Update models/hate_speech_classifier.py

models/hate_speech_classifier.py

@@ -204,15 +204,121 @@ class HateSpeechClassifier:
                 print(f"❌ Error loading {model_key} pretrained model: {e}")
                 model_info["pipeline"] = None
     
+    # async def classify_with_custom_model(self, text: str, language: str) -> Dict:
+    #     """Classify using language-specific custom model"""
+    #     if language == "english":
+    #         if not self.english_model_loaded:
+    #             return None
+    #         model = self.english_model
+    #         vectorizer = self.english_vectorizer
+    #     elif language == "bengali":
+    #         if not self.bengali_model_loaded:
+    #             return None
+    #         model = self.bengali_model
+    #         vectorizer = self.bengali_vectorizer
+    #     else:
+    #         return None
+        
+    #     try:
+    #         X = vectorizer.transform([text])
+    #         prediction = model.predict(X)[0]
+            
+    #         if hasattr(model, 'predict_proba'):
+    #             probabilities = model.predict_proba(X)[0]
+    #             confidence = float(max(probabilities))
+    #         else:
+    #             confidence = 0.75
+            
+    #         if language == "english":
+    #             if prediction == 0:
+    #                 category = "neutral"
+    #             else:
+    #                 category = "hate_speech"
+    #         else:
+    #             if prediction == 0:
+    #                 category = "neutral"
+    #             elif prediction == 1:
+    #                 category = "offensive"
+    #             else:
+    #                 category = "hate_speech"
+            
+    #         return {
+    #             "category": category,
+    #             "confidence": confidence,
+    #             "method": f"custom_model_{language}",
+    #             "raw_prediction": int(prediction)
+    #         }
+    #     except Exception as e:
+    #         print(f"❌ Custom model classification failed: {e}")
+    #         return None
+    # async def classify_with_custom_model(self, text: str, language: str) -> Dict:
+    #     """Classify using language-specific custom model"""
+    #     if language == "english":
+    #         if not self.english_model_loaded:
+    #             return None
+    #         model = self.english_model
+    #         vectorizer = self.english_vectorizer
+    #     elif language == "bengali":
+    #         if not self.bengali_model_loaded:
+    #             return None
+    #         model = self.bengali_model
+    #         vectorizer = self.bengali_vectorizer
+    #     else:
+    #         return None
+        
+    #     try:
+    #         X = vectorizer.transform([text])
+    #         prediction = model.predict(X)[0]
+            
+    #         if hasattr(model, 'predict_proba'):
+    #             probabilities = model.predict_proba(X)[0]
+    #             # ✅ FIX: Use probability of the PREDICTED class, not max
+    #             confidence = float(probabilities[prediction])
+                
+    #             # Debug logging
+    #             print(f"🔍 Custom Model Debug:")
+    #             print(f"   Prediction: {prediction}")
+    #             print(f"   Probabilities: {probabilities}")
+    #             print(f"   Confidence: {confidence:.4f}")
+    #         else:
+    #             confidence = 0.75
+            
+    #         if language == "english":
+    #             if prediction == 0:
+    #                 category = "neutral"
+    #             else:
+    #                 category = "hate_speech"
+    #         else:
+    #             if prediction == 0:
+    #                 category = "neutral"
+    #             elif prediction == 1:
+    #                 category = "offensive"
+    #             else:
+    #                 category = "hate_speech"
+            
+    #         return {
+    #             "category": category,
+    #             "confidence": confidence,
+    #             "method": f"custom_model_{language}",
+    #             "raw_prediction": int(prediction),
+    #             "probabilities": probabilities.tolist() if hasattr(model, 'predict_proba') else None
+    #         }
+    #     except Exception as e:
+    #         print(f"❌ Custom model classification failed: {e}")
+    #         import traceback
+    #         traceback.print_exc()
+    #         return None
     async def classify_with_custom_model(self, text: str, language: str) -> Dict:
         """Classify using language-specific custom model"""
         if language == "english":
             if not self.english_model_loaded:
+                print("❌ English model not loaded, returning None")
                 return None
             model = self.english_model
             vectorizer = self.english_vectorizer
         elif language == "bengali":
             if not self.bengali_model_loaded:
+                print("❌ Bengali model not loaded, returning None")
                 return None
             model = self.bengali_model
             vectorizer = self.bengali_vectorizer
@@ -220,13 +326,25 @@ class HateSpeechClassifier:
             return None
         
         try:
+            # Debug: Check model type
+            print(f"🔍 Model type: {type(model)}")
+            print(f"🔍 Has predict_proba: {hasattr(model, 'predict_proba')}")
+            
             X = vectorizer.transform([text])
             prediction = model.predict(X)[0]
             
+            print(f"🔍 Raw prediction: {prediction}")
+            
             if hasattr(model, 'predict_proba'):
                 probabilities = model.predict_proba(X)[0]
-                confidence = float(max(probabilities))
+                confidence = float(probabilities[prediction])
+                
+                print(f"🔍 Custom Model Debug:")
+                print(f"   Prediction: {prediction}")
+                print(f"   Probabilities: {probabilities}")
+                print(f"   Confidence (probabilities[{prediction}]): {confidence:.4f}")
             else:
+                print("⚠️ Model doesn't have predict_proba, using fallback 0.75")
                 confidence = 0.75
             
             if language == "english":
@@ -246,12 +364,14 @@ class HateSpeechClassifier:
                 "category": category,
                 "confidence": confidence,
                 "method": f"custom_model_{language}",
-                "raw_prediction": int(prediction)
+                "raw_prediction": int(prediction),
+                "probabilities": probabilities.tolist() if hasattr(model, 'predict_proba') else None
             }
         except Exception as e:
             print(f"❌ Custom model classification failed: {e}")
+            import traceback
+            traceback.print_exc()
             return None
-    
     async def classify_with_pretrained_model(self, text: str, language: str = "english") -> Dict:
         """Classify using ensemble of pretrained models with translation support"""