File size: 15,519 Bytes
fec9168 7e6c03a fec9168 | 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 | """
Question generation utilities for MCQ and open-text formats.
"""
import random
from typing import Dict, List, Optional, Tuple
from .logger import setup_logger
logger = setup_logger(__name__)
class QuestionGenerator:
"""Generates questions in MCQ and open-text formats."""
def __init__(
self,
num_options: int = 4,
option_labels: Optional[List[str]] = None,
distractor_strategy: str = "balanced"
):
"""
Initialize question generator.
Args:
num_options: Number of MCQ options
option_labels: Labels for options (e.g., ['A', 'B', 'C', 'D'])
distractor_strategy: Strategy for generating distractor options
- "present_only": only use sounds present in audio
- "mixed": mix of present and absent sounds
- "balanced": balanced distribution
"""
self.num_options = num_options
self.option_labels = option_labels or ["A", "B", "C", "D"]
self.distractor_strategy = distractor_strategy
if len(self.option_labels) != num_options:
raise ValueError(f"Number of option labels must match num_options ({num_options})")
def generate_count_mcq(
self,
question_template: str,
correct_count: int,
all_categories: List[str]
) -> Dict:
"""
Generate an MCQ for counting task.
Args:
question_template: Question text template
correct_count: Correct number of unique sounds
all_categories: List of all available categories
Returns:
Dictionary with question, options, and correct answer
"""
# Generate options (including the correct answer)
options = self._generate_count_options(correct_count)
# Shuffle options
random.shuffle(options)
# Find correct answer label
correct_label = self.option_labels[options.index(correct_count)]
# Create option mapping
option_map = {label: value for label, value in zip(self.option_labels, options)}
return {
"question": question_template,
"options": option_map,
"correct_answer": correct_label,
"correct_value": correct_count
}
def generate_count_open_text(
self,
question_template: str,
correct_count: int
) -> Dict:
"""
Generate an open-text question for counting task.
Args:
question_template: Question text template
correct_count: Correct number of unique sounds
Returns:
Dictionary with question and correct answer
"""
return {
"question": question_template,
"correct_answer": str(correct_count)
}
def generate_category_mcq(
self,
question_template: str,
correct_category: str,
present_categories: List[str],
all_categories: List[str]
) -> Dict:
"""
Generate an MCQ where answer is a sound category.
Args:
question_template: Question text template
correct_category: Correct category
present_categories: Categories present in the audio
all_categories: All available categories
Returns:
Dictionary with question, options, and correct answer
"""
# Generate distractor options
distractors = self._generate_category_distractors(
correct_category,
present_categories,
all_categories,
self.num_options - 1
)
# Combine with correct answer
options = [correct_category] + distractors
random.shuffle(options)
# Find correct answer label
correct_label = self.option_labels[options.index(correct_category)]
# Create option mapping
option_map = {label: value for label, value in zip(self.option_labels, options)}
return {
"question": question_template,
"options": option_map,
"correct_answer": correct_label,
"correct_value": correct_category
}
def generate_category_open_text(
self,
question_template: str,
correct_category: str
) -> Dict:
"""
Generate an open-text question where answer is a sound category.
Args:
question_template: Question text template
correct_category: Correct category
Returns:
Dictionary with question and correct answer
"""
return {
"question": question_template,
"correct_answer": correct_category
}
def generate_sequence_open_text(
self,
question_template: str,
sequence: List[str]
) -> Dict:
"""
Generate an open-text question for sequence/ordering.
Args:
question_template: Question text template
sequence: List of categories in order
Returns:
Dictionary with question and correct answer
"""
return {
"question": question_template,
"correct_answer": ", ".join(sequence)
}
def generate_pair_mcq(
self,
question_template: str,
correct_pair: Tuple[str, str],
all_present_categories: List[str],
all_categories: List[str]
) -> Dict:
"""
Generate an MCQ where the answer is a pair of sounds.
Used for silence_gap (longest/shortest gap between which two sounds).
Options are formatted as "soundA and soundB".
Args:
question_template: Question text
correct_pair: Tuple of (soundA, soundB) that form the correct pair
all_present_categories: Categories present in the audio
all_categories: All available categories
Returns:
Dictionary with question, options, correct_answer
"""
correct_option = f"{correct_pair[0]} and {correct_pair[1]}"
# Generate distractor pairs from present categories
import itertools
present_pairs = []
for a, b in itertools.combinations(all_present_categories, 2):
pair_str = f"{a} and {b}"
if pair_str != correct_option and f"{b} and {a}" != correct_option:
present_pairs.append(pair_str)
random.shuffle(present_pairs)
distractors = present_pairs[:self.num_options - 1]
# Fill remaining with mixed pairs if needed
while len(distractors) < self.num_options - 1:
absent = [c for c in all_categories if c not in all_present_categories]
if absent and all_present_categories:
pair_str = f"{random.choice(all_present_categories)} and {random.choice(absent)}"
if pair_str not in distractors and pair_str != correct_option:
distractors.append(pair_str)
else:
break
options = [correct_option] + distractors[:self.num_options - 1]
random.shuffle(options)
correct_label = self.option_labels[options.index(correct_option)]
option_map = {label: value for label, value in zip(self.option_labels, options)}
return {
"question": question_template,
"options": option_map,
"correct_answer": correct_label,
"correct_value": correct_option
}
def generate_pair_open_text(
self,
question_template: str,
correct_pair: Tuple[str, str]
) -> Dict:
"""
Generate an open-text question where answer is a pair of sounds.
Args:
question_template: Question text
correct_pair: Tuple of (soundA, soundB)
Returns:
Dictionary with question and correct_answer
"""
return {
"question": question_template,
"correct_answer": f"{correct_pair[0]} and {correct_pair[1]}"
}
def generate_yes_no_mcq(
self,
question_template: str,
correct_answer: bool
) -> Dict:
"""
Generate an MCQ with Yes/No answer.
Used for overlap (yes_no_overlap) and during_contains (yes_no_during).
Args:
question_template: Question text
correct_answer: True for Yes, False for No
Returns:
Dictionary with question, options, correct_answer
"""
options = ["Yes", "No"]
random.shuffle(options)
correct_text = "Yes" if correct_answer else "No"
# Use only A and B for yes/no
option_map = {self.option_labels[i]: options[i] for i in range(2)}
# Fill C and D with N/A
for i in range(2, self.num_options):
option_map[self.option_labels[i]] = "N/A"
correct_label = self.option_labels[options.index(correct_text)]
return {
"question": question_template,
"options": option_map,
"correct_answer": correct_label,
"correct_value": correct_text
}
def generate_yes_no_open_text(
self,
question_template: str,
correct_answer: bool
) -> Dict:
"""
Generate an open-text yes/no question.
Args:
question_template: Question text
correct_answer: True for Yes, False for No
Returns:
Dictionary with question and correct_answer
"""
return {
"question": question_template,
"correct_answer": "Yes" if correct_answer else "No"
}
def generate_pairwise_comparison_mcq(
self,
question_template: str,
correct_category: str,
compared_categories: Tuple[str, str],
all_categories: List[str]
) -> Dict:
"""
Generate an MCQ for pairwise comparison (duration/loudness).
The question names two specific sounds and asks which is longer/louder.
Options include both compared sounds plus distractors.
Args:
question_template: Question text (already formatted with {sound1}, {sound2})
correct_category: The correct answer category
compared_categories: Tuple of the two categories being compared
all_categories: All available categories for distractors
Returns:
Dictionary with question, options, correct_answer
"""
# Include both compared categories in options
options = list(compared_categories)
# Add distractors
other = [c for c in all_categories if c not in compared_categories]
random.shuffle(other)
options.extend(other[:self.num_options - 2])
# Trim to num_options
options = options[:self.num_options]
random.shuffle(options)
correct_label = self.option_labels[options.index(correct_category)]
option_map = {label: value for label, value in zip(self.option_labels, options)}
return {
"question": question_template,
"options": option_map,
"correct_answer": correct_label,
"correct_value": correct_category
}
def _generate_count_options(self, correct_count: int) -> List[int]:
"""
Generate count options including the correct count.
Args:
correct_count: Correct count value
Returns:
List of count options
"""
options = [correct_count]
# Generate distractors (minimum count is 1, not 0)
possible_values = list(range(1, max(correct_count + 3, 12)))
possible_values = [v for v in possible_values if v != correct_count]
distractors = random.sample(possible_values, min(self.num_options - 1, len(possible_values)))
options.extend(distractors)
return options[:self.num_options]
def _generate_category_distractors(
self,
correct_category: str,
present_categories: List[str],
all_categories: List[str],
num_distractors: int
) -> List[str]:
"""
Generate distractor categories based on strategy.
Args:
correct_category: Correct category
present_categories: Categories present in audio
all_categories: All available categories
num_distractors: Number of distractors to generate
Returns:
List of distractor categories
"""
present_non_answer = [c for c in present_categories if c != correct_category]
absent_categories = [c for c in all_categories if c not in present_categories]
distractors = []
if self.distractor_strategy == "present_only":
# Only use categories present in the audio
if len(present_non_answer) >= num_distractors:
distractors = random.sample(present_non_answer, num_distractors)
else:
distractors = present_non_answer.copy()
# Fill remaining with random absent categories
remaining = num_distractors - len(distractors)
distractors.extend(random.sample(absent_categories, min(remaining, len(absent_categories))))
elif self.distractor_strategy == "mixed":
# Mix of present and absent (random proportion)
num_present = random.randint(0, min(len(present_non_answer), num_distractors))
num_absent = num_distractors - num_present
if num_present > 0:
distractors.extend(random.sample(present_non_answer, min(num_present, len(present_non_answer))))
if num_absent > 0:
distractors.extend(random.sample(absent_categories, min(num_absent, len(absent_categories))))
else: # balanced
# Balanced distribution: 0, 1, or 2 present sounds as distractors
num_present_distractor = random.choice([0, 1, 2])
num_present_distractor = min(num_present_distractor, len(present_non_answer), num_distractors)
num_absent_distractor = num_distractors - num_present_distractor
if num_present_distractor > 0:
distractors.extend(random.sample(present_non_answer, num_present_distractor))
if num_absent_distractor > 0:
distractors.extend(random.sample(absent_categories, min(num_absent_distractor, len(absent_categories))))
# Fill remaining slots if needed
while len(distractors) < num_distractors:
remaining_options = [c for c in all_categories if c not in distractors and c != correct_category]
if not remaining_options:
break
distractors.append(random.choice(remaining_options))
return distractors[:num_distractors]
|