pluem9 commited on
Commit ·
3d94eac
1
Parent(s): 9b23490
fixminilmTFIDF
Browse files
src/streamlit_app.py
CHANGED
|
@@ -301,6 +301,7 @@ if generate_btn:
|
|
| 301 |
n_questions=n_questions,
|
| 302 |
level=level,
|
| 303 |
seed=seed_value,
|
|
|
|
| 304 |
)
|
| 305 |
|
| 306 |
rows = mcq_to_baseline_rows(
|
|
@@ -319,6 +320,7 @@ if generate_btn:
|
|
| 319 |
n_questions=n_questions,
|
| 320 |
level=level,
|
| 321 |
seed=seed_value,
|
|
|
|
| 322 |
)
|
| 323 |
|
| 324 |
rows = mcq_to_baseline_rows(
|
|
|
|
| 301 |
n_questions=n_questions,
|
| 302 |
level=level,
|
| 303 |
seed=seed_value,
|
| 304 |
+
q_types=q_types,
|
| 305 |
)
|
| 306 |
|
| 307 |
rows = mcq_to_baseline_rows(
|
|
|
|
| 320 |
n_questions=n_questions,
|
| 321 |
level=level,
|
| 322 |
seed=seed_value,
|
| 323 |
+
q_types=q_types,
|
| 324 |
)
|
| 325 |
|
| 326 |
rows = mcq_to_baseline_rows(
|
src/utils/__pycache__/utils_quiz_embedding.cpython-313.pyc
CHANGED
|
Binary files a/src/utils/__pycache__/utils_quiz_embedding.cpython-313.pyc and b/src/utils/__pycache__/utils_quiz_embedding.cpython-313.pyc differ
|
|
|
src/utils/__pycache__/utils_quiz_quality.cpython-313.pyc
CHANGED
|
Binary files a/src/utils/__pycache__/utils_quiz_quality.cpython-313.pyc and b/src/utils/__pycache__/utils_quiz_quality.cpython-313.pyc differ
|
|
|
src/utils/utils_quiz_embedding.py
CHANGED
|
@@ -132,10 +132,117 @@ def gen_question(slide: str, emb: EmbIndex, level="medium") -> Dict:
|
|
| 132 |
stem = random.choice(stem_templates)
|
| 133 |
return {"question": stem, "options": options, "answer": ans, "meta": {"level": level}}
|
| 134 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 135 |
# ---------------------------
|
| 136 |
# Public API
|
| 137 |
# ---------------------------
|
| 138 |
-
def generate_quiz(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 139 |
random.seed(seed)
|
| 140 |
np.random.seed(seed)
|
| 141 |
slides = [s for s in slides if s.strip()]
|
|
@@ -150,11 +257,14 @@ def generate_quiz(slides: List[str], n_questions=10, level="medium", seed=42) ->
|
|
| 150 |
all_q.append(q)
|
| 151 |
if len(all_q) > n_questions:
|
| 152 |
all_q = all_q[:n_questions]
|
|
|
|
|
|
|
|
|
|
| 153 |
return {
|
| 154 |
-
"questions":
|
| 155 |
"qc_report": {
|
| 156 |
"slides": len(slides),
|
| 157 |
-
"generated": len(
|
| 158 |
"level": clamp(level),
|
| 159 |
"seed": seed
|
| 160 |
}
|
|
|
|
| 132 |
stem = random.choice(stem_templates)
|
| 133 |
return {"question": stem, "options": options, "answer": ans, "meta": {"level": level}}
|
| 134 |
|
| 135 |
+
# ---------------------------
|
| 136 |
+
# Apply question types (MCQ / CLOZE / T/F)
|
| 137 |
+
# ---------------------------
|
| 138 |
+
def _apply_qtypes(
|
| 139 |
+
qq: List[Dict],
|
| 140 |
+
q_types: Optional[List[str]],
|
| 141 |
+
seed: int,
|
| 142 |
+
) -> List[Dict]:
|
| 143 |
+
if not qq:
|
| 144 |
+
return qq
|
| 145 |
+
|
| 146 |
+
if not q_types:
|
| 147 |
+
q_types = ["MCQ"]
|
| 148 |
+
|
| 149 |
+
norm_types = [t.upper() for t in q_types if t]
|
| 150 |
+
if not norm_types:
|
| 151 |
+
norm_types = ["MCQ"]
|
| 152 |
+
|
| 153 |
+
rng = random.Random(seed)
|
| 154 |
+
out: List[Dict] = []
|
| 155 |
+
|
| 156 |
+
for i, q in enumerate(qq):
|
| 157 |
+
stem = str(q.get("question", ""))
|
| 158 |
+
ans = str(q.get("answer", ""))
|
| 159 |
+
opts = [
|
| 160 |
+
str(o)
|
| 161 |
+
for o in q.get("options", [])
|
| 162 |
+
if isinstance(o, (str, int, float)) and str(o).strip()
|
| 163 |
+
]
|
| 164 |
+
|
| 165 |
+
if ans and all(norm(ans) != norm(o) for o in opts):
|
| 166 |
+
opts.insert(0, ans)
|
| 167 |
+
if not opts and ans:
|
| 168 |
+
opts = [ans]
|
| 169 |
+
|
| 170 |
+
while len(opts) < 4:
|
| 171 |
+
opts.append(f"option-{len(opts)+1}")
|
| 172 |
+
opts = dedup_keep_order(opts)[:4]
|
| 173 |
+
|
| 174 |
+
target_type = norm_types[i % len(norm_types)]
|
| 175 |
+
|
| 176 |
+
# CLOZE
|
| 177 |
+
if target_type == "CLOZE":
|
| 178 |
+
if ans:
|
| 179 |
+
pattern = re.compile(re.escape(ans), re.IGNORECASE)
|
| 180 |
+
new_stem, nsub = pattern.subn("____", stem, count=1)
|
| 181 |
+
if nsub > 0:
|
| 182 |
+
stem_cloze = new_stem
|
| 183 |
+
else:
|
| 184 |
+
stem_cloze = f"____ {stem}"
|
| 185 |
+
else:
|
| 186 |
+
stem_cloze = stem
|
| 187 |
+
|
| 188 |
+
out.append({
|
| 189 |
+
"type": "CLOZE",
|
| 190 |
+
"question": stem_cloze,
|
| 191 |
+
"options": opts,
|
| 192 |
+
"answer": ans,
|
| 193 |
+
"explanation": "",
|
| 194 |
+
"meta": q.get("meta", {}),
|
| 195 |
+
})
|
| 196 |
+
|
| 197 |
+
# T/F
|
| 198 |
+
elif target_type in ("T/F", "TF", "TRUE/FALSE"):
|
| 199 |
+
wrong = next(
|
| 200 |
+
(o for o in opts if ans and norm(o) != norm(ans)),
|
| 201 |
+
None,
|
| 202 |
+
)
|
| 203 |
+
make_true = (i % 2 == 0 or not wrong)
|
| 204 |
+
|
| 205 |
+
if make_true:
|
| 206 |
+
statement = f"\"{ans}\" ปรากฏเป็นคำสำคัญในสไลด์นี้"
|
| 207 |
+
tf_ans = "True"
|
| 208 |
+
expl = ""
|
| 209 |
+
else:
|
| 210 |
+
statement = f"\"{wrong}\" ปรากฏเป็นคำสำคัญในสไลด์นี้"
|
| 211 |
+
tf_ans = "False"
|
| 212 |
+
expl = f"คำตอบที่ถูกคือ {ans}"
|
| 213 |
+
|
| 214 |
+
out.append({
|
| 215 |
+
"type": "T/F",
|
| 216 |
+
"question": statement,
|
| 217 |
+
"options": ["True", "False"],
|
| 218 |
+
"answer": tf_ans,
|
| 219 |
+
"explanation": expl,
|
| 220 |
+
"meta": q.get("meta", {}),
|
| 221 |
+
})
|
| 222 |
+
|
| 223 |
+
# MCQ
|
| 224 |
+
else:
|
| 225 |
+
out.append({
|
| 226 |
+
"type": "MCQ",
|
| 227 |
+
"question": stem,
|
| 228 |
+
"options": opts,
|
| 229 |
+
"answer": ans,
|
| 230 |
+
"explanation": "",
|
| 231 |
+
"meta": q.get("meta", {}),
|
| 232 |
+
})
|
| 233 |
+
|
| 234 |
+
return out
|
| 235 |
+
|
| 236 |
# ---------------------------
|
| 237 |
# Public API
|
| 238 |
# ---------------------------
|
| 239 |
+
def generate_quiz(
|
| 240 |
+
slides: List[str],
|
| 241 |
+
n_questions=10,
|
| 242 |
+
level="medium",
|
| 243 |
+
seed=42,
|
| 244 |
+
q_types: Optional[List[str]] = None,
|
| 245 |
+
) -> Dict:
|
| 246 |
random.seed(seed)
|
| 247 |
np.random.seed(seed)
|
| 248 |
slides = [s for s in slides if s.strip()]
|
|
|
|
| 257 |
all_q.append(q)
|
| 258 |
if len(all_q) > n_questions:
|
| 259 |
all_q = all_q[:n_questions]
|
| 260 |
+
|
| 261 |
+
typed_questions = _apply_qtypes(all_q, q_types, seed)
|
| 262 |
+
|
| 263 |
return {
|
| 264 |
+
"questions": typed_questions,
|
| 265 |
"qc_report": {
|
| 266 |
"slides": len(slides),
|
| 267 |
+
"generated": len(typed_questions),
|
| 268 |
"level": clamp(level),
|
| 269 |
"seed": seed
|
| 270 |
}
|
src/utils/utils_quiz_quality.py
CHANGED
|
@@ -339,11 +339,122 @@ def distribute(buckets: List[List[Dict]], total: int) -> List[Dict]:
|
|
| 339 |
bs.pop(i % len(bs) if bs else 0)
|
| 340 |
return out
|
| 341 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 342 |
|
| 343 |
# ---------------------------
|
| 344 |
# Public API (force-complete)
|
| 345 |
# ---------------------------
|
| 346 |
-
def generate_quiz(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 347 |
set_all_seeds(seed)
|
| 348 |
slides = [s for s in slides if s and s.strip()]
|
| 349 |
if not slides:
|
|
@@ -397,15 +508,19 @@ def generate_quiz(slides: List[str], n_questions=12, level="medium", seed=42) ->
|
|
| 397 |
|
| 398 |
qc_set(final)
|
| 399 |
|
|
|
|
|
|
|
|
|
|
| 400 |
total_raw = sum(len(x) for x in draft_by_slide)
|
| 401 |
return {
|
| 402 |
-
"questions":
|
| 403 |
"qc_report": {
|
| 404 |
"slides": len(slides),
|
| 405 |
"draft_generated": total_raw,
|
| 406 |
-
"selected": len(
|
| 407 |
"level": clamp(level),
|
| 408 |
"seed": seed,
|
| 409 |
"forced_fill_used": len(final) < n_questions
|
| 410 |
}
|
| 411 |
}
|
|
|
|
|
|
| 339 |
bs.pop(i % len(bs) if bs else 0)
|
| 340 |
return out
|
| 341 |
|
| 342 |
+
# ---------------------------
|
| 343 |
+
# Apply question types (MCQ / CLOZE / T/F)
|
| 344 |
+
# ---------------------------
|
| 345 |
+
def _apply_qtypes(
|
| 346 |
+
qq: List[Dict],
|
| 347 |
+
q_types: Optional[List[str]],
|
| 348 |
+
seed: int,
|
| 349 |
+
) -> List[Dict]:
|
| 350 |
+
if not qq:
|
| 351 |
+
return qq
|
| 352 |
+
|
| 353 |
+
if not q_types:
|
| 354 |
+
q_types = ["MCQ"]
|
| 355 |
+
|
| 356 |
+
# normalize
|
| 357 |
+
norm_types = [t.upper() for t in q_types if t]
|
| 358 |
+
if not norm_types:
|
| 359 |
+
norm_types = ["MCQ"]
|
| 360 |
+
|
| 361 |
+
rng = random.Random(seed)
|
| 362 |
+
out: List[Dict] = []
|
| 363 |
+
|
| 364 |
+
for i, q in enumerate(qq):
|
| 365 |
+
stem = str(q.get("question", ""))
|
| 366 |
+
ans = str(q.get("answer", ""))
|
| 367 |
+
opts = [
|
| 368 |
+
str(o)
|
| 369 |
+
for o in q.get("options", [])
|
| 370 |
+
if isinstance(o, (str, int, float)) and str(o).strip()
|
| 371 |
+
]
|
| 372 |
+
|
| 373 |
+
# ensure answer in options
|
| 374 |
+
if ans and all(norm(ans) != norm(o) for o in opts):
|
| 375 |
+
opts.insert(0, ans)
|
| 376 |
+
if not opts and ans:
|
| 377 |
+
opts = [ans]
|
| 378 |
+
|
| 379 |
+
while len(opts) < 4:
|
| 380 |
+
opts.append(f"option-{len(opts)+1}")
|
| 381 |
+
opts = dedup_keep_order(opts)[:4]
|
| 382 |
+
|
| 383 |
+
# เลือก type
|
| 384 |
+
target_type = norm_types[i % len(norm_types)]
|
| 385 |
+
|
| 386 |
+
# ---- CLOZE ----
|
| 387 |
+
if target_type == "CLOZE":
|
| 388 |
+
if ans:
|
| 389 |
+
pattern = re.compile(re.escape(ans), re.IGNORECASE)
|
| 390 |
+
new_stem, nsub = pattern.subn("____", stem, count=1)
|
| 391 |
+
if nsub > 0:
|
| 392 |
+
stem_cloze = new_stem
|
| 393 |
+
else:
|
| 394 |
+
stem_cloze = f"____ {stem}"
|
| 395 |
+
else:
|
| 396 |
+
stem_cloze = stem
|
| 397 |
+
|
| 398 |
+
out.append({
|
| 399 |
+
"type": "CLOZE",
|
| 400 |
+
"question": stem_cloze,
|
| 401 |
+
"options": opts,
|
| 402 |
+
"answer": ans,
|
| 403 |
+
"explanation": "",
|
| 404 |
+
"meta": q.get("meta", {}),
|
| 405 |
+
})
|
| 406 |
+
|
| 407 |
+
# ---- T/F ----
|
| 408 |
+
elif target_type in ("T/F", "TF", "TRUE/FALSE"):
|
| 409 |
+
# ทำ statement จริง/เท็จจาก answer / distractor
|
| 410 |
+
wrong = next(
|
| 411 |
+
(o for o in opts if ans and norm(o) != norm(ans)),
|
| 412 |
+
None,
|
| 413 |
+
)
|
| 414 |
+
make_true = (i % 2 == 0 or not wrong)
|
| 415 |
+
|
| 416 |
+
if make_true:
|
| 417 |
+
statement = f"\"{ans}\" ปรากฏเป็นคำสำคัญในสไลด์นี้"
|
| 418 |
+
tf_ans = "True"
|
| 419 |
+
expl = ""
|
| 420 |
+
else:
|
| 421 |
+
statement = f"\"{wrong}\" ปรากฏเป็นคำสำคัญในสไลด์นี้"
|
| 422 |
+
tf_ans = "False"
|
| 423 |
+
expl = f"คำตอบที่ถูกคือ {ans}"
|
| 424 |
+
|
| 425 |
+
out.append({
|
| 426 |
+
"type": "T/F",
|
| 427 |
+
"question": statement,
|
| 428 |
+
"options": ["True", "False"],
|
| 429 |
+
"answer": tf_ans,
|
| 430 |
+
"explanation": expl,
|
| 431 |
+
"meta": q.get("meta", {}),
|
| 432 |
+
})
|
| 433 |
+
|
| 434 |
+
# ---- MCQ ----
|
| 435 |
+
else:
|
| 436 |
+
out.append({
|
| 437 |
+
"type": "MCQ",
|
| 438 |
+
"question": stem,
|
| 439 |
+
"options": opts,
|
| 440 |
+
"answer": ans,
|
| 441 |
+
"explanation": "",
|
| 442 |
+
"meta": q.get("meta", {}),
|
| 443 |
+
})
|
| 444 |
+
|
| 445 |
+
return out
|
| 446 |
+
|
| 447 |
|
| 448 |
# ---------------------------
|
| 449 |
# Public API (force-complete)
|
| 450 |
# ---------------------------
|
| 451 |
+
def generate_quiz(
|
| 452 |
+
slides: List[str],
|
| 453 |
+
n_questions=12,
|
| 454 |
+
level="medium",
|
| 455 |
+
seed=42,
|
| 456 |
+
q_types: Optional[List[str]] = None,
|
| 457 |
+
) -> Dict:
|
| 458 |
set_all_seeds(seed)
|
| 459 |
slides = [s for s in slides if s and s.strip()]
|
| 460 |
if not slides:
|
|
|
|
| 508 |
|
| 509 |
qc_set(final)
|
| 510 |
|
| 511 |
+
# apply question types (MCQ / CLOZE / T/F)
|
| 512 |
+
typed_questions = _apply_qtypes(final[:n_questions], q_types, seed)
|
| 513 |
+
|
| 514 |
total_raw = sum(len(x) for x in draft_by_slide)
|
| 515 |
return {
|
| 516 |
+
"questions": typed_questions,
|
| 517 |
"qc_report": {
|
| 518 |
"slides": len(slides),
|
| 519 |
"draft_generated": total_raw,
|
| 520 |
+
"selected": len(typed_questions),
|
| 521 |
"level": clamp(level),
|
| 522 |
"seed": seed,
|
| 523 |
"forced_fill_used": len(final) < n_questions
|
| 524 |
}
|
| 525 |
}
|
| 526 |
+
|