Spaces:
Sleeping
Sleeping
File size: 20,316 Bytes
8205d6b |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 |
# DEPENDENCIES
import math
import numpy as np
from typing import Any
from typing import Dict
from typing import List
from loguru import logger
from collections import Counter
from metrics.base_metric import BaseMetric
from config.threshold_config import Domain
from metrics.base_metric import MetricResult
from models.model_manager import get_model_manager
from config.threshold_config import get_threshold_for_domain
class EntropyMetric(BaseMetric):
"""
Enhanced entropy analysis for text randomness and predictability
Measures (Aligned with Documentation):
- Character-level entropy and diversity
- Word-level entropy and burstiness
- Token-level diversity and unpredictability in sequences
- Entropy distribution across text chunks
- AI-specific pattern detection
"""
def __init__(self):
super().__init__(name = "entropy",
description = "Token-level diversity and unpredictability in text sequences",
)
self.tokenizer = None
def initialize(self) -> bool:
"""
Initialize the entropy metric
"""
try:
logger.info("Initializing entropy metric...")
# Load tokenizer for token-level analysis
model_manager = get_model_manager()
gpt_model = model_manager.load_model("perplexity_gpt2")
if isinstance(gpt_model, tuple):
self.tokenizer = gpt_model[1]
else:
logger.warning("Could not get tokenizer, using character-level entropy only")
self.is_initialized = True
logger.success("Entropy metric initialized successfully")
return True
except Exception as e:
logger.error(f"Failed to initialize entropy metric: {repr(e)}")
return False
def compute(self, text: str, **kwargs) -> MetricResult:
"""
Compute enhanced entropy measures for text with FULL DOMAIN THRESHOLD INTEGRATION
"""
try:
if (not text or (len(text.strip()) < 50)):
return MetricResult(metric_name = self.name,
ai_probability = 0.5,
human_probability = 0.5,
mixed_probability = 0.0,
confidence = 0.1,
error = "Text too short for entropy analysis",
)
# Get domain-specific thresholds
domain = kwargs.get('domain', Domain.GENERAL)
domain_thresholds = get_threshold_for_domain(domain)
entropy_thresholds = domain_thresholds.entropy
# Calculate comprehensive entropy features
features = self._calculate_enhanced_entropy_features(text)
# Calculate raw entropy score (0-1 scale)
raw_entropy_score, confidence = self._analyze_entropy_patterns(features)
# Apply domain-specific thresholds to convert raw score to probabilities
ai_prob, human_prob, mixed_prob = self._apply_domain_thresholds(raw_entropy_score, entropy_thresholds, features)
# Apply confidence multiplier from domain thresholds
confidence *= entropy_thresholds.confidence_multiplier
confidence = max(0.0, min(1.0, confidence))
return MetricResult(metric_name = self.name,
ai_probability = ai_prob,
human_probability = human_prob,
mixed_probability = mixed_prob,
confidence = confidence,
details = {**features,
'domain_used' : domain.value,
'ai_threshold' : entropy_thresholds.ai_threshold,
'human_threshold' : entropy_thresholds.human_threshold,
'raw_score' : raw_entropy_score,
},
)
except Exception as e:
logger.error(f"Error in entropy computation: {repr(e)}")
return MetricResult(metric_name = self.name,
ai_probability = 0.5,
human_probability = 0.5,
mixed_probability = 0.0,
confidence = 0.0,
error = str(e),
)
def _apply_domain_thresholds(self, raw_score: float, thresholds: Any, features: Dict[str, Any]) -> tuple:
"""
Apply domain-specific thresholds to convert raw score to probabilities
"""
ai_threshold = thresholds.ai_threshold # e.g., 0.55 for GENERAL, 0.50 for ACADEMIC
human_threshold = thresholds.human_threshold # e.g., 0.45 for GENERAL, 0.40 for ACADEMIC
# Calculate probabilities based on threshold distances
if (raw_score >= ai_threshold):
# Above AI threshold - strongly AI
distance_from_threshold = raw_score - ai_threshold
ai_prob = 0.7 + (distance_from_threshold * 0.3) # 0.7 to 1.0
human_prob = 0.3 - (distance_from_threshold * 0.3) # 0.3 to 0.0
elif (raw_score <= human_threshold):
# Below human threshold - strongly human
distance_from_threshold = human_threshold - raw_score
ai_prob = 0.3 - (distance_from_threshold * 0.3) # 0.3 to 0.0
human_prob = 0.7 + (distance_from_threshold * 0.3) # 0.7 to 1.0
else:
# Between thresholds - uncertain zone
range_width = ai_threshold - human_threshold
if (range_width > 0):
position_in_range = (raw_score - human_threshold) / range_width
ai_prob = 0.3 + (position_in_range * 0.4) # 0.3 to 0.7
human_prob = 0.7 - (position_in_range * 0.4) # 0.7 to 0.3
else:
ai_prob = 0.5
human_prob = 0.5
# Ensure probabilities are valid
ai_prob = max(0.0, min(1.0, ai_prob))
human_prob = max(0.0, min(1.0, human_prob))
# Calculate mixed probability based on entropy variance
mixed_prob = self._calculate_mixed_probability(features)
# Normalize to sum to 1.0
total = ai_prob + human_prob + mixed_prob
if (total > 0):
ai_prob /= total
human_prob /= total
mixed_prob /= total
return ai_prob, human_prob, mixed_prob
def _calculate_enhanced_entropy_features(self, text: str) -> Dict[str, Any]:
"""
Calculate comprehensive entropy measures including document-required features
"""
# Basic entropy measures
char_entropy = self._calculate_character_entropy(text)
word_entropy = self._calculate_word_entropy(text)
token_entropy = self._calculate_token_entropy(text) if self.tokenizer else 0.0
# DOCUMENT-REQUIRED: Token-level diversity
token_diversity = self._calculate_token_diversity(text)
# DOCUMENT-REQUIRED: Unpredictability in sequences
sequence_unpredictability = self._calculate_sequence_unpredictability(text)
# Chunk-based analysis for whole-text understanding
chunk_entropies = self._calculate_chunk_entropy(text, chunk_size=100)
entropy_variance = np.var(chunk_entropies) if chunk_entropies else 0.0
avg_chunk_entropy = np.mean(chunk_entropies) if chunk_entropies else 0.0
# AI-specific pattern detection
ai_pattern_score = self._detect_ai_entropy_patterns(text)
# Predictability measures
predictability = 1.0 - min(1.0, char_entropy / 4.0)
return {"char_entropy" : round(char_entropy, 4),
"word_entropy" : round(word_entropy, 4),
"token_entropy" : round(token_entropy, 4),
"token_diversity" : round(token_diversity, 4),
"sequence_unpredictability" : round(sequence_unpredictability, 4),
"entropy_variance" : round(entropy_variance, 4),
"avg_chunk_entropy" : round(avg_chunk_entropy, 4),
"predictability_score" : round(predictability, 4),
"ai_pattern_score" : round(ai_pattern_score, 4),
"num_chunks_analyzed" : len(chunk_entropies),
}
def _calculate_character_entropy(self, text: str) -> float:
"""
Calculate character-level entropy
"""
# Clean text and convert to lowercase
clean_text = ''.join(c for c in text.lower() if c.isalnum() or c.isspace())
if not clean_text:
return 0.0
# Count character frequencies
char_counts = Counter(clean_text)
total_chars = len(clean_text)
# Calculate entropy
entropy = 0.0
for count in char_counts.values():
probability = count / total_chars
entropy -= probability * math.log2(probability)
return entropy
def _calculate_word_entropy(self, text: str) -> float:
"""
Calculate word-level entropy
"""
words = text.lower().split()
if (len(words) < 5):
return 0.0
word_counts = Counter(words)
total_words = len(words)
entropy = 0.0
for count in word_counts.values():
probability = count / total_words
entropy -= probability * math.log2(probability)
return entropy
def _calculate_token_entropy(self, text: str) -> float:
"""
Calculate token-level entropy using GPT-2 tokenizer
"""
try:
if not self.tokenizer:
return 0.0
# Length check before tokenization
if (len(text.strip()) < 10):
return 0.0
# Tokenize text
tokens = self.tokenizer.encode(text,
add_special_tokens = False,
truncation = True,
)
if (len(tokens) < 10):
return 0.0
token_counts = Counter(tokens)
total_tokens = len(tokens)
entropy = 0.0
for count in token_counts.values():
probability = count / total_tokens
entropy -= probability * math.log2(probability)
return entropy
except Exception as e:
logger.warning(f"Token entropy calculation failed: {repr(e)}")
return 0.0
def _calculate_token_diversity(self, text: str) -> float:
"""
Calculate token-level diversity : Higher diversity = more human-like
"""
if not self.tokenizer:
return 0.0
try:
tokens = self.tokenizer.encode(text, add_special_tokens=False)
if (len(tokens) < 10):
return 0.0
unique_tokens = len(set(tokens))
total_tokens = len(tokens)
# Type-token ratio for tokens
diversity = unique_tokens / total_tokens
return diversity
except Exception as e:
logger.warning(f"Token diversity calculation failed: {repr(e)}")
return 0.0
def _calculate_sequence_unpredictability(self, text: str) -> float:
"""
Calculate unpredictability in text sequences, it measures how unpredictable the token sequences are
"""
if not self.tokenizer:
return 0.0
try:
tokens = self.tokenizer.encode(text, add_special_tokens=False)
if (len(tokens) < 20):
return 0.0
# Calculate bigram unpredictability
bigrams = [(tokens[i], tokens[i+1]) for i in range(len(tokens)-1)]
bigram_counts = Counter(bigrams)
total_bigrams = len(bigrams)
# Higher entropy = more unpredictable sequences
sequence_entropy = 0.0
for count in bigram_counts.values():
probability = count / total_bigrams
sequence_entropy -= probability * math.log2(probability)
# Normalize to 0-1 scale : Assuming max ~8 bits
normalized_entropy = min(1.0, sequence_entropy / 8.0)
return normalized_entropy
except Exception as e:
logger.warning(f"Sequence unpredictability calculation failed: {repr(e)}")
return 0.0
def _calculate_chunk_entropy(self, text: str, chunk_size: int = 100) -> List[float]:
"""
Calculate entropy distribution across text chunks
"""
chunks = list()
words = text.split()
# Create overlapping chunks for better analysis
for i in range(0, len(words), chunk_size // 2):
chunk = ' '.join(words[i:i + chunk_size])
# Minimum chunk size
if (len(chunk) > 20):
entropy = self._calculate_character_entropy(chunk)
chunks.append(entropy)
return chunks
def _detect_ai_entropy_patterns(self, text: str) -> float:
"""
Detect AI-specific entropy patterns: AI text often shows specific entropy signatures
"""
patterns_detected = 0
total_patterns = 4
# Overly consistent character distribution
char_entropy = self._calculate_character_entropy(text)
# AI tends to be more consistent
if (char_entropy < 3.8):
patterns_detected += 1
# Low token diversity
token_diversity = self._calculate_token_diversity(text)
# AI reuses tokens more
if (token_diversity < 0.7):
patterns_detected += 1
# Predictable sequences
sequence_unpredictability = self._calculate_sequence_unpredictability(text)
# AI sequences are more predictable
if (sequence_unpredictability < 0.4):
patterns_detected += 1
# Low entropy variance across chunks
chunk_entropies = self._calculate_chunk_entropy(text, chunk_size = 100)
entropy_variance = np.var(chunk_entropies) if chunk_entropies else 0.0
# AI maintains consistent entropy
if (entropy_variance < 0.2):
patterns_detected += 1
return patterns_detected / total_patterns
def _analyze_entropy_patterns(self, features: Dict[str, Any]) -> tuple:
"""
Analyze entropy patterns to determine RAW entropy score (0-1 scale)
This raw score will later be converted using domain thresholds
"""
# Check feature validity
valid_features = [score for score in [features.get('char_entropy', 0),
features.get('token_diversity', 0),
features.get('sequence_unpredictability', 0),
features.get('ai_pattern_score', 0)
] if score > 0
]
if (len(valid_features) < 2):
# Low confidence if insufficient features
return 0.5, 0.3
ai_indicators = list()
# AI text often has lower character entropy (more predictable)
if (features['char_entropy'] < 3.5):
# Strong AI indicator
ai_indicators.append(0.8)
elif (features['char_entropy'] < 4.0):
# Moderate AI indicator
ai_indicators.append(0.6)
else:
# Weak AI indicator
ai_indicators.append(0.2)
# Low entropy variance suggests AI (consistent patterns)
if (features['entropy_variance'] < 0.1):
# Very strong AI indicator
ai_indicators.append(0.9)
elif (features['entropy_variance'] < 0.3):
# Neutral
ai_indicators.append(0.5)
else:
# Strong human indicator
ai_indicators.append(0.1)
# Low token diversity suggests AI
if (features['token_diversity'] < 0.6):
ai_indicators.append(0.7)
elif (features['token_diversity'] < 0.8):
ai_indicators.append(0.4)
else:
ai_indicators.append(0.2)
# Low sequence unpredictability suggests AI
if (features['sequence_unpredictability'] < 0.3):
ai_indicators.append(0.8)
elif (features['sequence_unpredictability'] < 0.5):
ai_indicators.append(0.5)
else:
ai_indicators.append(0.2)
# High AI pattern score suggests AI
if (features['ai_pattern_score'] > 0.75):
ai_indicators.append(0.9)
elif (features['ai_pattern_score'] > 0.5):
ai_indicators.append(0.7)
else:
ai_indicators.append(0.3)
# Calculate raw score and confidence
raw_score = np.mean(ai_indicators) if ai_indicators else 0.5
confidence = 1.0 - (np.std(ai_indicators) / 0.5) if ai_indicators else 0.5
confidence = max(0.1, min(0.9, confidence))
return raw_score, confidence
def _calculate_mixed_probability(self, features: Dict[str, Any]) -> float:
"""
Calculate probability of mixed AI/Human content with better indicators
"""
mixed_indicators = list()
# High entropy variance suggests mixed content
entropy_variance = features.get('entropy_variance', 0)
if (entropy_variance > 0.5):
# Strong mixed indicator
mixed_indicators.append(0.6)
elif (entropy_variance > 0.3):
mixed_indicators.append(0.3)
else:
mixed_indicators.append(0.0)
# Inconsistent patterns across different entropy measures
char_entropy = features.get('char_entropy', 0)
word_entropy = features.get('word_entropy', 0)
if ((char_entropy > 0) and (word_entropy > 0)):
entropy_discrepancy = abs(char_entropy - word_entropy)
# Large discrepancy suggests mixing
if (entropy_discrepancy > 1.0):
mixed_indicators.append(0.4)
# Moderate AI pattern score might indicate mixing
ai_pattern_score = features.get('ai_pattern_score', 0)
if (0.4 <= ai_pattern_score <= 0.6):
mixed_indicators.append(0.3)
mixed_probability = min(0.4, np.mean(mixed_indicators)) if mixed_indicators else 0.0
return mixed_probability
def cleanup(self):
"""
Clean up resources
"""
self.tokenizer = None
super().cleanup()
# Export
__all__ = ["EntropyMetric"]
|