Update app.py

app.py

@@ -13,221 +13,118 @@ st.set_page_config(
 # Custom CSS styling
 st.markdown("""
 <style>
-.analysis-card {
+.analysis-section {
     padding: 1.5rem;
-    border-radius: 0.5rem;
-    box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.1);
     margin: 1rem 0;
 }
-.positive { border-left: 0.5rem solid #10B981 !important; }
-.negative { border-left: 0.5rem solid #EF4444 !important; }
-.step-indicator {
-    font-size: 1.2rem;
-    font-weight: 600;
-    color: #6B7280;
-    margin: 1.5rem 0;
+.response-box {
+    background-color: #F8FAFC;
+    padding: 1.5rem;
+    border-radius: 0.5rem;
+    margin: 1rem 0;
 }
 </style>
 """, unsafe_allow_html=True)
 
-# Response templates
-response_templates = {
-    "billing": {
-        "positive": "Thank you for your positive feedback on our billing services. We're delighted to hear about your experience...",
-        "negative": "We sincerely apologize for the issues with our billing services. Your concerns are important to us..."
-    },
-    # Add other template categories...
-}
-
 @st.cache_resource
 def load_models():
-    """Load all required ML models"""
+    """Load ML models"""
     with st.spinner("🚀 Loading AI models..."):
         device = 0 if torch.cuda.is_available() else -1
         
-        # Zero-shot classification model
-        topic_classifier = pipeline(
+        classifier = pipeline(
             "zero-shot-classification",
             model="MoritzLaurer/deberta-v3-base-zeroshot-v1",
             device=device,
             torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32
         )
         
-        # Sentiment analysis model
         sentiment_analyzer = pipeline(
             "sentiment-analysis", 
             model="cardiffnlp/twitter-roberta-base-sentiment-latest",
             device=device
         )
         
-        # Response generation model
         response_generator = pipeline(
             "text2text-generation",
             model="Leo66277/custom-response-generator",
             device=device
         )
         
-    return topic_classifier, sentiment_analyzer, response_generator
+    return classifier, sentiment_analyzer, response_generator
 
 def analyze_review(text, models):
-    """Full analysis pipeline"""
-    topic_classifier, sentiment_analyzer, response_generator = models
+    """Analysis pipeline"""
+    classifier, sentiment_analyzer, response_generator = models
     
-    # Step 1: Topic classification
-    with st.spinner("🔍 Analyzing review topic..."):
-        topic_labels = list(response_templates.keys())
-        topic_result = topic_classifier(text, topic_labels, multi_label=False)
-        main_topic = topic_result['labels'][0]
-        topic_confidence = topic_result['scores'][0]
+    # Topic classification
+    topics = ["customer service", "mobile app", "credit card", "account security"]
+    topic_result = classifier(text, topics, multi_label=False)
+    main_topic = topic_result['labels'][0]
     
-    # Step 2: Sentiment analysis
-    with st.spinner("💡 Detecting sentiment..."):
-        sentiment_result = sentiment_analyzer(text)[0]
-        sentiment_label = "positive" if sentiment_result['label'] in ['POSITIVE', 'positive'] else "negative"
-        sentiment_score = sentiment_result['score']
+    # Sentiment analysis
+    sentiment_result = sentiment_analyzer(text)[0]
+    sentiment = "positive" if sentiment_result['label'] in ['POSITIVE', 'positive'] else "negative"
     
-    # Step 3: Generate response
-    with st.spinner("✍️ Generating professional response..."):
-        prompt = f"""Review: {text}
+    # Generate response
+    prompt = f"""Review: {text}
 Topic: {main_topic}
-Sentiment: {sentiment_label}
+Sentiment: {sentiment}
 Generate response:"""
-        
-        generated_response = response_generator(
-            prompt,
-            max_length=300,
-            num_return_sequences=1,
-            do_sample=True,
-            temperature=0.7
-        )[0]['generated_text'].strip()
     
-    return {
-        "topic": main_topic,
-        "topic_confidence": f"{topic_confidence:.1%}",
-        "sentiment": sentiment_label,
-        "sentiment_score": f"{sentiment_score:.1%}",
-        "response": generated_response,
-        "template_response": response_templates.get(main_topic, {}).get(sentiment_label)
-    }
-
-def format_text(text, line_width=80):
-    """Text formatting"""
-    return "\n".join(textwrap.wrap(text, width=line_width))
+    response = response_generator(
+        prompt,
+        max_length=300,
+        num_return_sequences=1,
+        do_sample=True,
+        temperature=0.7
+    )[0]['generated_text'].strip()
+    
+    return main_topic, sentiment, response
 
-# Main interface
 def main():
-    st.title("🏦 Bank Review Analysis System")
-    st.markdown("---")
+    st.title("🏦 Bank Review Analysis")
     
-    # Sidebar configuration
-    with st.sidebar:
-        st.header("Settings")
-        show_details = st.checkbox("Show analysis details", True)
-        show_template = st.checkbox("Show template response", True)
-        st.markdown("---")
-        st.caption("Version: 1.0 | Developer: Leo66277")
+    # Main input
+    review = st.text_area(
+        "Enter customer review",
+        height=150,
+        placeholder="Paste your bank review here..."
+    )
     
-    # Main input area
-    with st.form(key="analysis_form"):
-        col1, col2 = st.columns([3, 1])
-        with col1:
-            review_text = st.text_area(
-                "Enter customer review",
-                placeholder="Paste your bank review here...",
-                height=150
-            )
-        with col2:
-            st.markdown("### Example Reviews")
-            st.caption("▶️ Difficulty logging into mobile banking")
-            st.caption("▶️ Unreasonable credit card annual fee")
-            st.caption("▶️ Excellent counter service attitude")
-        
-        submitted = st.form_submit_button("Start Analysis", type="primary")
-
     # Load models
     models = load_models()
-
-    if submitted and review_text.strip():
-        # Perform analysis
-        result = analyze_review(review_text, models)
-        
-        # Display results
-        with st.container():
-            st.markdown("## Analysis Results")
-            
-            # Sentiment indicators
-            sentiment_icon = "✅" if result['sentiment'] == "positive" else "⚠️"
-            
-            # Main metrics card
-            with st.expander("Key Metrics", expanded=True):
-                cols = st.columns(4)
-                cols[0].metric("Detected Topic", result['topic'])
-                cols[1].metric("Topic Confidence", result['topic_confidence'])
-                cols[2].metric("Sentiment", f"{sentiment_icon} {result['sentiment'].capitalize()}")
-                cols[3].metric("Sentiment Strength", result['sentiment_score'])
-            
-            # Detailed analysis
-            if show_details:
-                with st.expander("Detailed Analysis", expanded=False):
-                    tab1, tab2, tab3 = st.tabs(["Original Review", "Topic Distribution", "Sentiment Analysis"])
-                    
-                    with tab1:
-                        st.code(format_text(review_text), language="text")
-                        
-                    with tab2:
-                        st.caption("Topic Probability Distribution")
-                        topic_data = {
-                            'Topic': result['topic_distribution']['labels'],
-                            'Confidence': result['topic_distribution']['scores']
-                        }
-                        st.bar_chart(topic_data, x='Topic', y='Confidence')
-                    
-                    with tab3:
-                        st.caption("Sentiment Analysis Raw Results")
-                        st.json({
-                            "label": result['sentiment'],
-                            "score": float(result['sentiment_score'].strip('%'))/100
-                        })
+    
+    if st.button("Analyze Review"):
+        if review.strip():
+            topic, sentiment, response = analyze_review(review, models)
             
-            # Response generation
-            st.markdown("## 💬 Response Suggestions")
+            # Display results
+            st.markdown("---")
+            st.markdown("### Analysis Results")
             
+            # Results columns
             col1, col2 = st.columns(2)
             with col1:
-                with st.container(border=True):
-                    st.markdown("### 🚀 AI Generated Response")
-                    st.markdown(f'<div class="analysis-card positive">{format_text(result["response"])}</div>', 
-                               unsafe_allow_html=True)
+                st.markdown(f"**Detected Topic:** {topic}")
+                st.markdown(f"**Sentiment:** {sentiment.capitalize()}")
             
-            if show_template and result['template_response']:
-                with col2:
-                    with st.container(border=True):
-                        st.markdown("### 📋 Template Response")
-                        st.markdown(f'<div class="analysis-card">{format_text(result["template_response"])}</div>', 
-                                   unsafe_allow_html=True)
+            # Response section
+            st.markdown("---")
+            st.markdown("### Suggested Response")
+            st.markdown(f'<div class="response-box">{textwrap.fill(response, width=80)}</div>', 
+                       unsafe_allow_html=True)
             
-            # Download report
+            # Download button
             st.download_button(
-                label="Download Full Report",
-                data=f"""Analysis Report
-                
-                Original Review:
-                {review_text}
-                
-                Topic Analysis: {result['topic']} ({result['topic_confidence']})
-                Sentiment Analysis: {result['sentiment'].capitalize()} ({result['sentiment_score']})
-                
-                AI Generated Response:
-                {result['response']}
-                
-                {"Template Response: " + result['template_response'] if result['template_response'] else ""}
-                """,
-                file_name="bank_review_analysis.txt",
+                label="Download Report",
+                data=f"Topic: {topic}\nSentiment: {sentiment}\n\nResponse:\n{response}",
+                file_name="review_analysis.txt",
                 mime="text/plain"
             )
-
-    elif submitted:
-        st.warning("⚠️ Please enter valid review content")
+            
+        else:
+            st.warning("Please enter a review to analyze")
 
 if __name__ == "__main__":
     main()