File size: 16,134 Bytes
94a65e4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f36a10a
94a65e4
 
f36a10a
94a65e4
f36a10a
 
 
 
 
94a65e4
f36a10a
 
207a2e4
 
 
94a65e4
 
 
 
207a2e4
94a65e4
 
207a2e4
 
94a65e4
 
 
 
 
 
 
 
f36a10a
 
207a2e4
f36a10a
 
 
 
207a2e4
 
f36a10a
 
 
 
 
94a65e4
 
f36a10a
207a2e4
 
f36a10a
 
 
 
207a2e4
 
 
94a65e4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
207a2e4
 
 
94a65e4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f36a10a
 
207a2e4
f36a10a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
207a2e4
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
# import torch
# from transformers import AutoTokenizer, AutoModelForSequenceClassification
# import networkx as nx
# import spacy
# import pickle
# import pandas as pd
# import google.generativeai as genai
# import json

# # Load spaCy for NER
# nlp = spacy.load("en_core_web_sm")

# # Load the trained ML model
# model_path = "./results/checkpoint-5030"  # Replace with the actual path to your model
# tokenizer = AutoTokenizer.from_pretrained('microsoft/deberta-v3-small')
# model = AutoModelForSequenceClassification.from_pretrained(model_path)
# model.eval()

# #########################
# def setup_gemini():
#     genai.configure(api_key='AIzaSyAQzWpSyWyYCM1G5f-G0ulRCQkXuY7admA')
#     model = genai.GenerativeModel('gemini-pro')
#     return model
# #########################

# # Load the knowledge graph
# graph_path = "./models/knowledge_graph.pkl"  # Replace with the actual path to your knowledge graph
# with open(graph_path, 'rb') as f:
#     graph_data = pickle.load(f)

# knowledge_graph = nx.DiGraph()
# knowledge_graph.add_nodes_from(graph_data['nodes'].items())
# for u, edges in graph_data['edges'].items():
#     for v, data in edges.items():
#         knowledge_graph.add_edge(u, v, **data)

# def predict_with_model(text):
#     """Predict whether the news is real or fake using the ML model."""
#     inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=512)
#     with torch.no_grad():
#         outputs = model(**inputs)
#     probabilities = torch.nn.functional.softmax(outputs.logits, dim=-1)
#     predicted_label = torch.argmax(probabilities, dim=-1).item()
#     return "FAKE" if predicted_label == 1 else "REAL"

# def update_knowledge_graph(text, is_real):
#     """Update the knowledge graph with the new article."""
#     entities = extract_entities(text)
#     for entity, entity_type in entities:
#         if not knowledge_graph.has_node(entity):
#             knowledge_graph.add_node(
#                 entity,
#                 type=entity_type,
#                 real_count=1 if is_real else 0,
#                 fake_count=0 if is_real else 1
#             )
#         else:
#             if is_real:
#                 knowledge_graph.nodes[entity]['real_count'] += 1
#             else:
#                 knowledge_graph.nodes[entity]['fake_count'] += 1

#     for i, (entity1, _) in enumerate(entities):
#         for entity2, _ in entities[i+1:]:
#             if not knowledge_graph.has_edge(entity1, entity2):
#                 knowledge_graph.add_edge(
#                     entity1,
#                     entity2,
#                     weight=1,
#                     is_real=is_real
#                 )
#             else:
#                 knowledge_graph[entity1][entity2]['weight'] += 1

# def extract_entities(text):
#     """Extract named entities from text using spaCy."""
#     doc = nlp(text)
#     entities = [(ent.text, ent.label_) for ent in doc.ents]
#     return entities

# def predict_with_knowledge_graph(text):
#     """Predict whether the news is real or fake using the knowledge graph."""
#     entities = extract_entities(text)
#     real_score = 0
#     fake_score = 0

#     for entity, _ in entities:
#         if knowledge_graph.has_node(entity):
#             real_count = knowledge_graph.nodes[entity].get('real_count', 0)
#             fake_count = knowledge_graph.nodes[entity].get('fake_count', 0)
#             total = real_count + fake_count
#             if total > 0:
#                 real_score += real_count / total
#                 fake_score += fake_count / total

#     if real_score > fake_score:
#         return "REAL"
#     else:
#         return "FAKE"

# def predict_news(text):
#     """Predict whether the news is real or fake using both the ML model and the knowledge graph."""
#     # Predict with the ML model
#     ml_prediction = predict_with_model(text)
#     is_real = ml_prediction == "REAL"

#     # Update the knowledge graph
#     update_knowledge_graph(text, is_real)

#     # Predict with the knowledge graph
#     kg_prediction = predict_with_knowledge_graph(text)

#     # Combine predictions (for simplicity, we use the ML model's prediction here)
#     # You can enhance this by combining the scores from both predictions
#     return ml_prediction if ml_prediction == kg_prediction else "UNCERTAIN"

# #########################
# # def analyze_content_gemini(model, text):
# #     prompt = f"""Analyze this news text and provide results in the following JSON-like format:

# #     TEXT: {text}

# #     Please provide analysis in these specific sections:

# #     1. GEMINI ANALYSIS:
# #        - Predicted Classification: [Real/Fake]
# #        - Confidence Score: [0-100%]
# #        - Reasoning: [Key points for classification]

# #     2. TEXT CLASSIFICATION:
# #         - Content category/topic
# #         - Writing style: [Formal/Informal/Clickbait]
# #         - Target audience
# #         - Content type: [news/opinion/editorial]

# #     3. SENTIMENT ANALYSIS:
# #        - Primary emotion
# #        - Emotional intensity (1-10)
# #        - Sensationalism Level: [High/Medium/Low]
# #        - Bias Indicators: [List if any]
# #        - Tone: (formal/informal), [Professional/Emotional/Neutral]
# #        - Key emotional triggers

# #     4. ENTITY RECOGNITION:
# #         - Source Credibility: [High/Medium/Low]
# #        - People mentioned
# #        - Organizations
# #        - Locations
# #        - Dates/Time references
# #        - Key numbers/statistics

# #     5. CONTEXT EXTRACTION:
# #        - Main narrative/story
# #        - Supporting elements
# #        - Key claims
# #        - Narrative structure

# #     6. FACT CHECKING:
# #        - Verifiable Claims: [List main claims]
# #        - Evidence Present: [Yes/No]
# #        - Fact Check Score: [0-100%]

# #     Format the response clearly with distinct sections."""
    
# #     response = model.generate_content(prompt)
# #     return response.text

# def analyze_content_gemini(model, text):
#     prompt = f"""Analyze this news text and return a JSON object with the following structure:
#     {{
#         "gemini_analysis": {{
#             "predicted_classification": "Real or Fake",
#             "confidence_score": "0-100",
#             "reasoning": ["point1", "point2"]
#         }},
#         "text_classification": {{
#             "category": "",
#             "writing_style": "Formal/Informal/Clickbait",
#             "target_audience": "",
#             "content_type": "news/opinion/editorial"
#         }},
#         "sentiment_analysis": {{
#             "primary_emotion": "",
#             "emotional_intensity": "1-10",
#             "sensationalism_level": "High/Medium/Low",
#             "bias_indicators": ["bias1", "bias2"],
#             "tone": {{"formality": "formal/informal", "style": "Professional/Emotional/Neutral"}},
#             "emotional_triggers": ["trigger1", "trigger2"]
#         }},
#         "entity_recognition": {{
#             "source_credibility": "High/Medium/Low",
#             "people": ["person1", "person2"],
#             "organizations": ["org1", "org2"],
#             "locations": ["location1", "location2"],
#             "dates": ["date1", "date2"],
#             "statistics": ["stat1", "stat2"]
#         }},
#         "context": {{
#             "main_narrative": "",
#             "supporting_elements": ["element1", "element2"],
#             "key_claims": ["claim1", "claim2"],
#             "narrative_structure": ""
#         }},
#         "fact_checking": {{
#             "verifiable_claims": ["claim1", "claim2"],
#             "evidence_present": "Yes/No",
#             "fact_check_score": "0-100"
#         }}
#     }}

#     Analyze this text and return only the JSON response: {text}"""
    
#     response = model.generate_content(prompt)
#     # return json.loads(response.text)
#     # Add error handling and response cleaning
#     try:
#         # Clean the response text to ensure it's valid JSON
#         cleaned_text = response.text.strip()
#         if cleaned_text.startswith('```json'):
#             cleaned_text = cleaned_text[7:-3]  # Remove ```json and ``` markers
#         return json.loads(cleaned_text)
#     except json.JSONDecodeError:
#         # Return a default structured response if JSON parsing fails
#         return {
#             "gemini_analysis": {
#                 "predicted_classification": "UNCERTAIN",
#                 "confidence_score": "50",
#                 "reasoning": ["Analysis failed to generate valid JSON"]
#             }
#         }


# def clean_gemini_output(text):
#     """Remove markdown formatting from Gemini output"""
#     text = text.replace('##', '')
#     text = text.replace('**', '')
#     return text

# def get_gemini_analysis(text):
#     """Get detailed content analysis from Gemini."""
#     gemini_model = setup_gemini()
#     gemini_analysis = analyze_content_gemini(gemini_model, text)
#     # cleaned_analysis = clean_gemini_output(gemini_analysis)
#     # return cleaned_analysis
#     return gemini_analysis
# #########################

# def main():
#     print("Welcome to the News Classifier!")
#     print("Enter your news text below. Type 'Exit' to quit.")
    
#     while True:
#         news_text = input("\nEnter news text: ")
        
#         if news_text.lower() == 'exit':
#             print("Thank you for using the News Classifier!")
#             return
            
#         # First get ML and Knowledge Graph prediction
#         prediction = predict_news(news_text)
#         print(f"\nML and Knowledge Graph Analysis: {prediction}")
        
#         # Then get Gemini analysis
#         print("\n=== Detailed Gemini Analysis ===")
#         gemini_result = get_gemini_analysis(news_text)
#         print(gemini_result)


# if __name__ == "__main__":
#     main()

import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification, DebertaV2Tokenizer
import networkx as nx
import spacy
import pickle
import google.generativeai as genai
import json
import os
import dotenv

# Load environment variables
dotenv.load_dotenv()

def load_models():
    """Load all required ML models"""
    nlp = spacy.load("en_core_web_sm")
    model_path = "./results/checkpoint-753"
    tokenizer = DebertaV2Tokenizer.from_pretrained('microsoft/deberta-v3-small')
    model = AutoModelForSequenceClassification.from_pretrained(model_path)
    model.eval()
    return nlp, tokenizer, model

def load_knowledge_graph():
    """Load and initialize knowledge graph"""
    graph_path = "./models/knowledge_graph.pkl"
    with open(graph_path, 'rb') as f:
        graph_data = pickle.load(f)
    knowledge_graph = nx.DiGraph()
    knowledge_graph.add_nodes_from(graph_data['nodes'].items())
    for u, edges in graph_data['edges'].items():
        for v, data in edges.items():
            knowledge_graph.add_edge(u, v, **data)
    return knowledge_graph

def setup_gemini():
    """Initialize Gemini model"""
    genai.configure(api_key=os.getenv("GEMINI_API"))
    model = genai.GenerativeModel('gemini-pro')
    return model

def predict_with_model(text, tokenizer, model):
    """Make predictions using the ML model"""
    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=512)
    with torch.no_grad():
        outputs = model(**inputs)
    probabilities = torch.nn.functional.softmax(outputs.logits, dim=-1)
    predicted_label = torch.argmax(probabilities, dim=-1).item()
    confidence = probabilities[0][predicted_label].item() * 100
    return "FAKE" if predicted_label == 1 else "REAL", confidence

def extract_entities(text, nlp):
    """Extract named entities from text"""
    doc = nlp(text)
    entities = [(ent.text, ent.label_) for ent in doc.ents]
    return entities

def update_knowledge_graph(text, is_real, knowledge_graph, nlp):
    """Update knowledge graph with new information"""
    entities = extract_entities(text, nlp)
    for entity, entity_type in entities:
        if not knowledge_graph.has_node(entity):
            knowledge_graph.add_node(
                entity,
                type=entity_type,
                real_count=1 if is_real else 0,
                fake_count=0 if is_real else 1
            )
        else:
            if is_real:
                knowledge_graph.nodes[entity]['real_count'] += 1
            else:
                knowledge_graph.nodes[entity]['fake_count'] += 1

    for i, (entity1, _) in enumerate(entities):
        for entity2, _ in entities[i+1:]:
            if not knowledge_graph.has_edge(entity1, entity2):
                knowledge_graph.add_edge(
                    entity1,
                    entity2,
                    weight=1,
                    is_real=is_real
                )
            else:
                knowledge_graph[entity1][entity2]['weight'] += 1

def predict_with_knowledge_graph(text, knowledge_graph, nlp):
    """Make predictions using the knowledge graph"""
    entities = extract_entities(text, nlp)
    real_score = 0
    fake_score = 0

    for entity, _ in entities:
        if knowledge_graph.has_node(entity):
            real_count = knowledge_graph.nodes[entity].get('real_count', 0)
            fake_count = knowledge_graph.nodes[entity].get('fake_count', 0)
            total = real_count + fake_count
            if total > 0:
                real_score += real_count / total
                fake_score += fake_count / total

    total_score = real_score + fake_score
    if total_score == 0:
        return "UNCERTAIN", 50.0
    
    if real_score > fake_score:
        confidence = (real_score / total_score) * 100
        return "REAL", confidence
    else:
        confidence = (fake_score / total_score) * 100
        return "FAKE", confidence

def analyze_content_gemini(model, text):
    """Analyze content using Gemini model"""
    prompt = f"""Analyze this news text and return a JSON object with the following structure:
    {{
        "gemini_analysis": {{
            "predicted_classification": "Real or Fake",
            "confidence_score": "0-100",
            "reasoning": ["point1", "point2"]
        }},
        "text_classification": {{
            "category": "",
            "writing_style": "Formal/Informal/Clickbait",
            "target_audience": "",
            "content_type": "news/opinion/editorial"
        }},
        "sentiment_analysis": {{
            "primary_emotion": "",
            "emotional_intensity": "1-10",
            "sensationalism_level": "High/Medium/Low",
            "bias_indicators": ["bias1", "bias2"],
            "tone": {{"formality": "formal/informal", "style": "Professional/Emotional/Neutral"}},
            "emotional_triggers": ["trigger1", "trigger2"]
        }},
        "entity_recognition": {{
            "source_credibility": "High/Medium/Low",
            "people": ["person1", "person2"],
            "organizations": ["org1", "org2"],
            "locations": ["location1", "location2"],
            "dates": ["date1", "date2"],
            "statistics": ["stat1", "stat2"]
        }},
        "context": {{
            "main_narrative": "",
            "supporting_elements": ["element1", "element2"],
            "key_claims": ["claim1", "claim2"],
            "narrative_structure": ""
        }},
        "fact_checking": {{
            "verifiable_claims": ["claim1", "claim2"],
            "evidence_present": "Yes/No",
            "fact_check_score": "0-100"
        }}
    }}

    Analyze this text and return only the JSON response: {text}"""
    
    response = model.generate_content(prompt)
    try:
        cleaned_text = response.text.strip()
        if cleaned_text.startswith('```json'):
            cleaned_text = cleaned_text[7:-3]
        return json.loads(cleaned_text)
    except json.JSONDecodeError:
        return {
            "gemini_analysis": {
                "predicted_classification": "UNCERTAIN",
                "confidence_score": "50",
                "reasoning": ["Analysis failed to generate valid JSON"]
            }
        }