File size: 12,609 Bytes
c293f7c | 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 | #!/usr/bin/env python3
"""
Explainability Engine for Misinformation Detection
Provides interpretable explanations for why content is classified as fake/real.
"""
import re
import logging
from typing import Dict, List, Tuple
import numpy as np
logger = logging.getLogger(__name__)
class ExplainabilityEngine:
"""
Generate human-readable explanations for misinformation predictions
"""
def __init__(self):
# Sensational keywords that trigger fake news alerts
self.sensational_keywords = [
'breaking', 'urgent', 'shocking', 'exclusive', 'exposed', 'revealed',
'secret', 'hidden', 'conspiracy', 'cover-up', 'scandal', 'bombshell',
'viral', 'must see', 'unbelievable', 'incredible'
]
# Emotional manipulation words
self.emotional_keywords = [
'outraged', 'furious', 'devastated', 'terrified', 'shocked',
'disgusted', 'betrayed', 'angry', 'hate'
]
# Clickbait patterns
self.clickbait_patterns = [
r'you (won\'t|will not) believe',
r'\d+ (things|ways|reasons|facts)',
r'this will shock',
r'number \d+ will',
r'what happens next'
]
# Credible attribution patterns
self.attribution_patterns = [
'according to', 'sources say', 'officials confirm', 'study shows',
'research indicates', 'data reveals', 'experts believe', 'report states',
'ministry announced', 'government stated'
]
def explain_prediction(self, text: str, prediction_result: Dict) -> Dict:
"""
Generate comprehensive explanation for a prediction
Args:
text: The article text
prediction_result: Result from FakeNewsDetector
Returns:
Dictionary with explanation details
"""
# Extract highlighted snippets
highlighted = self.extract_highlighted_snippets(text, prediction_result)
# Generate reason summary
reasons = self.generate_reasons(text, prediction_result)
# Get flagged keywords
flagged_keywords = self.get_flagged_keywords(text)
# Calculate explanation confidence
explanation_confidence = self._calculate_explanation_confidence(
reasons, flagged_keywords
)
return {
'highlighted_snippets': highlighted,
'reasons': reasons,
'flagged_keywords': flagged_keywords,
'explanation_confidence': explanation_confidence,
'summary': self._generate_summary(prediction_result, reasons)
}
def extract_highlighted_snippets(self, text: str, prediction_result: Dict) -> List[Dict]:
"""
Extract and highlight text snippets that contributed to the prediction
"""
snippets = []
sentences = self._split_into_sentences(text)
for i, sentence in enumerate(sentences):
sentence_lower = sentence.lower()
# Check for sensational language
sensational_found = [
keyword for keyword in self.sensational_keywords
if keyword in sentence_lower
]
# Check for emotional manipulation
emotional_found = [
keyword for keyword in self.emotional_keywords
if keyword in sentence_lower
]
# Check for clickbait patterns
clickbait_found = any(
re.search(pattern, sentence_lower)
for pattern in self.clickbait_patterns
)
# Check for lack of attribution
has_attribution = any(
pattern in sentence_lower
for pattern in self.attribution_patterns
)
# Calculate sentence risk score
risk_score = 0
reasons = []
if sensational_found:
risk_score += 0.3
reasons.append(f"Sensational language: {', '.join(sensational_found)}")
if emotional_found:
risk_score += 0.3
reasons.append(f"Emotional manipulation: {', '.join(emotional_found)}")
if clickbait_found:
risk_score += 0.2
reasons.append("Clickbait pattern detected")
if not has_attribution and len(sentence.split()) > 10:
risk_score += 0.2
reasons.append("Lacks source attribution")
# Only include high-risk sentences
if risk_score >= 0.3:
snippets.append({
'sentence': sentence,
'position': i,
'risk_score': min(risk_score, 1.0),
'reasons': reasons
})
# Sort by risk score and return top 5
snippets.sort(key=lambda x: x['risk_score'], reverse=True)
return snippets[:5]
def generate_reasons(self, text: str, prediction_result: Dict) -> List[str]:
"""Generate list of reasons for the classification"""
reasons = []
text_lower = text.lower()
# Check components from prediction
components = prediction_result.get('components', {})
# ML Classification reason
ml_result = components.get('ml_classification', {})
if ml_result.get('fake_probability', 0) > 0.6:
reasons.append(f"ML model confidence: {ml_result.get('fake_probability', 0):.2%} fake probability")
# Linguistic analysis
linguistic = components.get('linguistic_analysis', {})
if linguistic.get('sensational_words', 0) > 2:
reasons.append(f"Contains {linguistic.get('sensational_words')} sensational words")
if linguistic.get('emotional_words', 0) > 1:
reasons.append(f"Uses {linguistic.get('emotional_words')} emotional manipulation words")
if linguistic.get('clickbait_patterns', 0) > 0:
reasons.append("Contains clickbait patterns")
# Source credibility
source_cred = components.get('source_credibility', {})
if source_cred.get('credibility_score', 0.5) < 0.4:
reasons.append(f"Low source credibility ({source_cred.get('source_type', 'unknown')} source)")
# Fact checking
fact_check = components.get('fact_checking', {})
if fact_check.get('checked') and fact_check.get('verdict') == 'false':
reasons.append(f"Debunked by {fact_check.get('source', 'fact-checkers')}")
# Satellite verification
satellite = components.get('satellite_verification')
if satellite and not satellite.get('verified'):
reasons.append("Location claim could not be verified")
# Cross-reference
cross_ref = components.get('cross_reference_score', 0.5)
if cross_ref < 0.4:
reasons.append("Limited corroboration from other sources")
# Add general linguistic flags
if not any(pattern in text_lower for pattern in self.attribution_patterns):
reasons.append("Lacks proper source attribution")
if text.count('!') > 3:
reasons.append("Excessive use of exclamation marks")
# Limit to top 8 reasons
return reasons[:8]
def get_flagged_keywords(self, text: str) -> Dict[str, List[str]]:
"""Get all flagged keywords by category"""
text_lower = text.lower()
flagged = {
'sensational': [
word for word in self.sensational_keywords
if word in text_lower
],
'emotional': [
word for word in self.emotional_keywords
if word in text_lower
],
'clickbait_patterns': [
pattern for pattern in self.clickbait_patterns
if re.search(pattern, text_lower)
]
}
# Remove empty categories
return {k: v for k, v in flagged.items() if v}
def _split_into_sentences(self, text: str) -> List[str]:
"""Split text into sentences"""
# Simple sentence splitter
sentences = re.split(r'[.!?]+', text)
return [s.strip() for s in sentences if s.strip()]
def _calculate_explanation_confidence(self, reasons: List[str],
flagged_keywords: Dict) -> float:
"""Calculate confidence in the explanation"""
# More reasons and flagged keywords = higher confidence
reason_score = min(len(reasons) / 8, 1.0) * 0.6
keyword_score = min(sum(len(v) for v in flagged_keywords.values()) / 10, 1.0) * 0.4
return reason_score + keyword_score
def _generate_summary(self, prediction_result: Dict, reasons: List[str]) -> str:
"""Generate a human-readable summary"""
verdict = prediction_result.get('verdict', 'uncertain')
confidence = prediction_result.get('confidence', 0)
if verdict == 'fake':
summary = f"This content is classified as **LIKELY FAKE** with {confidence:.0%} confidence. "
elif verdict == 'real':
summary = f"This content appears to be **LIKELY REAL** with {confidence:.0%} confidence. "
else:
summary = f"This content is **UNCERTAIN** and requires further verification. "
if reasons:
top_reasons = reasons[:3]
summary += f"Key indicators: {', '.join(top_reasons)}."
return summary
def generate_detailed_report(self, text: str, prediction_result: Dict) -> str:
"""Generate a detailed explainability report"""
explanation = self.explain_prediction(text, prediction_result)
report = []
report.append("=" * 70)
report.append("MISINFORMATION DETECTION EXPLANATION")
report.append("=" * 70)
# Summary
report.append(f"\n{explanation['summary']}\n")
# Reasons
report.append("REASONS FOR CLASSIFICATION:")
report.append("-" * 70)
for i, reason in enumerate(explanation['reasons'], 1):
report.append(f"{i}. {reason}")
# Flagged keywords
if explanation['flagged_keywords']:
report.append(f"\nFLAGGED KEYWORDS:")
report.append("-" * 70)
for category, keywords in explanation['flagged_keywords'].items():
report.append(f" {category.upper()}: {', '.join(keywords)}")
# Highlighted snippets
if explanation['highlighted_snippets']:
report.append(f"\nHIGH-RISK SNIPPETS:")
report.append("-" * 70)
for snippet in explanation['highlighted_snippets'][:3]:
report.append(f"\n📍 \"{snippet['sentence']}\"")
report.append(f" Risk Score: {snippet['risk_score']:.2f}")
report.append(f" Issues: {', '.join(snippet['reasons'])}")
report.append("\n" + "=" * 70)
return "\n".join(report)
def demo():
"""Demo the explainability engine"""
# Sample text
sample_text = """
BREAKING: Government hiding shocking vaccine microchip truth!
This urgent revelation will shock you! Secret documents exposed by anonymous sources
reveal the terrifying conspiracy. You won't believe what they're hiding from us!
"""
# Mock prediction result
mock_result = {
'verdict': 'fake',
'confidence': 0.87,
'components': {
'ml_classification': {'fake_probability': 0.85},
'linguistic_analysis': {'sensational_words': 4, 'emotional_words': 2, 'clickbait_patterns': 1},
'source_credibility': {'credibility_score': 0.2, 'source_type': 'unknown'},
'fact_checking': {'checked': False}
}
}
engine = ExplainabilityEngine()
# Generate explanation
explanation = engine.explain_prediction(sample_text, mock_result)
print("\n🔍 Explainability Demo:")
print(engine.generate_detailed_report(sample_text, mock_result))
if __name__ == "__main__":
demo()
|