Spaces:
Sleeping
Sleeping
File size: 55,696 Bytes
8205d6b 38d5865 8205d6b 38d5865 8205d6b c79f796 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b c79f796 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b c79f796 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b c79f796 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b 38d5865 8205d6b c79f796 8205d6b c79f796 8205d6b c79f796 8205d6b c79f796 8205d6b c79f796 8205d6b 38d5865 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 |
# DEPENDENCIES
import re
from typing import List
from typing import Dict
from typing import Tuple
from loguru import logger
from typing import Optional
from dataclasses import dataclass
from config.threshold_config import Domain
from metrics.base_metric import MetricResult
from detector.ensemble import EnsembleResult
from detector.ensemble import EnsembleClassifier
from processors.text_processor import TextProcessor
from config.threshold_config import ConfidenceLevel
from config.threshold_config import MetricThresholds
from config.threshold_config import get_confidence_level
from config.threshold_config import get_threshold_for_domain
from config.threshold_config import get_active_metric_weights
@dataclass
class HighlightedSentence:
"""
A sentence with highlighting information
"""
text : str
ai_probability : float
human_probability : float
mixed_probability : float
confidence : float
confidence_level : ConfidenceLevel
color_class : str
tooltip : str
index : int
is_mixed_content : bool
metric_breakdown : Optional[Dict[str, float]] = None
class TextHighlighter:
"""
Generates sentence-level highlighting with ensemble results integration
FEATURES:
- Sentence-level highlighting with confidence scores
- Domain-aware calibration
- Ensemble-based probability aggregation
- Mixed content detection
- Explainable tooltips
- Highlighting metrics calculation
"""
# Color thresholds with mixed content support - FIXED: No gaps
COLOR_THRESHOLDS = [(0.00, 0.10, "very-high-human", "#dcfce7", "Very likely human-written"),
(0.10, 0.25, "high-human", "#bbf7d0", "Likely human-written"),
(0.25, 0.40, "medium-human", "#86efac", "Possibly human-written"),
(0.40, 0.60, "uncertain", "#fef9c3", "Uncertain"),
(0.60, 0.75, "medium-ai", "#fde68a", "Possibly AI-generated"),
(0.75, 0.90, "high-ai", "#fed7aa", "Likely AI-generated"),
(0.90, 1.00, "very-high-ai", "#fecaca", "Very likely AI-generated"),
]
# Mixed content pattern
MIXED_THRESHOLD = 0.25
# Risk weights
RISK_WEIGHTS = {'very-high-ai' : 1.0,
'high-ai' : 0.8,
'medium-ai' : 0.6,
'uncertain' : 0.4,
'medium-human' : 0.2,
'high-human' : 0.1,
'very-high-human' : 0.0,
'mixed-content' : 0.7,
}
def __init__(self, domain: Domain = Domain.GENERAL, ensemble_classifier: Optional[EnsembleClassifier] = None):
"""
Initialize text highlighter with ENSEMBLE INTEGRATION
Arguments:
----------
domain { Domain } : Text domain for adaptive thresholding
ensemble_classifier { EnsembleClassifier } : Optional ensemble for sentence-level analysis
"""
self.text_processor = TextProcessor()
self.domain = domain
self.domain_thresholds = get_threshold_for_domain(domain)
self.ensemble = ensemble_classifier or self._create_default_ensemble()
def _create_default_ensemble(self) -> EnsembleClassifier:
"""
Create default ensemble classifier with proper error handling
"""
try:
return EnsembleClassifier(primary_method = "confidence_calibrated",
fallback_method = "domain_weighted",
)
except Exception as e:
logger.warning(f"Failed to create default ensemble: {e}. Using fallback mode.")
# Return a minimal ensemble or raise based on requirements
return EnsembleClassifier(primary_method = "weighted_average")
def generate_highlights(self, text: str, metric_results: Dict[str, MetricResult], ensemble_result: Optional[EnsembleResult] = None,
enabled_metrics: Optional[Dict[str, bool]] = None, use_sentence_level: bool = True) -> List[HighlightedSentence]:
"""
Generate sentence-level highlights with ensemble integration
Arguments:
----------
text { str } : Original text
metric_results { dict } : Results from all 6 metrics
ensemble_result { EnsembleResult } : Optional document-level ensemble result
enabled_metrics { dict } : Dict of metric_name -> is_enabled
use_sentence_level { bool } : Whether to compute sentence-level probabilities
Returns:
--------
{ list } : List of HighlightedSentence objects
"""
try:
# Validate inputs
if not text or not text.strip():
return self._handle_empty_text(text, metric_results, ensemble_result)
# Get domain-appropriate weights for enabled metrics
if enabled_metrics is None:
enabled_metrics = {name: True for name in metric_results.keys()}
weights = get_active_metric_weights(self.domain, enabled_metrics)
# Split text into sentences with error handling
sentences = self._split_sentences_with_fallback(text)
if not sentences:
return self._handle_no_sentences(text, metric_results, ensemble_result)
# Calculate probabilities for each sentence using ENSEMBLE METHODS
highlighted_sentences = list()
for idx, sentence in enumerate(sentences):
try:
if use_sentence_level:
# Use ENSEMBLE for sentence-level analysis
ai_prob, human_prob, mixed_prob, confidence, breakdown = self._calculate_sentence_ensemble_probability(sentence = sentence,
metric_results = metric_results,
weights = weights,
ensemble_result = ensemble_result,
)
else:
# Use document-level ensemble probabilities
ai_prob, human_prob, mixed_prob, confidence, breakdown = self._get_document_ensemble_probability(ensemble_result = ensemble_result,
metric_results = metric_results,
weights = weights,
)
# Apply domain-specific adjustments with limits
ai_prob = self._apply_domain_specific_adjustments(sentence = sentence,
ai_prob = ai_prob,
sentence_length = len(sentence.split()),
)
# Determine if this is mixed content
is_mixed_content = (mixed_prob > self.MIXED_THRESHOLD)
# Get confidence level
confidence_level = get_confidence_level(confidence)
# Get color class (consider mixed content)
color_class, color_hex, tooltip_base = self._get_color_for_probability(probability = ai_prob,
is_mixed_content = is_mixed_content,
mixed_prob = mixed_prob,
)
# Generate enhanced tooltip
tooltip = self._generate_ensemble_tooltip(sentence = sentence,
ai_prob = ai_prob,
human_prob = human_prob,
mixed_prob = mixed_prob,
confidence = confidence,
confidence_level = confidence_level,
tooltip_base = tooltip_base,
breakdown = breakdown,
is_mixed_content = is_mixed_content,
)
highlighted_sentences.append(HighlightedSentence(text = sentence,
ai_probability = ai_prob,
human_probability = human_prob,
mixed_probability = mixed_prob,
confidence = confidence,
confidence_level = confidence_level,
color_class = color_class,
tooltip = tooltip,
index = idx,
is_mixed_content = is_mixed_content,
metric_breakdown = breakdown,
)
)
except Exception as e:
logger.warning(f"Failed to process sentence {idx}: {e}")
# Add fallback sentence
highlighted_sentences.append(self._create_fallback_sentence(sentence, idx))
return highlighted_sentences
except Exception as e:
logger.error(f"Highlight generation failed: {e}")
return self._create_error_fallback(text, metric_results)
def _handle_empty_text(self, text: str, metric_results: Dict[str, MetricResult], ensemble_result: Optional[EnsembleResult]) -> List[HighlightedSentence]:
"""
Handle empty input text
"""
if ensemble_result:
return [self._create_fallback_sentence(text = "No text content",
index = 0,
ai_prob = ensemble_result.ai_probability,
human_prob = ensemble_result.human_probability,
)
]
return [self._create_fallback_sentence("No text content", 0)]
def _handle_no_sentences(self, text: str, metric_results: Dict[str, MetricResult], ensemble_result: Optional[EnsembleResult]) -> List[HighlightedSentence]:
"""
Handle case where no sentences could be extracted
"""
if (text and (len(text.strip()) > 0)):
# Treat entire text as one sentence
return [self._create_fallback_sentence(text.strip(), 0)]
return [self._create_fallback_sentence("No processable content", 0)]
def _create_fallback_sentence(self, text: str, index: int, ai_prob: float = 0.5, human_prob: float = 0.5) -> HighlightedSentence:
"""
Create a fallback sentence when processing fails
"""
confidence_level = get_confidence_level(0.3)
color_class, _, tooltip_base = self._get_color_for_probability(probability = ai_prob,
is_mixed_content = False,
mixed_prob = 0.0,
)
return HighlightedSentence(text = text,
ai_probability = ai_prob,
human_probability = human_prob,
mixed_probability = 0.0,
confidence = 0.3,
confidence_level = confidence_level,
color_class = color_class,
tooltip = f"Fallback: {tooltip_base}\nProcessing failed for this sentence",
index = index,
is_mixed_content = False,
metric_breakdown = {"fallback": ai_prob},
)
def _create_error_fallback(self, text: str, metric_results: Dict[str, MetricResult]) -> List[HighlightedSentence]:
"""
Create fallback when entire processing fails
"""
return [HighlightedSentence(text = text[:100] + "..." if len(text) > 100 else text,
ai_probability = 0.5,
human_probability = 0.5,
mixed_probability = 0.0,
confidence = 0.1,
confidence_level = get_confidence_level(0.1),
color_class = "uncertain",
tooltip = "Error in text processing",
index = 0,
is_mixed_content = False,
metric_breakdown = {"error": 0.5},
)
]
def _split_sentences_with_fallback(self, text: str) -> List[str]:
"""
Split text into sentences with comprehensive fallback handling
"""
try:
sentences = self.text_processor.split_sentences(text)
filtered_sentences = [s.strip() for s in sentences if len(s.strip()) >= 3]
if filtered_sentences:
return filtered_sentences
# Fallback: split by common sentence endings
fallback_sentences = re.split(r'[.!?]+', text)
fallback_sentences = [s.strip() for s in fallback_sentences if len(s.strip()) >= 3]
if fallback_sentences:
return fallback_sentences
# Ultimate fallback: treat as single sentence if meaningful
if text.strip():
return [text.strip()]
return []
except Exception as e:
logger.warning(f"Sentence splitting failed, using fallback: {e}")
# Return text as single sentence
return [text] if text.strip() else []
def _calculate_sentence_ensemble_probability(self, sentence: str, metric_results: Dict[str, MetricResult], weights: Dict[str, float],
ensemble_result: Optional[EnsembleResult] = None) -> Tuple[float, float, float, float, Dict[str, float]]:
"""
Calculate sentence probabilities using ensemble methods with domain calibration
"""
sentence_length = len(sentence.split())
# Handling short sentences - don't force neutral
if (sentence_length < 3):
# Return probabilities with lower confidence for very short sentences
base_ai_prob = 0.5
# Low confidence for very short sentences
base_confidence = 0.2
breakdown = {"short_sentence" : base_ai_prob}
# Try to get some signal from available metrics
for name, result in metric_results.items():
if ((result.error is None) and (weights.get(name, 0) > 0)):
base_ai_prob = result.ai_probability
breakdown[name] = base_ai_prob
break
return base_ai_prob, 1.0 - base_ai_prob, 0.0, base_confidence, breakdown
# Calculate sentence-level metric results
sentence_metric_results = dict()
breakdown = dict()
for name, doc_result in metric_results.items():
if doc_result.error is None:
try:
# Compute sentence-level probability for this metric
sentence_prob = self._compute_sentence_metric(metric_name = name,
sentence = sentence,
result = doc_result,
weight = weights.get(name, 0.0),
)
# Create sentence-level MetricResult
sentence_metric_results[name] = self._create_sentence_metric_result(metric_name = name,
ai_prob = sentence_prob,
doc_result = doc_result,
sentence_length = sentence_length,
)
breakdown[name] = sentence_prob
except Exception as e:
logger.warning(f"Metric {name} failed for sentence: {e}")
# Use document probability as fallback
breakdown[name] = doc_result.ai_probability
# Use ensemble to combine sentence-level metrics
if sentence_metric_results:
try:
ensemble_sentence_result = self.ensemble.predict(metric_results = sentence_metric_results,
domain = self.domain,
)
return (ensemble_sentence_result.ai_probability,
ensemble_sentence_result.human_probability,
ensemble_sentence_result.mixed_probability,
ensemble_sentence_result.overall_confidence,
breakdown)
except Exception as e:
logger.warning(f"Sentence ensemble failed: {e}")
# Fallback: weighted average
return self._calculate_weighted_probability(metric_results, weights, breakdown)
def _compute_sentence_metric(self, metric_name: str, sentence: str, result: MetricResult, weight: float) -> float:
"""
Compute metric probability for a single sentence using domain-specific thresholds
"""
sentence_length = len(sentence.split())
# Get domain-specific threshold for this metric
metric_thresholds = getattr(self.domain_thresholds, metric_name, None)
if not metric_thresholds:
return result.ai_probability
# Base probability from document-level result
base_prob = result.ai_probability
# Apply domain-aware sentence-level adjustments
adjusted_prob = self._apply_metric_specific_adjustments(metric_name = metric_name,
sentence = sentence,
base_prob = base_prob,
sentence_length = sentence_length,
thresholds = metric_thresholds,
)
return adjusted_prob
def _create_sentence_metric_result(self, metric_name: str, ai_prob: float, doc_result: MetricResult, sentence_length: int) -> MetricResult:
"""
Create sentence-level MetricResult from document-level result
"""
# IMPROVED: Calculate confidence based on sentence characteristics
sentence_confidence = self._calculate_sentence_confidence(doc_result.confidence, sentence_length)
return MetricResult(metric_name = metric_name,
ai_probability = ai_prob,
human_probability = 1.0 - ai_prob,
mixed_probability = 0.0,
confidence = sentence_confidence,
details = doc_result.details,
error = None,
)
def _calculate_sentence_confidence(self, doc_confidence: float, sentence_length: int) -> float:
"""
IMPROVED: Calculate confidence for sentence-level analysis with length consideration
"""
base_reduction = 0.8
# Scale confidence penalty with sentence length
length_penalty = max(0.3, min(1.0, sentence_length / 12.0)) # Normalize around 12 words
return max(0.1, doc_confidence * base_reduction * length_penalty)
def _calculate_weighted_probability(self, metric_results: Dict[str, MetricResult], weights: Dict[str, float], breakdown: Dict[str, float]) -> Tuple[float, float, float, float, Dict[str, float]]:
"""
Fallback weighted probability calculation
"""
weighted_ai_probs = list()
weighted_human_probs = list()
confidences = list()
total_weight = 0.0
for name, result in metric_results.items():
if (result.error is None):
weight = weights.get(name, 0.0)
if (weight > 0):
weighted_ai_probs.append(result.ai_probability * weight)
weighted_human_probs.append(result.human_probability * weight)
confidences.append(result.confidence)
total_weight += weight
if ((not weighted_ai_probs) or (total_weight == 0)):
return 0.5, 0.5, 0.0, 0.5, breakdown or {}
ai_prob = sum(weighted_ai_probs) / total_weight
human_prob = sum(weighted_human_probs) / total_weight
mixed_prob = 0.0 # Fallback
avg_confidence = sum(confidences) / len(confidences) if confidences else 0.5
return ai_prob, human_prob, mixed_prob, avg_confidence, breakdown
def _get_document_ensemble_probability(self, ensemble_result: Optional[EnsembleResult], metric_results: Dict[str, MetricResult],
weights: Dict[str, float]) -> Tuple[float, float, float, float, Dict[str, float]]:
"""
Get document-level ensemble probability
"""
if ensemble_result:
# Use existing ensemble result
breakdown = {name: result.ai_probability for name, result in metric_results.items()}
return (ensemble_result.ai_probability, ensemble_result.human_probability, ensemble_result.mixed_probability,
ensemble_result.overall_confidence, breakdown)
else:
# Calculate from metrics
return self._calculate_weighted_probability(metric_results, weights, {})
def _apply_domain_specific_adjustments(self, sentence: str, ai_prob: float, sentence_length: int) -> float:
"""
Apply domain-specific adjustments to AI probability with limits
"""
original_prob = ai_prob
adjustments = list()
sentence_lower = sentence.lower()
# Technical & AI/ML domains
if (self.domain in [Domain.AI_ML, Domain.SOFTWARE_DEV, Domain.TECHNICAL_DOC, Domain.ENGINEERING, Domain.SCIENCE]):
if self._has_technical_terms(sentence_lower):
adjustments.append(1.1)
elif self._has_code_like_patterns(sentence):
adjustments.append(1.15)
elif (sentence_length > 35):
adjustments.append(1.05)
# Creative & informal domains
elif (self.domain in [Domain.CREATIVE, Domain.SOCIAL_MEDIA, Domain.BLOG_PERSONAL]):
if self._has_informal_language(sentence_lower):
adjustments.append(0.7)
elif self._has_emotional_language(sentence):
adjustments.append(0.8)
elif (sentence_length < 10):
adjustments.append(0.8)
# Academic & formal domains
elif (self.domain in [Domain.ACADEMIC, Domain.LEGAL, Domain.MEDICAL]):
if self._has_citation_patterns(sentence):
adjustments.append(0.8)
elif self._has_technical_terms(sentence_lower):
adjustments.append(1.1)
elif (sentence_length > 40):
adjustments.append(1.1)
# Business & professional domains
elif (self.domain in [Domain.BUSINESS, Domain.MARKETING, Domain.JOURNALISM]):
if self._has_business_jargon(sentence_lower):
adjustments.append(1.05)
elif self._has_ambiguous_phrasing(sentence_lower):
adjustments.append(0.9)
elif (15 <= sentence_length <= 25):
adjustments.append(0.9)
# Tutorial & educational domains
elif (self.domain == Domain.TUTORIAL):
if self._has_instructional_language(sentence_lower):
adjustments.append(0.85)
elif self._has_step_by_step_pattern(sentence):
adjustments.append(0.8)
elif self._has_examples(sentence):
adjustments.append(0.9)
# General domain - minimal adjustments
elif (self.domain == Domain.GENERAL):
if self._has_complex_structure(sentence):
adjustments.append(0.9)
elif self._has_repetition(sentence):
adjustments.append(1.1)
# Apply adjustments with limits - take strongest 2 adjustments maximum
if adjustments:
# Sort by impact (farthest from 1.0)
adjustments.sort(key = lambda x: abs(x - 1.0), reverse = True)
# Limit to 2 strongest
strongest_adjustments = adjustments[:2]
for adjustment in strongest_adjustments:
ai_prob *= adjustment
# Ensure probability stays within bounds and doesn't change too drastically : Maximum 30% change from original
max_change = 0.3
bounded_prob = max(original_prob - max_change, min(original_prob + max_change, ai_prob))
return max(0.0, min(1.0, bounded_prob))
def _apply_metric_specific_adjustments(self, metric_name: str, sentence: str, base_prob: float, sentence_length: int, thresholds: MetricThresholds) -> float:
"""
Apply metric-specific adjustments
"""
# Use metrics from ensemble
if (metric_name == "perplexity"):
if (sentence_length < 8):
return min(1.0, base_prob * 1.2)
elif (sentence_length > 25):
return max(0.0, base_prob * 0.8)
elif (metric_name == "entropy"):
words = sentence.split()
if (len(words) > 3):
unique_words = len(set(words))
diversity = unique_words / len(words)
if (diversity < 0.6):
return min(1.0, base_prob * 1.2)
elif (diversity > 0.8):
return max(0.0, base_prob * 0.8)
elif (metric_name == "linguistic"):
complexity_score = self._analyze_sentence_complexity(sentence)
if (complexity_score < 0.3):
return min(1.0, base_prob * 1.1)
elif (complexity_score > 0.7):
return max(0.0, base_prob * 0.9)
elif (metric_name == "structural"):
if ((sentence_length < 5) or (sentence_length > 40)):
return max(0.0, base_prob * 0.8)
elif (8 <= sentence_length <= 20):
return min(1.0, base_prob * 1.1)
elif (metric_name == "semantic_analysis"):
if self._has_repetition(sentence):
return min(1.0, base_prob * 1.2)
elif (metric_name == "multi_perturbation_stability"):
# MultiPerturbationStability adjustments for sentence level
if (sentence_length > 15):
return min(1.0, base_prob * 1.1)
return base_prob
def _get_color_for_probability(self, probability: float, is_mixed_content: bool = False, mixed_prob: float = 0.0) -> Tuple[str, str, str]:
"""
Get color class with mixed content support and no threshold gaps
"""
# Handle probability = 1.0 explicitly
if (probability >= 1.0):
return "very-high-ai", "#fecaca", "Very likely AI-generated (100%)"
# Check mixed content first
if (is_mixed_content and (mixed_prob > self.MIXED_THRESHOLD)):
return "mixed-content", "#e9d5ff", f"Mixed AI/Human content ({mixed_prob:.1%} mixed)"
# Iterate through thresholds correctly
for min_thresh, max_thresh, color_class, color_hex, tooltip in self.COLOR_THRESHOLDS:
if (min_thresh <= probability < max_thresh):
return color_class, color_hex, tooltip
# Fallback for probability = 1.0 (should be caught above, but just in case)
return "very-high-ai", "#fecaca", "Very likely AI-generated"
def _generate_ensemble_tooltip(self, sentence: str, ai_prob: float, human_prob: float, mixed_prob: float, confidence: float, confidence_level: ConfidenceLevel,
tooltip_base: str, breakdown: Optional[Dict[str, float]] = None, is_mixed_content: bool = False) -> str:
"""
Generate enhanced tooltip with ENSEMBLE information
"""
tooltip = f"{tooltip_base}\n"
if is_mixed_content:
tooltip += "🔀 MIXED CONTENT DETECTED\n"
tooltip += f"AI Probability: {ai_prob:.1%}\n"
tooltip += f"Human Probability: {human_prob:.1%}\n"
tooltip += f"Mixed Probability: {mixed_prob:.1%}\n"
tooltip += f"Confidence: {confidence:.1%} ({confidence_level.value.replace('_', ' ').title()})\n"
tooltip += f"Domain: {self.domain.value.replace('_', ' ').title()}\n"
tooltip += f"Length: {len(sentence.split())} words"
if breakdown:
tooltip += "\n\nMetric Breakdown:"
# Show top 4 metrics
for metric, prob in list(breakdown.items())[:4]:
tooltip += f"\n• {metric}: {prob:.1%}"
tooltip += f"\n\nEnsemble Method: {getattr(self.ensemble, 'primary_method', 'fallback')}"
return tooltip
def _has_citation_patterns(self, sentence: str) -> bool:
"""
Check for academic citation patterns
"""
citation_indicators = ['et al.', 'ibid.', 'cf.', 'e.g.', 'i.e.', 'vol.', 'pp.', 'ed.', 'trans.', 'reference', 'cited', 'according to']
return any(indicator in sentence.lower() for indicator in citation_indicators)
def _has_informal_language(self, sentence: str) -> bool:
"""
Check for informal language patterns
"""
informal_indicators = ['lol', 'omg', 'btw', 'imo', 'tbh', 'afaik', 'smh', '👋', '😂', '❤️', 'haha', 'wow', 'awesome']
return any(indicator in sentence.lower() for indicator in informal_indicators)
def _has_technical_terms(self, sentence: str) -> bool:
"""
Check for domain-specific technical terms
"""
technical_indicators = ['hereinafter', 'whereas', 'aforementioned', 'diagnosis', 'prognosis', 'etiology',
'algorithm', 'neural network', 'machine learning', 'api', 'endpoint', 'database',
'quantum', 'thermodynamics', 'hypothesis', 'methodology']
return any(indicator in sentence.lower() for indicator in technical_indicators)
def _has_ambiguous_phrasing(self, sentence: str) -> bool:
"""
Check for ambiguous phrasing that might indicate human writing
"""
ambiguous_indicators = ['perhaps', 'maybe', 'possibly', 'likely', 'appears to', 'seems to', 'might be', 'could be']
return any(indicator in sentence.lower() for indicator in ambiguous_indicators)
def _has_complex_structure(self, sentence: str) -> bool:
"""
Check if sentence has complex linguistic structure
"""
words = sentence.split()
if (len(words) < 8):
return False
complex_indicators = ['which', 'that', 'although', 'because', 'while', 'when', 'if', 'however', 'therefore']
return any(indicator in sentence.lower() for indicator in complex_indicators)
def _has_emotional_language(self, sentence: str) -> bool:
"""
Check for emotional or subjective language
"""
emotional_indicators = ['feel', 'believe', 'think', 'wonder', 'hope', 'wish', 'love', 'hate', 'frustrating', 'exciting']
return any(indicator in sentence.lower() for indicator in emotional_indicators)
def _has_business_jargon(self, sentence: str) -> bool:
"""
Check for business jargon
"""
jargon_indicators = ['synergy', 'leverage', 'bandwidth', 'circle back', 'touch base', 'value add', 'core competency']
return any(indicator in sentence.lower() for indicator in jargon_indicators)
def _has_instructional_language(self, sentence: str) -> bool:
"""
Check for instructional language patterns
"""
instructional_indicators = ['step by step', 'firstly', 'secondly', 'finally', 'note that', 'remember to', 'make sure']
return any(indicator in sentence.lower() for indicator in instructional_indicators)
def _has_step_by_step_pattern(self, sentence: str) -> bool:
"""
Check for step-by-step instructions
"""
step_patterns = ['step 1', 'step 2', 'step 3', 'step one', 'step two', 'first step', 'next step']
return any(pattern in sentence.lower() for pattern in step_patterns)
def _has_examples(self, sentence: str) -> bool:
"""
Check for example indicators
"""
example_indicators = ['for example', 'for instance', 'such as', 'e.g.', 'as an example']
return any(indicator in sentence.lower() for indicator in example_indicators)
def _has_code_like_patterns(self, sentence: str) -> bool:
"""
Check for code-like patterns in technical domains
"""
code_patterns = ['function', 'variable', 'class', 'method', 'import', 'def ', 'void ', 'public ', 'private ']
return any(pattern in sentence for pattern in code_patterns)
def _analyze_sentence_complexity(self, sentence: str) -> float:
"""
Analyze sentence complexity (0 = simple, 1 = complex)
"""
words = sentence.split()
if (len(words) < 5):
return 0.2
complexity_indicators = ['although', 'because', 'while', 'when', 'if', 'since', 'unless', 'until', 'which', 'that', 'who', 'whom', 'whose', 'and', 'but', 'or', 'yet', 'so', 'however', 'therefore', 'moreover', 'furthermore', 'nevertheless', ',', ';', ':', '—']
score = 0.0
if (len(words) > 15):
score += 0.3
elif (len(words) > 25):
score += 0.5
indicator_count = sum(1 for indicator in complexity_indicators if indicator in sentence.lower())
score += min(0.5, indicator_count * 0.1)
clause_indicators = [',', ';', 'and', 'but', 'or', 'because', 'although']
clause_count = sum(1 for indicator in clause_indicators if indicator in sentence.lower())
score += min(0.2, clause_count * 0.05)
return min(1.0, score)
def _has_repetition(self, sentence: str) -> bool:
"""
Check if sentence has word repetition (common in AI text)
"""
words = sentence.lower().split()
if (len(words) < 6):
return False
word_counts = dict()
for word in words:
if (len(word) > 3):
word_counts[word] = word_counts.get(word, 0) + 1
repeated_words = [word for word, count in word_counts.items() if count > 2]
return len(repeated_words) > 0
def _split_sentences(self, text: str) -> List[str]:
"""
Split the text chunk into multiple sentences
"""
sentences = self.text_processor.split_sentences(text)
filtered_sentences = list()
for sentence in sentences:
clean_sentence = sentence.strip()
if (len(clean_sentence) >= 3):
filtered_sentences.append(clean_sentence)
return filtered_sentences
def generate_html(self, highlighted_sentences: List[HighlightedSentence], include_legend: bool = False, include_metrics: bool = True) -> str:
"""
Generate HTML with highlighted sentences
Arguments:
----------
highlighted_sentences { List[HighlightedSentence] } : Sentences with highlighting data
include_legend { bool } : Whether to include legend (set to False to avoid duplicates)
include_metrics { bool } : Whether to include metrics summary
Returns:
--------
{ str } : HTML content
"""
html_parts = list()
# Add CSS
html_parts.append(self._generate_enhanced_css())
# Only include legend if explicitly requested (usually False to avoid duplicates)
if include_legend:
html_parts.append(self._generate_legend_html())
# Add highlighted text container
html_parts.append('<div class="highlighted-text">')
for sent in highlighted_sentences:
extra_class = " mixed-highlight" if sent.is_mixed_content else ""
html_parts.append(f'<span class="highlight {sent.color_class}{extra_class}" '
f'data-ai-prob="{sent.ai_probability:.4f}" '
f'data-human-prob="{sent.human_probability:.4f}" '
f'data-mixed-prob="{sent.mixed_probability:.4f}" '
f'data-confidence="{sent.confidence:.4f}" '
f'data-confidence-level="{sent.confidence_level.value}" '
f'data-domain="{self.domain.value}" '
f'data-sentence-idx="{sent.index}" '
f'data-is-mixed="{str(sent.is_mixed_content).lower()}" '
f'title="{sent.tooltip}">'
f'{sent.text}'
f'</span> '
)
html_parts.append('</div>')
# Add metrics summary if requested (separate from legend)
if include_metrics and highlighted_sentences:
html_parts.append(self._generate_metrics_summary(highlighted_sentences))
return '\n'.join(html_parts)
def _generate_enhanced_css(self) -> str:
"""
Generate CSS for highlighting for Better readability
"""
return """
<style>
.highlighted-text {
line-height: 1.8;
font-size: 16px;
font-family: 'Georgia', serif;
padding: 20px;
background: #ffffff;
border-radius: 8px;
box-shadow: 0 2px 4px rgba(0,0,0,0.1);
margin-bottom: 20px;
}
.highlight {
padding: 2px 4px;
margin: 0 1px;
border-radius: 3px;
transition: all 0.2s ease;
cursor: help;
border-bottom: 2px solid transparent;
color: #000000 !important;
font-weight: 500;
position: relative;
}
.highlight:hover {
transform: translateY(-1px);
box-shadow: 0 4px 12px rgba(0,0,0,0.15);
z-index: 10;
text-shadow: 0 1px 1px rgba(255,255,255,0.8);
}
/* AI indicators - Lighter backgrounds for better readability */
.very-high-ai {
background-color: #fee2e2;
border-bottom-color: #ef4444;
}
.high-ai {
background-color: #fed7aa;
border-bottom-color: #f97316;
}
.medium-ai {
background-color: #fef3c7;
border-bottom-color: #f59e0b;
}
/* Uncertain */
.uncertain {
background-color: #fef9c3;
border-bottom-color: #fbbf24;
}
/* Human indicators - Lighter backgrounds */
.medium-human {
background-color: #ecfccb;
border-bottom-color: #a3e635;
}
.high-human {
background-color: #bbf7d0;
border-bottom-color: #4ade80;
}
.very-high-human {
background-color: #dcfce7;
border-bottom-color: #22c55e;
}
/* Mixed content */
.mixed-content {
background-color: #e9d5ff;
border-bottom-color: #a855f7;
background-image: repeating-linear-gradient(45deg, transparent, transparent 5px, rgba(168, 85, 247, 0.1) 5px, rgba(168, 85, 247, 0.1) 10px);
}
.mixed-highlight:hover {
border: 2px dashed #a855f7;
}
/* Summary styles */
.highlight-summary {
margin-bottom: 20px;
padding: 15px;
background: #f9fafb;
border-radius: 8px;
border: 1px solid #e5e7eb;
}
.highlight-summary h4 {
margin: 0 0 10px 0;
font-size: 14px;
font-weight: 600;
color: #374151;
}
.summary-stats {
display: grid;
grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));
gap: 10px;
}
.stat-item {
display: flex;
justify-content: space-between;
align-items: center;
padding: 8px 12px;
background: white;
border-radius: 6px;
border: 1px solid #e5e7eb;
}
.stat-label {
font-size: 13px;
color: #6b7280;
}
.stat-value {
font-size: 13px;
font-weight: 600;
color: #374151;
}
</style>
"""
def _generate_metrics_summary(self, sentences: List[HighlightedSentence]) -> str:
"""
Generate summary statistics for highlighted sentences
"""
if not sentences:
return ""
# Calculate summary metrics
total_sentences = len(sentences)
# Count sentences by category
very_high_ai = len([s for s in sentences if s.color_class == "very-high-ai"])
high_ai = len([s for s in sentences if s.color_class == "high-ai"])
medium_ai = len([s for s in sentences if s.color_class == "medium-ai"])
uncertain = len([s for s in sentences if s.color_class == "uncertain"])
medium_human = len([s for s in sentences if s.color_class == "medium-human"])
high_human = len([s for s in sentences if s.color_class == "high-human"])
very_high_human = len([s for s in sentences if s.color_class == "very-high-human"])
mixed = len([s for s in sentences if s.color_class == "mixed-content"])
# Calculate overall risk score (weighted average)
weighted_risk = 0.0
for sent in sentences:
weight = self.RISK_WEIGHTS.get(sent.color_class, 0.4)
weighted_risk += sent.ai_probability * weight
overall_risk_score = weighted_risk / total_sentences if total_sentences else 0.0
# Calculate average probabilities
avg_ai_prob = sum(s.ai_probability for s in sentences) / total_sentences
avg_human_prob = sum(s.human_probability for s in sentences) / total_sentences
# Sentence counts
ai_sentences = very_high_ai + high_ai + medium_ai
human_sentences = very_high_human + high_human + medium_human
html = f"""
<div class="highlight-summary">
<h4>📊 Text Analysis Summary</h4>
<div class="summary-stats">
<div class="stat-item">
<span class="stat-label">Overall Risk Score</span>
<span class="stat-value">{overall_risk_score:.1%}</span>
</div>
<div class="stat-item">
<span class="stat-label">Average AI Probability</span>
<span class="stat-value">{avg_ai_prob:.1%}</span>
</div>
<div class="stat-item">
<span class="stat-label">AI Sentences</span>
<span class="stat-value">{ai_sentences} ({ai_sentences/total_sentences:.1%})</span>
</div>
<div class="stat-item">
<span class="stat-label">Human Sentences</span>
<span class="stat-value">{human_sentences} ({human_sentences/total_sentences:.1%})</span>
</div>
<div class="stat-item">
<span class="stat-label">Uncertain Sentences</span>
<span class="stat-value">{uncertain} ({uncertain/total_sentences:.1%})</span>
</div>
<div class="stat-item">
<span class="stat-label">Mixed Sentences</span>
<span class="stat-value">{mixed} ({mixed/total_sentences:.1%})</span>
</div>
<div class="stat-item">
<span class="stat-label">Total Sentences</span>
<span class="stat-value">{total_sentences}</span>
</div>
<div class="stat-item">
<span class="stat-label">Domain</span>
<span class="stat-value">{self.domain.value.replace('_', ' ').title()}</span>
</div>
</div>
</div>
"""
return html
def _generate_legend_html(self) -> str:
"""
Generate legend HTML - Only used if explicitly requested
"""
return """
<div class="highlight-legend" style="margin-bottom: 20px; padding: 15px; background: #f8fafc; border-radius: 8px; border: 1px solid #e2e8f0;">
<h4 style="margin: 0 0 10px 0; font-size: 14px; font-weight: 600; color: #374151;">AI Detection Legend</h4>
<div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(200px, 1fr)); gap: 8px;">
<div style="display: flex; align-items: center; gap: 8px;">
<div style="width: 16px; height: 16px; background: #dcfce7; border: 1px solid #22c55e; border-radius: 3px;"></div>
<span style="font-size: 12px; color: #374151;">Very Likely Human (0-10%)</span>
</div>
<div style="display: flex; align-items: center; gap: 8px;">
<div style="width: 16px; height: 16px; background: #bbf7d0; border: 1px solid #4ade80; border-radius: 3px;"></div>
<span style="font-size: 12px; color: #374151;">Likely Human (10-25%)</span>
</div>
<div style="display: flex; align-items: center; gap: 8px;">
<div style="width: 16px; height: 16px; background: #86efac; border: 1px solid #16a34a; border-radius: 3px;"></div>
<span style="font-size: 12px; color: #374151;">Possibly Human (25-40%)</span>
</div>
<div style="display: flex; align-items: center; gap: 8px;">
<div style="width: 16px; height: 16px; background: #fef9c3; border: 1px solid #fbbf24; border-radius: 3px;"></div>
<span style="font-size: 12px; color: #374151;">Uncertain (40-60%)</span>
</div>
<div style="display: flex; align-items: center; gap: 8px;">
<div style="width: 16px; height: 16px; background: #fde68a; border: 1px solid #f59e0b; border-radius: 3px;"></div>
<span style="font-size: 12px; color: #374151;">Possibly AI (60-75%)</span>
</div>
<div style="display: flex; align-items: center; gap: 8px;">
<div style="width: 16px; height: 16px; background: #fed7aa; border: 1px solid #f97316; border-radius: 3px;"></div>
<span style="font-size: 12px; color: #374151;">Likely AI (75-90%)</span>
</div>
<div style="display: flex; align-items: center; gap: 8px;">
<div style="width: 16px; height: 16px; background: #fecaca; border: 1px solid #ef4444; border-radius: 3px;"></div>
<span style="font-size: 12px; color: #374151;">Very Likely AI (90-100%)</span>
</div>
<div style="display: flex; align-items: center; gap: 8px;">
<div style="width: 16px; height: 16px; background: #e9d5ff; border: 1px solid #a855f7; border-radius: 3px;"></div>
<span style="font-size: 12px; color: #374151;">Mixed Content</span>
</div>
</div>
</div>
"""
def calculate_metrics(self, highlighted_sentences: List[HighlightedSentence]) -> Dict[str, float]:
"""
Calculate metrics for external use
Arguments:
----------
highlighted_sentences { List[HighlightedSentence] } : Sentences with highlighting data
Returns:
--------
{ Dict[str, float] } : Dictionary with metrics
"""
if not highlighted_sentences:
return {}
total_sentences = len(highlighted_sentences)
# Calculate weighted risk score
weighted_risk = 0.0
for sent in highlighted_sentences:
weight = self.RISK_WEIGHTS.get(sent.color_class, 0.4)
weighted_risk += sent.ai_probability * weight
overall_risk_score = weighted_risk / total_sentences
# Count sentences by category
ai_sentences = len([s for s in highlighted_sentences if s.ai_probability >= 0.6])
human_sentences = len([s for s in highlighted_sentences if s.ai_probability <= 0.4])
uncertain_sentences = len([s for s in highlighted_sentences if 0.4 < s.ai_probability < 0.6])
mixed_sentences = len([s for s in highlighted_sentences if s.is_mixed_content])
# Average probabilities
avg_ai_prob = sum(s.ai_probability for s in highlighted_sentences) / total_sentences
avg_human_prob = sum(s.human_probability for s in highlighted_sentences) / total_sentences
avg_confidence = sum(s.confidence for s in highlighted_sentences) / total_sentences
return {'overall_risk_score' : overall_risk_score,
'avg_ai_probability' : avg_ai_prob,
'avg_human_probability' : avg_human_prob,
'avg_confidence' : avg_confidence,
'ai_sentence_count' : ai_sentences,
'human_sentence_count' : human_sentences,
'uncertain_sentence_count' : uncertain_sentences,
'mixed_sentence_count' : mixed_sentences,
'total_sentences' : total_sentences,
'ai_sentence_percentage' : ai_sentences / total_sentences,
'human_sentence_percentage' : human_sentences / total_sentences,
}
# Export
__all__ = ["TextHighlighter",
"HighlightedSentence",
] |