File size: 20,698 Bytes
edf1149 |
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 |
# DEPENDENCIES
import re
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 config.threshold_config import Domain
from metrics.base_metric import BaseMetric
from metrics.base_metric import MetricResult
from models.model_manager import get_model_manager
from sklearn.metrics.pairwise import cosine_similarity
from config.threshold_config import get_threshold_for_domain
class SemanticAnalysisMetric(BaseMetric):
"""
Semantic coherence and consistency analysis
Measures (Aligned with Documentation):
- Semantic similarity between sentences
- Topic consistency across text
- Coherence and logical flow
- Repetition patterns and redundancy
- Contextual consistency
"""
def __init__(self):
super().__init__(name = "semantic_analysis",
description = "Semantic coherence, repetition patterns, and contextual consistency analysis",
)
self.sentence_model = None
def initialize(self) -> bool:
"""
Initialize the semantic analysis metric
"""
try:
logger.info("Initializing semantic analysis metric...")
# Load sentence transformer for semantic embeddings
model_manager = get_model_manager()
self.sentence_model = model_manager.load_model("semantic_primary")
self.is_initialized = True
logger.success("Semantic analysis metric initialized successfully")
return True
except Exception as e:
logger.error(f"Failed to initialize semantic analysis metric: {repr(e)}")
return False
def compute(self, text: str, **kwargs) -> MetricResult:
"""
Compute semantic analysis measures 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 semantic analysis",
)
# Get domain-specific thresholds
domain = kwargs.get('domain', Domain.GENERAL)
domain_thresholds = get_threshold_for_domain(domain)
semantic_thresholds = domain_thresholds.semantic_analysis
# Calculate comprehensive semantic features
features = self._calculate_semantic_features(text)
# Calculate raw semantic score (0-1 scale)
raw_semantic_score, confidence = self._analyze_semantic_patterns(features)
# Apply domain-specific thresholds to convert raw score to probabilities
ai_prob, human_prob, mixed_prob = self._apply_domain_thresholds(raw_semantic_score, semantic_thresholds, features)
# Apply confidence multiplier from domain thresholds
confidence *= semantic_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' : semantic_thresholds.ai_threshold,
'human_threshold' : semantic_thresholds.human_threshold,
'raw_score' : raw_semantic_score,
},
)
except Exception as e:
logger.error(f"Error in semantic analysis 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., 0.65 for GENERAL, 0.70 for ACADEMIC
human_threshold = thresholds.human_threshold # e.g., 0.35 for GENERAL, 0.30 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 semantic 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_semantic_features(self, text: str) -> Dict[str, Any]:
"""
Calculate comprehensive semantic analysis features
"""
# Split text into sentences
sentences = self._split_sentences(text)
if (len(sentences) < 3):
return self._get_default_features()
# Calculate semantic embeddings for all sentences
sentence_embeddings = self._get_sentence_embeddings(sentences)
if sentence_embeddings is None:
return self._get_default_features()
# Calculate semantic similarity matrix
similarity_matrix = cosine_similarity(sentence_embeddings)
# Calculate various semantic metrics
coherence_score = self._calculate_coherence(similarity_matrix)
consistency_score = self._calculate_consistency(similarity_matrix)
repetition_score = self._detect_repetition_patterns(sentences, similarity_matrix)
topic_drift_score = self._calculate_topic_drift(similarity_matrix)
contextual_consistency = self._calculate_contextual_consistency(sentences)
# Chunk-based analysis for whole-text understanding
chunk_coherence = self._calculate_chunk_coherence(text, chunk_size=200)
return {"coherence_score" : round(coherence_score, 4),
"consistency_score" : round(consistency_score, 4),
"repetition_score" : round(repetition_score, 4),
"topic_drift_score" : round(topic_drift_score, 4),
"contextual_consistency" : round(contextual_consistency, 4),
"avg_chunk_coherence" : round(np.mean(chunk_coherence) if chunk_coherence else 0.0, 4),
"coherence_variance" : round(np.var(chunk_coherence) if chunk_coherence else 0.0, 4),
"num_sentences" : len(sentences),
"num_chunks_analyzed" : len(chunk_coherence),
}
def _split_sentences(self, text: str) -> List[str]:
"""
Split text into sentences
"""
sentences = re.split(r'(?<!\w\.\w.)(?<![A-Z][a-z]\.)(?<=\.|\?|\!)\s', text)
return [s.strip() for s in sentences if s.strip() and len(s.strip()) > 10]
def _get_sentence_embeddings(self, sentences: List[str]) -> np.ndarray:
"""
Get semantic embeddings for sentences
"""
try:
if not self.sentence_model:
return None
# Filter out very short sentences that might cause issues
valid_sentences = [s for s in sentences if len(s.strip()) > 5]
if not valid_sentences:
return None
# Encode sentences to get embeddings
embeddings = self.sentence_model.encode(valid_sentences)
# Check if embeddings are valid
if ((embeddings is None) or (len(embeddings) == 0)):
return None
return embeddings
except Exception as e:
logger.warning(f"Sentence embedding failed: {repr(e)}")
return None
def _calculate_coherence(self, similarity_matrix: np.ndarray) -> float:
"""
Calculate overall text coherence : Higher coherence = more logically connected sentences
"""
if similarity_matrix.size == 0:
return 0.0
# Calculate average similarity between adjacent sentences
adjacent_similarities = list()
for i in range(len(similarity_matrix) - 1):
adjacent_similarities.append(similarity_matrix[i, i + 1])
if (not adjacent_similarities):
return 0.0
return np.mean(adjacent_similarities)
def _calculate_consistency(self, similarity_matrix: np.ndarray) -> float:
"""
Calculate topic consistency throughout the text : Lower variance in similarities = more consistent
"""
if (similarity_matrix.size == 0):
return 0.0
# Calculate variance of similarities (lower variance = more consistent)
all_similarities = similarity_matrix[np.triu_indices_from(similarity_matrix, k=1)]
if (len(all_similarities) == 0):
return 0.0
variance = np.var(all_similarities)
# Convert to consistency score (higher = more consistent)
consistency = 1.0 - min(1.0, variance * 5.0) # Normalize
return max(0.0, consistency)
def _detect_repetition_patterns(self, sentences: List[str], similarity_matrix: np.ndarray) -> float:
"""
Detect repetition patterns in semantic content : AI text sometimes shows more semantic repetition
"""
if (len(sentences) < 5):
return 0.0
# Look for high similarity between non-adjacent sentences
repetition_count = 0
total_comparisons = 0
for i in range(len(sentences)):
for j in range(i + 2, len(sentences)): # Skip adjacent sentences
# High semantic similarity
if (similarity_matrix[i, j] > 0.8):
repetition_count += 1
total_comparisons += 1
if (total_comparisons == 0):
return 0.0
repetition_score = repetition_count / total_comparisons
# Scale to make differences more noticeable
return min(1.0, repetition_score * 3.0)
def _calculate_topic_drift(self, similarity_matrix: np.ndarray) -> float:
"""
Calculate topic drift throughout the text : Higher drift = less focused content
"""
if (len(similarity_matrix) < 3):
return 0.0
# Calculate similarity between beginning and end sections
start_size = min(3, len(similarity_matrix) // 3)
end_size = min(3, len(similarity_matrix) // 3)
start_indices = list(range(start_size))
end_indices = list(range(len(similarity_matrix) - end_size, len(similarity_matrix)))
cross_similarities = list()
for i in start_indices:
for j in end_indices:
cross_similarities.append(similarity_matrix[i, j])
if not cross_similarities:
return 0.0
avg_cross_similarity = np.mean(cross_similarities)
# Lower similarity between start and end = higher topic drift
topic_drift = 1.0 - avg_cross_similarity
return max(0.0, topic_drift)
def _calculate_contextual_consistency(self, sentences: List[str]) -> float:
"""
Calculate contextual consistency using keyword and entity analysis
"""
if (len(sentences) < 3):
return 0.0
# Simple keyword consistency analysis : Extract meaningful words (nouns, adjectives)
all_words = list()
for sentence in sentences:
words = re.findall(r'\b[a-zA-Z]{4,}\b', sentence.lower())
all_words.extend(words)
if (len(all_words) < 10):
return 0.0
# Calculate how consistently keywords are used across sentences
word_freq = Counter(all_words)
top_keywords = [word for word, count in word_freq.most_common(10) if count > 1]
if not top_keywords:
return 0.0
# Check if top keywords appear consistently across sentences
keyword_presence = list()
for keyword in top_keywords:
sentences_with_keyword = sum(1 for sentence in sentences if keyword in sentence.lower())
presence_ratio = sentences_with_keyword / len(sentences)
keyword_presence.append(presence_ratio)
consistency = np.mean(keyword_presence)
return consistency
def _calculate_chunk_coherence(self, text: str, chunk_size: int = 200) -> List[float]:
"""
Calculate coherence across text chunks for whole-text analysis
"""
chunks = list()
words = text.split()
# Create overlapping chunks
for i in range(0, len(words), chunk_size // 2):
chunk = ' '.join(words[i:i + chunk_size])
# Minimum chunk size
if (len(chunk) > 50):
chunk_sentences = self._split_sentences(chunk)
if (len(chunk_sentences) >= 2):
embeddings = self._get_sentence_embeddings(chunk_sentences)
if ((embeddings is not None) and (len(embeddings) >= 2)):
similarity_matrix = cosine_similarity(embeddings)
coherence = self._calculate_coherence(similarity_matrix)
chunks.append(coherence)
return chunks if chunks else [0.0]
def _analyze_semantic_patterns(self, features: Dict[str, Any]) -> tuple:
"""
Analyze semantic patterns to determine RAW semantic score (0-1 scale)
"""
# Check feature validity first
required_features = ['coherence_score', 'consistency_score', 'repetition_score', 'topic_drift_score', 'coherence_variance']
valid_features = [features.get(feat, 0) for feat in required_features if features.get(feat, 0) > 0]
if (len(valid_features) < 3):
# Low confidence if insufficient features
return 0.5, 0.3
# Initialize ai_indicator list
ai_indicators = list()
# AI text often has very high coherence (too perfect)
if (features['coherence_score'] > 0.7):
# Suspiciously high coherence
ai_indicators.append(0.8)
elif (features['coherence_score'] > 0.5):
# Moderate coherence
ai_indicators.append(0.5)
else:
# Low coherence - more human-like
ai_indicators.append(0.2)
# Very high consistency suggests AI (unnaturally consistent)
if (features['consistency_score'] > 0.8):
ai_indicators.append(0.9)
elif (features['consistency_score'] > 0.6):
ai_indicators.append(0.6)
else:
ai_indicators.append(0.3)
# High repetition suggests AI
if (features['repetition_score'] > 0.3):
ai_indicators.append(0.7)
elif (features['repetition_score'] > 0.1):
ai_indicators.append(0.4)
else:
ai_indicators.append(0.2)
# Very low topic drift suggests AI (stays too focused)
if (features['topic_drift_score'] < 0.2):
ai_indicators.append(0.8)
elif (features['topic_drift_score'] < 0.4):
ai_indicators.append(0.5)
else:
ai_indicators.append(0.3)
# Low coherence variance across chunks suggests AI
if (features['coherence_variance'] < 0.05):
ai_indicators.append(0.7)
elif (features['coherence_variance'] < 0.1):
ai_indicators.append(0.4)
else:
ai_indicators.append(0.2)
# 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
"""
mixed_indicators = list()
# Moderate coherence values might indicate mixing
if (0.4 <= features['coherence_score'] <= 0.6):
mixed_indicators.append(0.3)
else:
mixed_indicators.append(0.0)
# High coherence variance suggests mixed content
if (features['coherence_variance'] > 0.15):
mixed_indicators.append(0.4)
elif (features['coherence_variance'] > 0.1):
mixed_indicators.append(0.2)
else:
mixed_indicators.append(0.0)
# Inconsistent repetition patterns
if (0.15 <= features['repetition_score'] <= 0.35):
mixed_indicators.append(0.3)
else:
mixed_indicators.append(0.0)
return min(0.3, np.mean(mixed_indicators)) if mixed_indicators else 0.0
def _get_default_features(self) -> Dict[str, Any]:
"""
Return default features when analysis is not possible
"""
return {"coherence_score" : 0.5,
"consistency_score" : 0.5,
"repetition_score" : 0.0,
"topic_drift_score" : 0.5,
"contextual_consistency" : 0.5,
"avg_chunk_coherence" : 0.5,
"coherence_variance" : 0.1,
"num_sentences" : 0,
"num_chunks_analyzed" : 0,
}
def cleanup(self):
"""
Clean up resources
"""
self.sentence_model = None
super().cleanup()
# Export
__all__ = ["SemanticAnalysisMetric"]
|