Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -13,13 +13,11 @@ from dataclasses import dataclass
|
|
| 13 |
from pathlib import Path
|
| 14 |
from typing import List, Tuple
|
| 15 |
|
| 16 |
-
import pandas as pd
|
| 17 |
from PIL import Image
|
| 18 |
from pypdf import PdfReader
|
| 19 |
import fitz # PyMuPDF
|
| 20 |
import regex as re2
|
| 21 |
import yake
|
| 22 |
-
from tqdm import tqdm
|
| 23 |
|
| 24 |
# =========================
|
| 25 |
# إعدادات عامة
|
|
@@ -66,12 +64,11 @@ def pdf_pages_to_images(pdf_path: str, zoom: float = 2.5) -> List[Image.Image]:
|
|
| 66 |
doc.close()
|
| 67 |
return imgs
|
| 68 |
|
| 69 |
-
def extract_text_with_ocr(pdf_path: str, model_id: str, zoom: float = 2.5
|
| 70 |
ocr = _get_ocr_pipeline(model_id)
|
| 71 |
images = pdf_pages_to_images(pdf_path, zoom=zoom)
|
| 72 |
page_texts = []
|
| 73 |
-
|
| 74 |
-
for idx, img in enumerate(pbar):
|
| 75 |
try:
|
| 76 |
out = ocr(img)
|
| 77 |
txt = out[0]["generated_text"].strip() if out and "generated_text" in out[0] else ""
|
|
@@ -87,36 +84,19 @@ def is_extraction_good(text: str, min_chars: int = 250, min_alpha_ratio: float =
|
|
| 87 |
ratio = alnum / max(1, len(text))
|
| 88 |
return ratio >= min_alpha_ratio
|
| 89 |
|
| 90 |
-
def
|
| 91 |
-
|
| 92 |
-
|
| 93 |
-
|
| 94 |
-
|
| 95 |
-
|
| 96 |
-
def pdf_to_txt(pdf_path: str, out_txt_path: str = None,
|
| 97 |
-
ocr_model: str = DEFAULT_TROCR_MODEL,
|
| 98 |
-
ocr_zoom: float = DEFAULT_TROCR_ZOOM) -> Tuple[str, str, str]:
|
| 99 |
assert os.path.isfile(pdf_path), f"File not found: {pdf_path}"
|
| 100 |
-
|
| 101 |
embedded_text = extract_text_with_pypdf(pdf_path)
|
| 102 |
if is_extraction_good(embedded_text):
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
final_text = embedded_text
|
| 108 |
-
method = "embedded (pypdf: weak)"
|
| 109 |
-
else:
|
| 110 |
-
final_text = extract_text_with_ocr(pdf_path, model_id=ocr_model, zoom=ocr_zoom)
|
| 111 |
-
method = "OCR (Hugging Face TrOCR)"
|
| 112 |
-
|
| 113 |
-
if out_txt_path is None:
|
| 114 |
-
base, _ = os.path.splitext(pdf_path)
|
| 115 |
-
out_txt_path = base + ".txt"
|
| 116 |
-
|
| 117 |
-
header = f"[[ Extraction method: {method} ]]\n\n"
|
| 118 |
-
save_text(header + final_text, out_txt_path)
|
| 119 |
-
return final_text, out_txt_path, method
|
| 120 |
|
| 121 |
# =========================
|
| 122 |
# 3) تطبيع/تصحيح عربي
|
|
@@ -139,7 +119,7 @@ def normalize_arabic(text: str) -> str:
|
|
| 139 |
text = re2.sub(r"[إأآا]", "ا", text)
|
| 140 |
text = re2.sub(r"[يى]", "ي", text)
|
| 141 |
text = re2.sub(r"\s+", " ", text)
|
| 142 |
-
# إزالة
|
| 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()
|
|
@@ -281,7 +261,7 @@ def make_mcqs_from_text(text: str, n: int = 8, lang: str = 'ar') -> List[MCQ]:
|
|
| 281 |
return items
|
| 282 |
|
| 283 |
# =========================
|
| 284 |
-
# 6)
|
| 285 |
# =========================
|
| 286 |
AR_PUNCT = "،؛؟"
|
| 287 |
EN_PUNCT = ",;?"
|
|
@@ -291,127 +271,82 @@ def normalize_punct(s: str) -> str:
|
|
| 291 |
s = s.replace(",", "،").replace(";", "؛").replace("?", "؟")
|
| 292 |
return s.strip().strip(AR_PUNCT + EN_PUNCT).strip()
|
| 293 |
|
| 294 |
-
def
|
| 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
|
| 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,
|
| 321 |
-
"
|
|
|
|
|
|
|
|
|
|
| 322 |
}
|
| 323 |
json_data.append(record)
|
| 324 |
return json_data
|
| 325 |
|
| 326 |
# =========================
|
| 327 |
-
# 7)
|
| 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 |
-
|
| 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 |
-
|
| 373 |
-
|
| 374 |
-
|
| 375 |
-
|
| 376 |
-
|
| 377 |
-
|
| 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 |
-
|
| 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:
|
|
@@ -438,207 +373,141 @@ def on_reveal(state):
|
|
| 438 |
return state, fb
|
| 439 |
|
| 440 |
def on_finish(state):
|
| 441 |
-
if not state: return state, ""
|
| 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
|
| 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
|
| 473 |
-
return state, score, csv_path
|
| 474 |
|
| 475 |
def on_reset():
|
| 476 |
-
return None, "", "", "", "", "",
|
| 477 |
|
| 478 |
# =========================
|
| 479 |
-
# 8)
|
| 480 |
# =========================
|
| 481 |
-
def
|
| 482 |
-
|
| 483 |
-
|
| 484 |
-
|
| 485 |
-
|
| 486 |
-
|
| 487 |
-
|
| 488 |
-
|
| 489 |
-
|
| 490 |
-
|
| 491 |
-
|
| 492 |
-
|
| 493 |
-
|
| 494 |
-
|
| 495 |
-
|
| 496 |
-
|
| 497 |
-
|
| 498 |
-
|
| 499 |
-
|
| 500 |
-
|
| 501 |
-
|
| 502 |
-
local_path = os.path.join(workdir, filename)
|
| 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()
|
| 510 |
-
method = "plain text (no PDF)"
|
| 511 |
-
else:
|
| 512 |
-
raw_text, out_txt_path, method = pdf_to_txt(
|
| 513 |
-
pdf_path=local_path,
|
| 514 |
-
ocr_model=trocr_model,
|
| 515 |
-
ocr_zoom=float(trocr_zoom)
|
| 516 |
-
)
|
| 517 |
-
logs.append(f"طريقة الاستخراج: {method}")
|
| 518 |
-
|
| 519 |
-
# 2) تنظيف/تطبيع
|
| 520 |
-
cleaned_text = postprocess_text(raw_text, lang=lang)
|
| 521 |
-
save_text(cleaned_text, os.path.join(workdir, "cleaned.txt"))
|
| 522 |
-
logs.append("تم تنظيف/تطبيع النص.")
|
| 523 |
-
|
| 524 |
-
# 3) توليد أسئلة
|
| 525 |
-
items = make_mcqs_from_text(cleaned_text, n=int(num_questions), lang=lang)
|
| 526 |
-
logs.append(f"تم توليد {len(items)} سؤالاً.")
|
| 527 |
-
|
| 528 |
-
# 4) بناء JSON
|
| 529 |
-
json_records = build_json_records(
|
| 530 |
-
items, lang=lang, source_pdf=Path(filename).name, method=method, num_questions=num_questions
|
| 531 |
-
)
|
| 532 |
-
json_str = json.dumps(json_records, ensure_ascii=False, indent=2)
|
| 533 |
-
|
| 534 |
-
# 5) حفظ ملف JSON للتنزيل
|
| 535 |
-
json_path = os.path.join(workdir, "mcqs.json")
|
| 536 |
-
with open(json_path, "w", encoding="utf-8") as fj:
|
| 537 |
-
fj.write(json_str)
|
| 538 |
-
logs.append("تم إنشاء ملف mcqs.json.")
|
| 539 |
-
|
| 540 |
-
return json_records, json_path, "\n".join(logs)
|
| 541 |
|
| 542 |
-
|
| 543 |
-
|
| 544 |
-
|
|
|
|
| 545 |
|
| 546 |
# =========================
|
| 547 |
-
# 9) واجهة Gradio (
|
| 548 |
# =========================
|
| 549 |
import gradio as gr
|
| 550 |
|
| 551 |
-
|
| 552 |
body { direction: rtl; font-family: system-ui, 'Cairo', 'IBM Plex Arabic', sans-serif; }
|
| 553 |
-
label, .gr-markdown { text-align: right; }
|
| 554 |
-
|
| 555 |
-
|
| 556 |
-
|
| 557 |
-
|
| 558 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 559 |
toast = gr.Markdown("")
|
| 560 |
|
| 561 |
-
with gr.
|
| 562 |
-
|
| 563 |
-
|
| 564 |
-
|
| 565 |
-
|
| 566 |
-
|
| 567 |
-
|
| 568 |
-
|
| 569 |
-
|
| 570 |
-
|
| 571 |
-
|
| 572 |
-
|
| 573 |
-
|
| 574 |
-
|
| 575 |
-
|
| 576 |
-
|
| 577 |
-
|
| 578 |
-
|
| 579 |
-
|
| 580 |
-
|
| 581 |
-
|
| 582 |
-
|
| 583 |
-
|
| 584 |
-
|
| 585 |
-
|
| 586 |
-
|
| 587 |
-
|
| 588 |
-
)
|
| 589 |
-
|
| 590 |
-
|
| 591 |
-
|
| 592 |
-
|
| 593 |
-
|
| 594 |
-
|
| 595 |
-
|
| 596 |
-
|
| 597 |
-
|
| 598 |
-
|
| 599 |
-
|
| 600 |
-
|
| 601 |
-
|
| 602 |
-
|
| 603 |
-
|
| 604 |
-
|
| 605 |
-
|
| 606 |
-
|
| 607 |
-
|
| 608 |
-
|
| 609 |
-
|
| 610 |
-
|
| 611 |
-
|
| 612 |
-
|
| 613 |
-
|
| 614 |
-
|
| 615 |
-
|
| 616 |
-
|
| 617 |
-
|
| 618 |
-
|
| 619 |
-
|
| 620 |
-
|
| 621 |
-
|
| 622 |
-
|
| 623 |
-
|
| 624 |
-
|
| 625 |
-
|
| 626 |
-
|
| 627 |
-
|
| 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__":
|
|
|
|
| 13 |
from pathlib import Path
|
| 14 |
from typing import List, Tuple
|
| 15 |
|
|
|
|
| 16 |
from PIL import Image
|
| 17 |
from pypdf import PdfReader
|
| 18 |
import fitz # PyMuPDF
|
| 19 |
import regex as re2
|
| 20 |
import yake
|
|
|
|
| 21 |
|
| 22 |
# =========================
|
| 23 |
# إعدادات عامة
|
|
|
|
| 64 |
doc.close()
|
| 65 |
return imgs
|
| 66 |
|
| 67 |
+
def extract_text_with_ocr(pdf_path: str, model_id: str, zoom: float = 2.5) -> str:
|
| 68 |
ocr = _get_ocr_pipeline(model_id)
|
| 69 |
images = pdf_pages_to_images(pdf_path, zoom=zoom)
|
| 70 |
page_texts = []
|
| 71 |
+
for idx, img in enumerate(images):
|
|
|
|
| 72 |
try:
|
| 73 |
out = ocr(img)
|
| 74 |
txt = out[0]["generated_text"].strip() if out and "generated_text" in out[0] else ""
|
|
|
|
| 84 |
ratio = alnum / max(1, len(text))
|
| 85 |
return ratio >= min_alpha_ratio
|
| 86 |
|
| 87 |
+
def pdf_to_text(pdf_path: str,
|
| 88 |
+
ocr_model: str = DEFAULT_TROCR_MODEL,
|
| 89 |
+
ocr_zoom: float = DEFAULT_TROCR_ZOOM) -> Tuple[str, str]:
|
| 90 |
+
"""
|
| 91 |
+
يرجع (النص النهائي، طريقة الاستخراج) بدون أي حفظ ملفات.
|
| 92 |
+
"""
|
|
|
|
|
|
|
|
|
|
| 93 |
assert os.path.isfile(pdf_path), f"File not found: {pdf_path}"
|
|
|
|
| 94 |
embedded_text = extract_text_with_pypdf(pdf_path)
|
| 95 |
if is_extraction_good(embedded_text):
|
| 96 |
+
return embedded_text, "embedded (pypdf)"
|
| 97 |
+
if not ocr_model:
|
| 98 |
+
return embedded_text, "embedded (pypdf: weak)"
|
| 99 |
+
return extract_text_with_ocr(pdf_path, model_id=ocr_model, zoom=ocr_zoom), "OCR (Hugging Face TrOCR)"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 100 |
|
| 101 |
# =========================
|
| 102 |
# 3) تطبيع/تصحيح عربي
|
|
|
|
| 119 |
text = re2.sub(r"[إأآا]", "ا", text)
|
| 120 |
text = re2.sub(r"[يى]", "ي", text)
|
| 121 |
text = re2.sub(r"\s+", " ", text)
|
| 122 |
+
# إزالة تكرار الحروف
|
| 123 |
text = re2.sub(r'(\p{L})\1{2,}', r'\1', text)
|
| 124 |
text = re2.sub(r'(\p{L})\1', r'\1', text)
|
| 125 |
return text.strip()
|
|
|
|
| 261 |
return items
|
| 262 |
|
| 263 |
# =========================
|
| 264 |
+
# 6) تحويل عناصر الأسئلة إلى سجلات لواجهة الحلّ
|
| 265 |
# =========================
|
| 266 |
AR_PUNCT = "،؛؟"
|
| 267 |
EN_PUNCT = ",;?"
|
|
|
|
| 271 |
s = s.replace(",", "،").replace(";", "؛").replace("?", "؟")
|
| 272 |
return s.strip().strip(AR_PUNCT + EN_PUNCT).strip()
|
| 273 |
|
| 274 |
+
def build_quiz_records(items: List[MCQ], lang: str, source_name: str, method: str, num_questions: int):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 275 |
json_data = []
|
| 276 |
letters = ["A", "B", "C", "D"]
|
| 277 |
for it in items:
|
| 278 |
+
opts = []
|
| 279 |
for idx, lbl in enumerate(letters):
|
| 280 |
raw = it.choices[idx] if idx < len(it.choices) else ""
|
| 281 |
+
txt = normalize_punct(raw) or "—"
|
|
|
|
|
|
|
|
|
|
| 282 |
opts.append({"id": lbl, "text": txt, "is_correct": (it.answer_index == idx)})
|
| 283 |
q_clean = normalize_punct(it.question)
|
| 284 |
exp_clean = normalize_punct(it.explanation)
|
| 285 |
record = {
|
| 286 |
+
"id": it.id,
|
| 287 |
+
"question": q_clean,
|
| 288 |
+
"options": opts,
|
| 289 |
+
"explanation": exp_clean,
|
| 290 |
+
"meta": {"lang": lang, "source": source_name, "extraction_method": method, "num_questions": int(num_questions)}
|
| 291 |
}
|
| 292 |
json_data.append(record)
|
| 293 |
return json_data
|
| 294 |
|
| 295 |
# =========================
|
| 296 |
+
# 7) منطق الاختبار (State + Handlers)
|
| 297 |
# =========================
|
| 298 |
def _format_question(rec):
|
| 299 |
q = rec.get("question","").strip()
|
| 300 |
return f"### السؤال:\n{q}"
|
| 301 |
|
| 302 |
def _radio_choices(rec):
|
|
|
|
|
|
|
| 303 |
out = []
|
| 304 |
for opt in rec.get("options", []):
|
| 305 |
lid, text = opt.get("id",""), opt.get("text","")
|
| 306 |
out.append(f"{lid}) {text}")
|
|
|
|
| 307 |
while len(out) < 4:
|
| 308 |
+
letters = ["A","B","C","D"]
|
| 309 |
out.append(f"{letters[len(out)]}) —")
|
| 310 |
return out
|
| 311 |
|
| 312 |
def _correct_letter(rec):
|
| 313 |
for opt in rec.get("options", []):
|
| 314 |
+
if opt.get("is_correct"): return opt.get("id","")
|
|
|
|
| 315 |
return ""
|
| 316 |
|
| 317 |
+
def _explanation(rec): return rec.get("explanation","")
|
|
|
|
| 318 |
|
| 319 |
def init_quiz_state(records):
|
| 320 |
+
return {"records": records, "idx": 0, "answers": {}, "revealed": set(), "finished": False}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 321 |
|
| 322 |
def render_current(rec, user_choice=None, revealed=False):
|
| 323 |
q_md = _format_question(rec)
|
| 324 |
choices = _radio_choices(rec)
|
| 325 |
exp = _explanation(rec) if revealed else ""
|
|
|
|
| 326 |
correct = _correct_letter(rec)
|
| 327 |
+
if user_choice and revealed:
|
| 328 |
+
feedback = "✅ إجابة صحيحة" if user_choice == correct else f"❌ إجابة خاطئة — الصحيح: {correct}"
|
| 329 |
+
elif user_choice:
|
| 330 |
+
feedback = f"تم اختيار: {user_choice}"
|
| 331 |
+
else:
|
| 332 |
+
feedback = ""
|
| 333 |
return q_md, choices, exp, feedback
|
| 334 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 335 |
def on_show_question(state):
|
| 336 |
if not state: return "", [], "", "",""
|
| 337 |
recs, idx = state["records"], state["idx"]
|
| 338 |
rec = recs[idx]
|
| 339 |
+
q_md, choices, exp, feedback = render_current(rec, user_choice=state["answers"].get(rec["id"]),
|
| 340 |
+
revealed=(rec["id"] in state["revealed"]))
|
|
|
|
|
|
|
|
|
|
| 341 |
pos = f"{idx+1} / {len(recs)}"
|
| 342 |
return q_md, choices, exp, feedback, pos
|
| 343 |
|
| 344 |
def on_select_choice(state, choice_label):
|
| 345 |
if not state or not choice_label: return state, ""
|
| 346 |
rec = state["records"][state["idx"]]
|
|
|
|
| 347 |
chosen_letter = choice_label.split(")")[0].strip()
|
| 348 |
state["answers"][rec["id"]] = chosen_letter
|
| 349 |
if rec["id"] in state["revealed"]:
|
|
|
|
| 350 |
correct = _correct_letter(rec)
|
| 351 |
fb = "✅ إجابة صحيحة" if chosen_letter == correct else f"❌ إجابة خاطئة — الصحيح: {correct}"
|
| 352 |
else:
|
|
|
|
| 373 |
return state, fb
|
| 374 |
|
| 375 |
def on_finish(state):
|
| 376 |
+
if not state: return state, ""
|
| 377 |
recs = state["records"]
|
| 378 |
correct_count, wrong_count, skipped = 0,0,0
|
|
|
|
| 379 |
for rec in recs:
|
| 380 |
qid = rec["id"]
|
| 381 |
user = state["answers"].get(qid)
|
| 382 |
correct = _correct_letter(rec)
|
|
|
|
| 383 |
if user is None: skipped += 1
|
| 384 |
+
elif user == correct: correct_count += 1
|
| 385 |
else: wrong_count += 1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 386 |
total = len(recs)
|
| 387 |
score = f"النتيجة: {correct_count}/{total} (صحيح: {correct_count}، خطأ: {wrong_count}، متروك: {skipped})"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 388 |
state["finished"] = True
|
| 389 |
+
return state, score
|
|
|
|
| 390 |
|
| 391 |
def on_reset():
|
| 392 |
+
return None, "", "", "", "", "", "تمت إعادة الضبط."
|
| 393 |
|
| 394 |
# =========================
|
| 395 |
+
# 8) معالجة الملف وبناء الأسئلة (بدون أي ملفات ناتجة)
|
| 396 |
# =========================
|
| 397 |
+
def process_input_file(uploaded_path,
|
| 398 |
+
num_questions=DEFAULT_NUM_QUESTIONS,
|
| 399 |
+
lang=DEFAULT_LANG,
|
| 400 |
+
trocr_model=DEFAULT_TROCR_MODEL,
|
| 401 |
+
trocr_zoom=DEFAULT_TROCR_ZOOM):
|
| 402 |
+
if not uploaded_path:
|
| 403 |
+
return None, "يرجى رفع ملف PDF/TXT أولاً."
|
| 404 |
+
src_path = str(uploaded_path)
|
| 405 |
+
filename = Path(src_path).name or "input"
|
| 406 |
+
ext = Path(filename).suffix.lower()
|
| 407 |
+
if ext not in [".pdf", ".txt"]:
|
| 408 |
+
return None, "الرجاء رفع PDF أو TXT فقط."
|
| 409 |
+
|
| 410 |
+
# قراءة النص
|
| 411 |
+
if ext == ".txt":
|
| 412 |
+
with open(src_path, "r", encoding="utf-8", errors="ignore") as f:
|
| 413 |
+
raw_text = f.read()
|
| 414 |
+
method = "plain text (no PDF)"
|
| 415 |
+
else:
|
| 416 |
+
raw_text, method = pdf_to_text(src_path, ocr_model=trocr_model, ocr_zoom=float(trocr_zoom))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 417 |
|
| 418 |
+
cleaned_text = postprocess_text(raw_text, lang=lang)
|
| 419 |
+
items = make_mcqs_from_text(cleaned_text, n=int(num_questions), lang=lang)
|
| 420 |
+
records = build_quiz_records(items, lang=lang, source_name=filename, method=method, num_questions=num_questions)
|
| 421 |
+
return init_quiz_state(records), f"تم توليد {len(records)} سؤالاً. بالتوفيق!"
|
| 422 |
|
| 423 |
# =========================
|
| 424 |
+
# 9) واجهة Gradio (تبويب واحد)
|
| 425 |
# =========================
|
| 426 |
import gradio as gr
|
| 427 |
|
| 428 |
+
THEME_CSS = """
|
| 429 |
body { direction: rtl; font-family: system-ui, 'Cairo', 'IBM Plex Arabic', sans-serif; }
|
| 430 |
+
label, .gr-markdown, .gr-button { text-align: right; }
|
| 431 |
+
.gradio-container { max-width: 880px; margin: auto; }
|
| 432 |
+
.card { background: #fff; border-radius: 1rem; padding: 1rem 1.2rem; box-shadow: 0 10px 25px rgba(0,0,0,0.06); }
|
| 433 |
+
.small { opacity: .85; font-size: .9rem; }
|
| 434 |
+
.progress { text-align: left; opacity:.75 }
|
| 435 |
+
"""
|
| 436 |
+
|
| 437 |
+
with gr.Blocks(title="اختبار من ملف (PDF/TXT)", css=THEME_CSS) as demo:
|
| 438 |
+
gr.Markdown("## ✨ صانع اختبار من ملف PDF/TXT — واجهة واحدة بسيطة")
|
| 439 |
+
gr.Markdown("ارفع ملفك، حدّد عدد الأسئلة، واضغط **ابدأ**. ثمّ أجب وتحقق من الإجابة.")
|
| 440 |
+
|
| 441 |
+
quiz_state = gr.State(value=None)
|
| 442 |
toast = gr.Markdown("")
|
| 443 |
|
| 444 |
+
with gr.Row():
|
| 445 |
+
inp_file = gr.File(label="ارفع ملف PDF أو TXT", file_count="single", file_types=[".pdf",".txt"], type="filepath")
|
| 446 |
+
num_q = gr.Slider(4, 20, value=DEFAULT_NUM_QUESTIONS, step=1, label="عدد الأسئلة")
|
| 447 |
+
with gr.Accordion("خيارات متقدمة (للـ PDF المصوّر)", open=False):
|
| 448 |
+
trocr_zoom = gr.Slider(2.0, 3.5, value=DEFAULT_TROCR_ZOOM, step=0.1, label="Zoom لتحويل الصفحات لصورة (OCR)")
|
| 449 |
+
trocr_model = gr.Dropdown(
|
| 450 |
+
choices=[
|
| 451 |
+
"microsoft/trocr-base-printed",
|
| 452 |
+
"microsoft/trocr-large-printed",
|
| 453 |
+
"microsoft/trocr-base-handwritten",
|
| 454 |
+
"microsoft/trocr-large-handwritten",
|
| 455 |
+
],
|
| 456 |
+
value=DEFAULT_TROCR_MODEL, label="نموذج TrOCR"
|
| 457 |
+
)
|
| 458 |
+
|
| 459 |
+
btn_start = gr.Button("ابدأ توليد الاختبار", variant="primary")
|
| 460 |
+
|
| 461 |
+
with gr.Group():
|
| 462 |
+
with gr.Row():
|
| 463 |
+
progress = gr.Label("", elem_classes=["progress"])
|
| 464 |
+
q_md = gr.Markdown("", elem_classes=["card"])
|
| 465 |
+
choices = gr.Radio(choices=[], label="اختر الإجابة", interactive=True)
|
| 466 |
+
feedback = gr.Markdown("")
|
| 467 |
+
exp_md = gr.Markdown("")
|
| 468 |
+
with gr.Row():
|
| 469 |
+
btn_prev = gr.Button("السابق")
|
| 470 |
+
btn_next = gr.Button("التالي")
|
| 471 |
+
btn_reveal = gr.Button("إظهار الإجابة")
|
| 472 |
+
btn_finish = gr.Button("إنهاء الاختبار", variant="stop")
|
| 473 |
+
btn_reset = gr.Button("إعادة ضبط")
|
| 474 |
+
|
| 475 |
+
score_md = gr.Markdown("")
|
| 476 |
+
|
| 477 |
+
# بدء المعالجة وبناء الأسئلة
|
| 478 |
+
btn_start.click(
|
| 479 |
+
process_input_file,
|
| 480 |
+
inputs=[inp_file, num_q, gr.State(DEFAULT_LANG), trocr_model, trocr_zoom],
|
| 481 |
+
outputs=[quiz_state, toast]
|
| 482 |
+
).then(
|
| 483 |
+
on_show_question, inputs=[quiz_state],
|
| 484 |
+
outputs=[q_md, choices, exp_md, feedback, progress]
|
| 485 |
+
)
|
| 486 |
+
|
| 487 |
+
# التنقل
|
| 488 |
+
btn_prev.click(on_prev, inputs=[quiz_state], outputs=[quiz_state]).then(
|
| 489 |
+
on_show_question, inputs=[quiz_state],
|
| 490 |
+
outputs=[q_md, choices, exp_md, feedback, progress]
|
| 491 |
+
)
|
| 492 |
+
btn_next.click(on_next, inputs=[quiz_state], outputs=[quiz_state]).then(
|
| 493 |
+
on_show_question, inputs=[quiz_state],
|
| 494 |
+
outputs=[q_md, choices, exp_md, feedback, progress]
|
| 495 |
+
)
|
| 496 |
+
btn_reveal.click(on_reveal, inputs=[quiz_state], outputs=[quiz_state, feedback]).then(
|
| 497 |
+
on_show_question, inputs=[quiz_state],
|
| 498 |
+
outputs=[q_md, choices, exp_md, feedback, progress]
|
| 499 |
+
)
|
| 500 |
+
|
| 501 |
+
# اختيار الإجابة
|
| 502 |
+
def _on_choice(state, choice):
|
| 503 |
+
return on_select_choice(state, choice)
|
| 504 |
+
choices.change(_on_choice, inputs=[quiz_state, choices], outputs=[quiz_state, feedback])
|
| 505 |
+
|
| 506 |
+
# إنهاء وإظهار نتيجة
|
| 507 |
+
btn_finish.click(on_finish, inputs=[quiz_state], outputs=[quiz_state, score_md])
|
| 508 |
+
|
| 509 |
+
# إعادة ضبط
|
| 510 |
+
btn_reset.click(lambda: on_reset(), outputs=[quiz_state, q_md, choices, exp_md, feedback, score_md, toast])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 511 |
|
| 512 |
# Spaces تتعرف على demo تلقائيًا
|
| 513 |
if __name__ == "__main__":
|