def strip_page_headers(text: str) -> str:
|
|
| 125 |
lines = text.splitlines()
|
| 126 |
out = []
|
| 127 |
for ln in lines:
|
| 128 |
-
if re2.match(r"^\s*--- \[Page \d+\] ---\s*$", ln):
|
| 129 |
-
|
| 130 |
-
if re2.match(r"^\s*
|
| 131 |
-
continue
|
| 132 |
-
if re2.match(r"^\s*[-–—_*]{3,}\s*$", ln):
|
| 133 |
-
continue
|
| 134 |
out.append(ln)
|
| 135 |
return "\n".join(out)
|
| 136 |
|
|
@@ -143,8 +140,8 @@ def normalize_arabic(text: str) -> str:
|
|
| 143 |
text = re2.sub(r"[يى]", "ي", text)
|
| 144 |
text = re2.sub(r"\s+", " ", text)
|
| 145 |
# إزالة التكرار الزائد للحروف (مثل جذرياا -> جذريا)
|
| 146 |
-
text = re2.sub(r'(\p{L})\1{2,}', r'\1', text)
|
| 147 |
-
text = re2.sub(r'(\p{L})\1', r'\1', text)
|
| 148 |
return text.strip()
|
| 149 |
|
| 150 |
def arabic_ocr_fixes(text: str) -> str:
|
|
@@ -193,14 +190,10 @@ def top_keywords_yake(text: str, max_k: int = 120, lan: str = 'ar') -> List[str]
|
|
| 193 |
seen, out = set(), []
|
| 194 |
for k in candidates:
|
| 195 |
kk = k.strip()
|
| 196 |
-
if not kk or kk in seen:
|
| 197 |
-
|
| 198 |
-
if
|
| 199 |
-
|
| 200 |
-
if len(kk) < 3:
|
| 201 |
-
continue
|
| 202 |
-
if re2.match(r"^[\p{P}\p{S}]+$", kk):
|
| 203 |
-
continue
|
| 204 |
seen.add(kk)
|
| 205 |
out.append(kk)
|
| 206 |
return out
|
|
@@ -231,26 +224,18 @@ def build_distractors(correct: str, pool: List[str], k: int = 3) -> List[str]:
|
|
| 231 |
target_len = len(correct.strip())
|
| 232 |
cand = []
|
| 233 |
for w in pool:
|
| 234 |
-
if not w:
|
| 235 |
-
continue
|
| 236 |
w2 = w.strip()
|
| 237 |
-
if w2 == correct.strip():
|
| 238 |
-
|
| 239 |
-
if
|
| 240 |
-
continue
|
| 241 |
-
if re2.match(r"^[\p{P}\p{S}\d_]+$", w2):
|
| 242 |
-
continue
|
| 243 |
-
# تقارب طولي
|
| 244 |
if abs(len(w2) - target_len) <= 3:
|
| 245 |
cand.append(w2)
|
| 246 |
-
|
| 247 |
random.shuffle(cand)
|
| 248 |
out = []
|
| 249 |
for w in cand:
|
| 250 |
out.append(w)
|
| 251 |
-
if len(out) == k:
|
| 252 |
-
break
|
| 253 |
-
|
| 254 |
fillers = ["—", "— —", "—-"]
|
| 255 |
while len(out) < k:
|
| 256 |
out.append(random.choice(fillers))
|
|
@@ -260,7 +245,6 @@ def make_mcqs_from_text(text: str, n: int = 8, lang: str = 'ar') -> List[MCQ]:
|
|
| 260 |
sentences = split_sentences(text)
|
| 261 |
if not sentences:
|
| 262 |
raise ValueError("النص قصير جدًا أو غير صالح لتوليد أسئلة.")
|
| 263 |
-
|
| 264 |
keywords = top_keywords_yake(text, max_k=160, lan=lang)
|
| 265 |
if not keywords:
|
| 266 |
toks = re2.findall(r"[\p{L}\p{N}_]+", text)
|
|
@@ -269,27 +253,20 @@ def make_mcqs_from_text(text: str, n: int = 8, lang: str = 'ar') -> List[MCQ]:
|
|
| 269 |
for t in toks:
|
| 270 |
freq[t] = freq.get(t, 0) + 1
|
| 271 |
keywords = [w for w, c in sorted(freq.items(), key=lambda x: -x[1])][:80]
|
| 272 |
-
|
| 273 |
sent_for_kw = {}
|
| 274 |
for s in sentences:
|
| 275 |
for kw in keywords:
|
| 276 |
-
if not _is_good_kw(kw):
|
| 277 |
-
continue
|
| 278 |
if re2.search(rf"(?<!\p{{L}}){re2.escape(kw)}(?!\p{{L}})", s) and kw not in sent_for_kw:
|
| 279 |
sent_for_kw[kw] = s
|
| 280 |
-
|
| 281 |
items: List[MCQ] = []
|
| 282 |
used_sents = set()
|
| 283 |
pool_iter = [kw for kw in keywords if kw in sent_for_kw]
|
| 284 |
-
|
| 285 |
for kw in pool_iter:
|
| 286 |
-
if len(items) >= n:
|
| 287 |
-
|
| 288 |
-
if not _is_good_kw(kw):
|
| 289 |
-
continue
|
| 290 |
s = sent_for_kw[kw]
|
| 291 |
-
if s in used_sents:
|
| 292 |
-
continue
|
| 293 |
blanked = re2.sub(rf"(?<!\p{{L}}){re2.escape(kw)}(?!\p{{L}})", "_____", s, count=1)
|
| 294 |
correct = kw
|
| 295 |
distractors = build_distractors(correct, [x for x in keywords if x != kw], k=3)
|
|
@@ -297,15 +274,8 @@ def make_mcqs_from_text(text: str, n: int = 8, lang: str = 'ar') -> List[MCQ]:
|
|
| 297 |
random.shuffle(choices)
|
| 298 |
ans_idx = choices.index(correct)
|
| 299 |
exp = f"مقتبس من الجملة: {s[:220]}" + ("..." if len(s) > 220 else "")
|
| 300 |
-
items.append(MCQ(
|
| 301 |
-
id=str(uuid.uuid4())[:8],
|
| 302 |
-
question=blanked,
|
| 303 |
-
choices=choices,
|
| 304 |
-
answer_index=ans_idx,
|
| 305 |
-
explanation=exp
|
| 306 |
-
))
|
| 307 |
used_sents.add(s)
|
| 308 |
-
|
| 309 |
if not items:
|
| 310 |
raise RuntimeError("تعذر توليد أسئلة من النص. جرّب نصاً أطول أو مختلفاً.")
|
| 311 |
return items
|
|
@@ -317,67 +287,196 @@ AR_PUNCT = "،؛؟"
|
|
| 317 |
EN_PUNCT = ",;?"
|
| 318 |
|
| 319 |
def normalize_punct(s: str) -> str:
|
| 320 |
-
if not s:
|
| 321 |
-
return ""
|
| 322 |
s = s.replace(",", "،").replace(";", "؛").replace("?", "؟")
|
| 323 |
return s.strip().strip(AR_PUNCT + EN_PUNCT).strip()
|
| 324 |
|
| 325 |
def is_bad_choice(txt: str) -> bool:
|
| 326 |
-
if not txt:
|
| 327 |
-
return True
|
| 328 |
txt = txt.strip()
|
| 329 |
BAD_NOISE = {"وهنا","اليه","الي","ليبق","لان","لانها","لانّه","ذلك","هذا","هذه"}
|
| 330 |
-
if txt in BAD_NOISE:
|
| 331 |
-
|
| 332 |
-
if len(txt)
|
| 333 |
-
|
| 334 |
-
if
|
| 335 |
-
return True
|
| 336 |
-
if txt in AR_STOP:
|
| 337 |
-
return True
|
| 338 |
-
if re2.match(r"^[\p{P}\p{S}]+$", txt):
|
| 339 |
-
return True
|
| 340 |
return False
|
| 341 |
|
| 342 |
def build_json_records(items: List[MCQ], lang: str, source_pdf: str, method: str, num_questions: int):
|
| 343 |
json_data = []
|
| 344 |
letters = ["A", "B", "C", "D"]
|
| 345 |
for it in items:
|
| 346 |
-
opts = []
|
| 347 |
-
seen = set()
|
| 348 |
for idx, lbl in enumerate(letters):
|
| 349 |
raw = it.choices[idx] if idx < len(it.choices) else ""
|
| 350 |
txt = normalize_punct(raw)
|
| 351 |
-
if is_bad_choice(txt):
|
| 352 |
-
|
| 353 |
-
if txt in seen:
|
| 354 |
-
txt += " "
|
| 355 |
seen.add(txt)
|
| 356 |
-
opts.append({
|
| 357 |
-
"id": lbl,
|
| 358 |
-
"text": txt,
|
| 359 |
-
"is_correct": (it.answer_index == idx)
|
| 360 |
-
})
|
| 361 |
q_clean = normalize_punct(it.question)
|
| 362 |
exp_clean = normalize_punct(it.explanation)
|
| 363 |
record = {
|
| 364 |
-
"id": it.id,
|
| 365 |
-
"
|
| 366 |
-
"options": opts,
|
| 367 |
-
"explanation": exp_clean,
|
| 368 |
-
"meta": {
|
| 369 |
-
"lang": lang,
|
| 370 |
-
"normalized": True,
|
| 371 |
-
"source_pdf": source_pdf,
|
| 372 |
-
"extraction_method": method,
|
| 373 |
-
"num_questions": int(num_questions),
|
| 374 |
-
}
|
| 375 |
}
|
| 376 |
json_data.append(record)
|
| 377 |
return json_data
|
| 378 |
|
| 379 |
# =========================
|
| 380 |
-
# 7)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 381 |
# =========================
|
| 382 |
def process_pdf(pdf_file_path,
|
| 383 |
num_questions=DEFAULT_NUM_QUESTIONS,
|
|
@@ -389,16 +488,13 @@ def process_pdf(pdf_file_path,
|
|
| 389 |
if not pdf_file_path:
|
| 390 |
return {}, None, "يرجى رفع ملف PDF/TXT أولاً."
|
| 391 |
|
| 392 |
-
# pdf_file_path قد يكون str أو NamedString -> خذه كمسار
|
| 393 |
src_path = str(pdf_file_path)
|
| 394 |
name_guess = getattr(pdf_file_path, "name", "") if hasattr(pdf_file_path, "name") else ""
|
| 395 |
filename = Path(name_guess).name or Path(src_path).name or "input"
|
| 396 |
workdir = tempfile.mkdtemp(prefix="mcq_")
|
| 397 |
|
| 398 |
-
# تأكد من الامتداد
|
| 399 |
ext = Path(filename).suffix.lower()
|
| 400 |
if ext not in [".pdf", ".txt"]:
|
| 401 |
-
# حاول تخمين نوعه، افتراض PDF
|
| 402 |
ext = ".pdf"
|
| 403 |
if not Path(filename).suffix:
|
| 404 |
filename += ext
|
|
@@ -407,7 +503,7 @@ def process_pdf(pdf_file_path,
|
|
| 407 |
shutil.copy(src_path, local_path)
|
| 408 |
logs.append(f"تم نسخ الملف إلى: {local_path}")
|
| 409 |
|
| 410 |
-
# 1) استخراج النص
|
| 411 |
if ext == ".txt":
|
| 412 |
with open(local_path, "r", encoding="utf-8", errors="ignore") as f:
|
| 413 |
raw_text = f.read()
|
|
@@ -448,45 +544,102 @@ def process_pdf(pdf_file_path,
|
|
| 448 |
return {}, None, "\n".join(logs)
|
| 449 |
|
| 450 |
# =========================
|
| 451 |
-
#
|
| 452 |
# =========================
|
| 453 |
import gradio as gr
|
| 454 |
|
| 455 |
-
with gr.Blocks(title="PDF/TXT → MCQ
|
| 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 |
if __name__ == "__main__":
|
| 492 |
demo.queue().launch()
|
|
|
|
| 27 |
random.seed(42)
|
| 28 |
DEFAULT_LANG = "ar"
|
| 29 |
DEFAULT_NUM_QUESTIONS = 8
|
| 30 |
+
DEFAULT_TROCR_MODEL = "microsoft/trocr-base-printed"
|
| 31 |
DEFAULT_TROCR_ZOOM = 2.8
|
| 32 |
|
| 33 |
# كاش بسيط للـ OCR pipeline (تحميل كسول)
|
|
|
|
| 42 |
return _OCR_PIPE[model_id]
|
| 43 |
|
| 44 |
# =========================
|
| 45 |
+
# 2) استخراج النص من PDF/TXT
|
| 46 |
# =========================
|
| 47 |
def extract_text_with_pypdf(pdf_path: str) -> str:
|
| 48 |
reader = PdfReader(pdf_path)
|
|
|
|
| 88 |
return ratio >= min_alpha_ratio
|
| 89 |
|
| 90 |
def save_text(text: str, out_path: str) -> None:
|
| 91 |
+
os.makedirs(os.path.dirname(out_path) or ".", exist_ok=True
|
| 92 |
+
)
|
| 93 |
with open(out_path, "w", encoding="utf-8") as f:
|
| 94 |
f.write(text)
|
| 95 |
|
|
|
|
| 104 |
method = "embedded (pypdf)"
|
| 105 |
else:
|
| 106 |
if not ocr_model:
|
|
|
|
| 107 |
final_text = embedded_text
|
| 108 |
method = "embedded (pypdf: weak)"
|
| 109 |
else:
|
|
|
|
| 125 |
lines = text.splitlines()
|
| 126 |
out = []
|
| 127 |
for ln in lines:
|
| 128 |
+
if re2.match(r"^\s*--- \[Page \d+\] ---\s*$", ln): continue
|
| 129 |
+
if re2.match(r"^\s*(Page\s*\d+|صفحة\s*\d+)\s*$", ln): continue
|
| 130 |
+
if re2.match(r"^\s*[-–—_*]{3,}\s*$", ln): continue
|
|
|
|
|
|
|
|
|
|
| 131 |
out.append(ln)
|
| 132 |
return "\n".join(out)
|
| 133 |
|
|
|
|
| 140 |
text = re2.sub(r"[يى]", "ي", text)
|
| 141 |
text = re2.sub(r"\s+", " ", text)
|
| 142 |
# إزالة التكرار الزائد للحروف (مثل جذرياا -> جذريا)
|
| 143 |
+
text = re2.sub(r'(\p{L})\1{2,}', r'\1', text)
|
| 144 |
+
text = re2.sub(r'(\p{L})\1', r'\1', text)
|
| 145 |
return text.strip()
|
| 146 |
|
| 147 |
def arabic_ocr_fixes(text: str) -> str:
|
|
|
|
| 190 |
seen, out = set(), []
|
| 191 |
for k in candidates:
|
| 192 |
kk = k.strip()
|
| 193 |
+
if not kk or kk in seen: continue
|
| 194 |
+
if lan == "ar" and kk in AR_STOP: continue
|
| 195 |
+
if len(kk) < 3: continue
|
| 196 |
+
if re2.match(r"^[\p{P}\p{S}]+$", kk): continue
|
|
|
|
|
|
|
|
|
|
|
|
|
| 197 |
seen.add(kk)
|
| 198 |
out.append(kk)
|
| 199 |
return out
|
|
|
|
| 224 |
target_len = len(correct.strip())
|
| 225 |
cand = []
|
| 226 |
for w in pool:
|
| 227 |
+
if not w: continue
|
|
|
|
| 228 |
w2 = w.strip()
|
| 229 |
+
if w2 == correct.strip(): continue
|
| 230 |
+
if len(w2) < 3 or w2 in AR_STOP: continue
|
| 231 |
+
if re2.match(r"^[\p{P}\p{S}\d_]+$", w2): continue
|
|
|
|
|
|
|
|
|
|
|
|
|
| 232 |
if abs(len(w2) - target_len) <= 3:
|
| 233 |
cand.append(w2)
|
|
|
|
| 234 |
random.shuffle(cand)
|
| 235 |
out = []
|
| 236 |
for w in cand:
|
| 237 |
out.append(w)
|
| 238 |
+
if len(out) == k: break
|
|
|
|
|
|
|
| 239 |
fillers = ["—", "— —", "—-"]
|
| 240 |
while len(out) < k:
|
| 241 |
out.append(random.choice(fillers))
|
|
|
|
| 245 |
sentences = split_sentences(text)
|
| 246 |
if not sentences:
|
| 247 |
raise ValueError("النص قصير جدًا أو غير صالح لتوليد أسئلة.")
|
|
|
|
| 248 |
keywords = top_keywords_yake(text, max_k=160, lan=lang)
|
| 249 |
if not keywords:
|
| 250 |
toks = re2.findall(r"[\p{L}\p{N}_]+", text)
|
|
|
|
| 253 |
for t in toks:
|
| 254 |
freq[t] = freq.get(t, 0) + 1
|
| 255 |
keywords = [w for w, c in sorted(freq.items(), key=lambda x: -x[1])][:80]
|
|
|
|
| 256 |
sent_for_kw = {}
|
| 257 |
for s in sentences:
|
| 258 |
for kw in keywords:
|
| 259 |
+
if not _is_good_kw(kw): continue
|
|
|
|
| 260 |
if re2.search(rf"(?<!\p{{L}}){re2.escape(kw)}(?!\p{{L}})", s) and kw not in sent_for_kw:
|
| 261 |
sent_for_kw[kw] = s
|
|
|
|
| 262 |
items: List[MCQ] = []
|
| 263 |
used_sents = set()
|
| 264 |
pool_iter = [kw for kw in keywords if kw in sent_for_kw]
|
|
|
|
| 265 |
for kw in pool_iter:
|
| 266 |
+
if len(items) >= n: break
|
| 267 |
+
if not _is_good_kw(kw): continue
|
|
|
|
|
|
|
| 268 |
s = sent_for_kw[kw]
|
| 269 |
+
if s in used_sents: continue
|
|
|
|
| 270 |
blanked = re2.sub(rf"(?<!\p{{L}}){re2.escape(kw)}(?!\p{{L}})", "_____", s, count=1)
|
| 271 |
correct = kw
|
| 272 |
distractors = build_distractors(correct, [x for x in keywords if x != kw], k=3)
|
|
|
|
| 274 |
random.shuffle(choices)
|
| 275 |
ans_idx = choices.index(correct)
|
| 276 |
exp = f"مقتبس من الجملة: {s[:220]}" + ("..." if len(s) > 220 else "")
|
| 277 |
+
items.append(MCQ(id=str(uuid.uuid4())[:8], question=blanked, choices=choices, answer_index=ans_idx, explanation=exp))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 278 |
used_sents.add(s)
|
|
|
|
| 279 |
if not items:
|
| 280 |
raise RuntimeError("تعذر توليد أسئلة من النص. جرّب نصاً أطول أو مختلفاً.")
|
| 281 |
return items
|
|
|
|
| 287 |
EN_PUNCT = ",;?"
|
| 288 |
|
| 289 |
def normalize_punct(s: str) -> str:
|
| 290 |
+
if not s: return ""
|
|
|
|
| 291 |
s = s.replace(",", "،").replace(";", "؛").replace("?", "؟")
|
| 292 |
return s.strip().strip(AR_PUNCT + EN_PUNCT).strip()
|
| 293 |
|
| 294 |
def is_bad_choice(txt: str) -> bool:
|
| 295 |
+
if not txt: return True
|
|
|
|
| 296 |
txt = txt.strip()
|
| 297 |
BAD_NOISE = {"وهنا","اليه","الي","ليبق","لان","لانها","لانّه","ذلك","هذا","هذه"}
|
| 298 |
+
if txt in BAD_NOISE: return True
|
| 299 |
+
if len(txt) > 18 and " " not in txt: return True
|
| 300 |
+
if len(txt) < 2: return True
|
| 301 |
+
if txt in AR_STOP: return True
|
| 302 |
+
if re2.match(r"^[\p{P}\p{S}]+$", txt): return True
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 303 |
return False
|
| 304 |
|
| 305 |
def build_json_records(items: List[MCQ], lang: str, source_pdf: str, method: str, num_questions: int):
|
| 306 |
json_data = []
|
| 307 |
letters = ["A", "B", "C", "D"]
|
| 308 |
for it in items:
|
| 309 |
+
opts, seen = [], set()
|
|
|
|
| 310 |
for idx, lbl in enumerate(letters):
|
| 311 |
raw = it.choices[idx] if idx < len(it.choices) else ""
|
| 312 |
txt = normalize_punct(raw)
|
| 313 |
+
if is_bad_choice(txt): txt = "—"
|
| 314 |
+
if txt in seen: txt += " "
|
|
|
|
|
|
|
| 315 |
seen.add(txt)
|
| 316 |
+
opts.append({"id": lbl, "text": txt, "is_correct": (it.answer_index == idx)})
|
|
|
|
|
|
|
|
|
|
|
|
|
| 317 |
q_clean = normalize_punct(it.question)
|
| 318 |
exp_clean = normalize_punct(it.explanation)
|
| 319 |
record = {
|
| 320 |
+
"id": it.id, "question": q_clean, "options": opts, "explanation": exp_clean,
|
| 321 |
+
"meta": {"lang": lang, "normalized": True, "source_pdf": source_pdf, "extraction_method": method, "num_questions": int(num_questions)}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 322 |
}
|
| 323 |
json_data.append(record)
|
| 324 |
return json_data
|
| 325 |
|
| 326 |
# =========================
|
| 327 |
+
# 7) دوال تبويب "حلّ الاختبار"
|
| 328 |
+
# =========================
|
| 329 |
+
def _format_question(rec):
|
| 330 |
+
q = rec.get("question","").strip()
|
| 331 |
+
return f"### السؤال:\n{q}"
|
| 332 |
+
|
| 333 |
+
def _radio_choices(rec):
|
| 334 |
+
# يعيد قائمة نصوص مثل "A) ...", "B) ..."
|
| 335 |
+
letters = ["A","B","C","D"]
|
| 336 |
+
out = []
|
| 337 |
+
for opt in rec.get("options", []):
|
| 338 |
+
lid, text = opt.get("id",""), opt.get("text","")
|
| 339 |
+
out.append(f"{lid}) {text}")
|
| 340 |
+
# إذا ناقص خيارات، كمّل لمواءمة المكوّن
|
| 341 |
+
while len(out) < 4:
|
| 342 |
+
out.append(f"{letters[len(out)]}) —")
|
| 343 |
+
return out
|
| 344 |
+
|
| 345 |
+
def _correct_letter(rec):
|
| 346 |
+
for opt in rec.get("options", []):
|
| 347 |
+
if opt.get("is_correct"):
|
| 348 |
+
return opt.get("id","")
|
| 349 |
+
return ""
|
| 350 |
+
|
| 351 |
+
def _explanation(rec):
|
| 352 |
+
return rec.get("explanation","")
|
| 353 |
+
|
| 354 |
+
def init_quiz_state(records):
|
| 355 |
+
# ترتيب عشوائي اختياري هنا (يمكن إبقاء كما هو)
|
| 356 |
+
# random.shuffle(records)
|
| 357 |
+
return {
|
| 358 |
+
"records": records,
|
| 359 |
+
"idx": 0,
|
| 360 |
+
"answers": {}, # id السؤال -> "A"/"B"/"C"/"D"
|
| 361 |
+
"revealed": set(), # ids تم إظهار حلّها
|
| 362 |
+
"finished": False,
|
| 363 |
+
"csv_path": None
|
| 364 |
+
}
|
| 365 |
+
|
| 366 |
+
def render_current(rec, user_choice=None, revealed=False):
|
| 367 |
+
q_md = _format_question(rec)
|
| 368 |
+
choices = _radio_choices(rec)
|
| 369 |
+
exp = _explanation(rec) if revealed else ""
|
| 370 |
+
progress = ""
|
| 371 |
+
correct = _correct_letter(rec)
|
| 372 |
+
feedback = ""
|
| 373 |
+
if user_choice:
|
| 374 |
+
if revealed:
|
| 375 |
+
feedback = "✅ إجابة صحيحة" if user_choice == correct else f"❌ إجابة خاطئة — الصحيح: {correct}"
|
| 376 |
+
else:
|
| 377 |
+
feedback = f"تم اختيار: {user_choice}"
|
| 378 |
+
return q_md, choices, exp, feedback
|
| 379 |
+
|
| 380 |
+
def on_start_quiz(json_records):
|
| 381 |
+
if not json_records or not isinstance(json_records, list):
|
| 382 |
+
return None, "لم يتم العثور على أسئلة صالحة."
|
| 383 |
+
return init_quiz_state(json_records), "تم بدء الاختبار. بالتوفيق!"
|
| 384 |
+
|
| 385 |
+
def on_load_json_file(file_path):
|
| 386 |
+
if not file_path: return None, "لم يتم اختيار ملف."
|
| 387 |
+
try:
|
| 388 |
+
with open(str(file_path), "r", encoding="utf-8") as f:
|
| 389 |
+
data = json.load(f)
|
| 390 |
+
if not isinstance(data, list): raise ValueError("صيغة JSON غير صحيحة (يجب أن تكون قائمة).")
|
| 391 |
+
return init_quiz_state(data), "تم تحميل ملف JSON بنجاح. اضغط بدء الاختبار."
|
| 392 |
+
except Exception as e:
|
| 393 |
+
return None, f"خطأ في قراءة JSON: {e}"
|
| 394 |
+
|
| 395 |
+
def on_show_question(state):
|
| 396 |
+
if not state: return "", [], "", "",""
|
| 397 |
+
recs, idx = state["records"], state["idx"]
|
| 398 |
+
rec = recs[idx]
|
| 399 |
+
q_md, choices, exp, feedback = render_current(
|
| 400 |
+
rec,
|
| 401 |
+
user_choice=state["answers"].get(rec["id"]),
|
| 402 |
+
revealed=(rec["id"] in state["revealed"])
|
| 403 |
+
)
|
| 404 |
+
pos = f"{idx+1} / {len(recs)}"
|
| 405 |
+
return q_md, choices, exp, feedback, pos
|
| 406 |
+
|
| 407 |
+
def on_select_choice(state, choice_label):
|
| 408 |
+
if not state or not choice_label: return state, ""
|
| 409 |
+
rec = state["records"][state["idx"]]
|
| 410 |
+
# choice_label على شكل "A) نص"
|
| 411 |
+
chosen_letter = choice_label.split(")")[0].strip()
|
| 412 |
+
state["answers"][rec["id"]] = chosen_letter
|
| 413 |
+
if rec["id"] in state["revealed"]:
|
| 414 |
+
# أعِد توليد الفيدباك
|
| 415 |
+
correct = _correct_letter(rec)
|
| 416 |
+
fb = "✅ إجابة صحيحة" if chosen_letter == correct else f"❌ إجابة خاطئة — الصحيح: {correct}"
|
| 417 |
+
else:
|
| 418 |
+
fb = f"تم اختيار: {chosen_letter}"
|
| 419 |
+
return state, fb
|
| 420 |
+
|
| 421 |
+
def on_prev(state):
|
| 422 |
+
if not state: return state
|
| 423 |
+
state["idx"] = max(0, state["idx"]-1)
|
| 424 |
+
return state
|
| 425 |
+
|
| 426 |
+
def on_next(state):
|
| 427 |
+
if not state: return state
|
| 428 |
+
state["idx"] = min(len(state["records"])-1, state["idx"]+1)
|
| 429 |
+
return state
|
| 430 |
+
|
| 431 |
+
def on_reveal(state):
|
| 432 |
+
if not state: return state, ""
|
| 433 |
+
rec = state["records"][state["idx"]]
|
| 434 |
+
state["revealed"].add(rec["id"])
|
| 435 |
+
user = state["answers"].get(rec["id"])
|
| 436 |
+
correct = _correct_letter(rec)
|
| 437 |
+
fb = "✅ إجابة صحيحة" if user == correct else (f"❌ إجابة خاطئة — الصحيح: {correct}" if user else f"الصحيح: {correct}")
|
| 438 |
+
return state, fb
|
| 439 |
+
|
| 440 |
+
def on_finish(state):
|
| 441 |
+
if not state: return state, "", None
|
| 442 |
+
recs = state["records"]
|
| 443 |
+
correct_count, wrong_count, skipped = 0,0,0
|
| 444 |
+
rows = []
|
| 445 |
+
for rec in recs:
|
| 446 |
+
qid = rec["id"]
|
| 447 |
+
user = state["answers"].get(qid)
|
| 448 |
+
correct = _correct_letter(rec)
|
| 449 |
+
is_correct = (user == correct) if user else False
|
| 450 |
+
if user is None: skipped += 1
|
| 451 |
+
elif is_correct: correct_count += 1
|
| 452 |
+
else: wrong_count += 1
|
| 453 |
+
# صف للـ CSV
|
| 454 |
+
# جمع النصوص للخيارات
|
| 455 |
+
opts = {opt["id"]: opt["text"] for opt in rec.get("options", [])}
|
| 456 |
+
rows.append({
|
| 457 |
+
"question": rec.get("question",""),
|
| 458 |
+
"A": opts.get("A",""), "B": opts.get("B",""),
|
| 459 |
+
"C": opts.get("C",""), "D": opts.get("D",""),
|
| 460 |
+
"user_choice": user or "",
|
| 461 |
+
"correct": correct,
|
| 462 |
+
"is_correct": bool(is_correct)
|
| 463 |
+
})
|
| 464 |
+
total = len(recs)
|
| 465 |
+
score = f"النتيجة: {correct_count}/{total} (صحيح: {correct_count}، خطأ: {wrong_count}، متروك: {skipped})"
|
| 466 |
+
# CSV
|
| 467 |
+
df = pd.DataFrame(rows)
|
| 468 |
+
workdir = tempfile.mkdtemp(prefix="quiz_")
|
| 469 |
+
csv_path = os.path.join(workdir, "results.csv")
|
| 470 |
+
df.to_csv(csv_path, index=False, encoding="utf-8-sig")
|
| 471 |
+
state["finished"] = True
|
| 472 |
+
state["csv_path"] = csv_path
|
| 473 |
+
return state, score, csv_path
|
| 474 |
+
|
| 475 |
+
def on_reset():
|
| 476 |
+
return None, "", "", "", "", "", None, "تمت إعادة الضبط."
|
| 477 |
+
|
| 478 |
+
# =========================
|
| 479 |
+
# 8) التبويب الأول: توليد الأسئلة (PDF/TXT → JSON)
|
| 480 |
# =========================
|
| 481 |
def process_pdf(pdf_file_path,
|
| 482 |
num_questions=DEFAULT_NUM_QUESTIONS,
|
|
|
|
| 488 |
if not pdf_file_path:
|
| 489 |
return {}, None, "يرجى رفع ملف PDF/TXT أولاً."
|
| 490 |
|
|
|
|
| 491 |
src_path = str(pdf_file_path)
|
| 492 |
name_guess = getattr(pdf_file_path, "name", "") if hasattr(pdf_file_path, "name") else ""
|
| 493 |
filename = Path(name_guess).name or Path(src_path).name or "input"
|
| 494 |
workdir = tempfile.mkdtemp(prefix="mcq_")
|
| 495 |
|
|
|
|
| 496 |
ext = Path(filename).suffix.lower()
|
| 497 |
if ext not in [".pdf", ".txt"]:
|
|
|
|
| 498 |
ext = ".pdf"
|
| 499 |
if not Path(filename).suffix:
|
| 500 |
filename += ext
|
|
|
|
| 503 |
shutil.copy(src_path, local_path)
|
| 504 |
logs.append(f"تم نسخ الملف إلى: {local_path}")
|
| 505 |
|
| 506 |
+
# 1) استخراج النص
|
| 507 |
if ext == ".txt":
|
| 508 |
with open(local_path, "r", encoding="utf-8", errors="ignore") as f:
|
| 509 |
raw_text = f.read()
|
|
|
|
| 544 |
return {}, None, "\n".join(logs)
|
| 545 |
|
| 546 |
# =========================
|
| 547 |
+
# 9) واجهة Gradio (تبويبان)
|
| 548 |
# =========================
|
| 549 |
import gradio as gr
|
| 550 |
|
| 551 |
+
with gr.Blocks(title="PDF/TXT → MCQ + Quiz", css="""
|
| 552 |
+
body { direction: rtl; font-family: system-ui, 'Cairo', 'IBM Plex Arabic', sans-serif; }
|
| 553 |
+
label, .gr-markdown { text-align: right; }
|
| 554 |
+
""") as demo:
|
| 555 |
+
gr.Markdown("## مولّد أسئلة + واجهة اختبار تفاعلي")
|
| 556 |
+
|
| 557 |
+
# حالة مشتركة بين التبويبين
|
| 558 |
+
quiz_state = gr.State(value=None) # سيحمل dict من init_quiz_state(...)
|
| 559 |
+
toast = gr.Markdown("")
|
| 560 |
+
|
| 561 |
+
with gr.Tabs():
|
| 562 |
+
# --- تبويب 1: توليد الأسئلة ---
|
| 563 |
+
with gr.TabItem("توليد الأسئلة (PDF/TXT → JSON)"):
|
| 564 |
+
with gr.Row():
|
| 565 |
+
inp_pdf = gr.File(label="ارفع PDF أو TXT", file_count="single", file_types=[".pdf",".txt"], type="filepath")
|
| 566 |
+
with gr.Column():
|
| 567 |
+
num_q = gr.Slider(4, 20, value=DEFAULT_NUM_QUESTIONS, step=1, label="عدد الأسئلة")
|
| 568 |
+
trocr_zoom = gr.Slider(2.0, 3.5, value=DEFAULT_TROCR_ZOOM, step=0.1, label="دقة تحويل PDF لصور (Zoom)")
|
| 569 |
+
trocr_model = gr.Dropdown(
|
| 570 |
+
choices=[
|
| 571 |
+
"microsoft/trocr-base-printed",
|
| 572 |
+
"microsoft/trocr-large-printed",
|
| 573 |
+
"microsoft/trocr-base-handwritten",
|
| 574 |
+
"microsoft/trocr-large-handwritten",
|
| 575 |
+
],
|
| 576 |
+
value=DEFAULT_TROCR_MODEL, label="موديل TrOCR (للـ PDF المصوّر)"
|
| 577 |
+
)
|
| 578 |
+
btn_gen = gr.Button("تشغيل المعالجة", variant="primary")
|
| 579 |
+
out_json = gr.JSON(label="النتيجة (JSON)")
|
| 580 |
+
out_file = gr.File(label="تحميل mcqs.json")
|
| 581 |
+
out_log = gr.Textbox(label="Logs", lines=10)
|
| 582 |
+
btn_send_to_quiz = gr.Button("إرسال الأسئلة إلى تبويب الاختبار")
|
| 583 |
+
|
| 584 |
+
btn_gen.click(
|
| 585 |
+
fn=process_pdf,
|
| 586 |
+
inputs=[inp_pdf, num_q, gr.State(DEFAULT_LANG), trocr_model, trocr_zoom],
|
| 587 |
+
outputs=[out_json, out_file, out_log]
|
| 588 |
)
|
| 589 |
|
| 590 |
+
# إرسال الناتج مباشرة إلى التبويب الثاني
|
| 591 |
+
def _send_to_quiz(records):
|
| 592 |
+
if not records: return None, "لا يوجد أسئلة لإرسالها."
|
| 593 |
+
return init_quiz_state(records), "تم إرسال الأسئلة إلى تبويب الاختبار. افتحه واضغط 'إظهار السؤال'."
|
| 594 |
+
btn_send_to_quiz.click(_send_to_quiz, inputs=[out_json], outputs=[quiz_state, toast])
|
| 595 |
+
|
| 596 |
+
# --- تبويب 2: حلّ الاختبار ---
|
| 597 |
+
with gr.TabItem("حلّ الاختبار (Quiz)"):
|
| 598 |
+
gr.Markdown("### 1) حمّل JSON للأسئلة أو استخدم زر الإرسال من التبويب الأول")
|
| 599 |
+
json_file = gr.File(label="أو ارفع ملف JSON", file_types=[".json"], type="filepath")
|
| 600 |
+
btn_load_json = gr.Button("تحميل ملف JSON")
|
| 601 |
+
btn_start = gr.Button("بدء الاختبار", variant="primary")
|
| 602 |
+
|
| 603 |
+
gr.Markdown("### 2) حل السؤال الحالي")
|
| 604 |
+
q_md = gr.Markdown("")
|
| 605 |
+
choices = gr.Radio(choices=[], label="اختر الإجابة")
|
| 606 |
+
exp_md = gr.Markdown("")
|
| 607 |
+
feedback = gr.Markdown("")
|
| 608 |
+
progress = gr.Label("")
|
| 609 |
+
|
| 610 |
+
with gr.Row():
|
| 611 |
+
btn_prev = gr.Button("السابق")
|
| 612 |
+
btn_next = gr.Button("التالي")
|
| 613 |
+
btn_reveal = gr.Button("إظهار الإجابة")
|
| 614 |
+
with gr.Row():
|
| 615 |
+
btn_finish = gr.Button("إنهاء الاختبار", variant="stop")
|
| 616 |
+
btn_reset = gr.Button("إعادة ضبط")
|
| 617 |
+
|
| 618 |
+
score_md = gr.Markdown("")
|
| 619 |
+
results_csv = gr.File(label="تحميل نتائج CSV")
|
| 620 |
+
|
| 621 |
+
# ربط الأزرار بالدوال
|
| 622 |
+
btn_load_json.click(on_load_json_file, inputs=[json_file], outputs=[quiz_state, toast])
|
| 623 |
+
btn_start.click(on_start_quiz, inputs=[quiz_state], outputs=[quiz_state, toast])
|
| 624 |
+
# عرض السؤال الحالي
|
| 625 |
+
def _show_and_render(state):
|
| 626 |
+
return on_show_question(state)
|
| 627 |
+
# عند البدء أو التنقل أو الإظهار نعيد رندر
|
| 628 |
+
btn_start.click(_show_and_render, inputs=[quiz_state], outputs=[q_md, choices, exp_md, feedback, progress])
|
| 629 |
+
btn_prev.click(on_prev, inputs=[quiz_state], outputs=[quiz_state]).then(_show_and_render, inputs=[quiz_state], outputs=[q_md, choices, exp_md, feedback, progress])
|
| 630 |
+
btn_next.click(on_next, inputs=[quiz_state], outputs=[quiz_state]).then(_show_and_render, inputs=[quiz_state], outputs=[q_md, choices, exp_md, feedback, progress])
|
| 631 |
+
btn_reveal.click(on_reveal, inputs=[quiz_state], outputs=[quiz_state, feedback]).then(_show_and_render, inputs=[quiz_state], outputs=[q_md, choices, exp_md, feedback, progress])
|
| 632 |
+
|
| 633 |
+
# اختيار الإجابة
|
| 634 |
+
def _on_choice(state, choice):
|
| 635 |
+
return on_select_choice(state, choice)
|
| 636 |
+
choices.change(_on_choice, inputs=[quiz_state, choices], outputs=[quiz_state, feedback])
|
| 637 |
+
|
| 638 |
+
# إنهاء
|
| 639 |
+
btn_finish.click(on_finish, inputs=[quiz_state], outputs=[quiz_state, score_md, results_csv])
|
| 640 |
+
# إعادة ضبط
|
| 641 |
+
btn_reset.click(on_reset, outputs=[quiz_state, q_md, exp_md, feedback, progress, score_md, results_csv, toast])
|
| 642 |
+
|
| 643 |
+
# Spaces تتعرف على demo تلقائيًا
|
| 644 |
if __name__ == "__main__":
|
| 645 |
demo.queue().launch()
|