Update app.py

app.py

@@ -17,6 +17,66 @@ def calculate_star_rating(positive_percent):
     else:
         return 1
 
+@st.cache_resource
+def analyze_sentiment(comments, progress_bar=None, status_text=None):
+    """Perform sentiment analysis on a list of comments"""
+    # Load model inside the function
+    classifier = pipeline(
+        "text-classification",
+        model="KeonBlackwell/movie_sentiment_model",
+        tokenizer="distilbert-base-uncased"
+    )
+    
+    results = []
+    total = len(comments)
+    
+    for i, comment in enumerate(comments):
+        if progress_bar and status_text:
+            progress = (i+1)/total
+            progress_bar.progress(progress)
+            status_text.text(f"Analyzing sentiment for {i+1}/{total} reviews...")
+        
+        prediction = classifier(comment)[0]
+        results.append({
+            'comment': comment,
+            'sentiment': 1 if prediction['label'] == 'LABEL_1' else 0,
+            'confidence': prediction['score']
+        })
+    
+    return results
+
+@st.cache_resource
+def extract_keyphrases(text, top_n=5):
+    """Extract top keyphrases from text"""
+    # Load model inside the function
+    keyphrase_extractor = pipeline(
+        "token-classification",
+        model="ml6team/keyphrase-extraction-distilbert-inspec",
+        aggregation_strategy="simple"
+    )
+    
+    keyphrases = keyphrase_extractor(text)
+    # Sort by confidence and take the top N
+    top_keyphrases = sorted(keyphrases, key=lambda x: x['score'], reverse=True)[:top_n]
+    return top_keyphrases
+
+@st.cache_resource
+def generate_summary(text, max_length=130, min_length=30):
+    """Generate summary from text"""
+    # Load model inside the function
+    summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
+    
+    # Limit text length to avoid model limitations
+    max_input_length = 1024  # Maximum input length for the model
+    if len(text) > max_input_length:
+        text = text[:max_input_length]
+    
+    summary = summarizer(text, 
+                        max_length=max_length, 
+                        min_length=min_length, 
+                        do_sample=False)
+    return summary[0]['summary_text']
+
 def main():
     st.set_page_config(page_title="Movie Review Analysis System", page_icon="🎬")
 
@@ -32,31 +92,6 @@ def main():
     </style>
     """, unsafe_allow_html=True)
 
-    # Model loading
-    with st.spinner("Loading all models, this may take a few minutes..."):
-        try:
-            # Sentiment analysis model
-            classifier = pipeline(
-                "text-classification",
-                model="KeonBlackwell/movie_sentiment_model",
-                tokenizer="distilbert-base-uncased"
-            )
-
-            # Keyphrase extraction model
-            keyphrase_extractor = pipeline(
-                "token-classification",
-                model="ml6team/keyphrase-extraction-distilbert-inspec",
-                aggregation_strategy="simple"
-            )
-
-            # Summarization model
-            summarizer = pipeline("summarization",
-                                model="facebook/bart-large-cnn")
-
-        except Exception as e:
-            st.error(f"Model loading failed: {str(e)}")
-            return
-
     # Page layout
     st.title("🎬 Movie Review Batch Analysis System")
     st.markdown("""
@@ -91,24 +126,11 @@ def main():
             progress_bar = st.progress(0)
             status_text = st.empty()
 
-            results = []
-            total = len(comments)
-
-            # Batch prediction
+            # Sentiment analysis
             try:
-                # Sentiment analysis
-                for i, comment in enumerate(comments):
-                    progress = (i+1)/total
-                    progress_bar.progress(progress)
-                    status_text.text(f"Analyzing sentiment for {i+1}/{total} reviews...")
-
-                    prediction = classifier(comment)[0]
-                    results.append({
-                        'comment': comment,
-                        'sentiment': 1 if prediction['label'] == 'LABEL_1' else 0,
-                        'confidence': prediction['score']
-                    })
-
+                with st.spinner("Loading sentiment analysis model..."):
+                    results = analyze_sentiment(comments, progress_bar, status_text)
+                
                 # Convert to DataFrame
                 result_df = pd.DataFrame(results)
 
@@ -144,10 +166,8 @@ def main():
                 combined_text = " ".join(comments)
 
                 # Keyphrase extraction
-                with st.spinner("Extracting keyphrases..."):
-                    keyphrases = keyphrase_extractor(combined_text)
-                    # Sort by confidence and take the top 5
-                    top_keyphrases = sorted(keyphrases, key=lambda x: x['score'], reverse=True)[:5]
+                with st.spinner("Loading keyphrase extraction model..."):
+                    top_keyphrases = extract_keyphrases(combined_text)
 
                 # Show keyphrases
                 st.markdown("**🔍 Extracted Keyphrases:**")
@@ -168,20 +188,12 @@ def main():
                     """, unsafe_allow_html=True)
 
                 # Generate summary
-                with st.spinner("Generating review summary..."):
-                    # Limit text length to avoid model limitations
-                    max_length = 1024  # Maximum input length for the model
-                    if len(combined_text) > max_length:
-                        combined_text = combined_text[:max_length]
-
-                    summary = summarizer(combined_text, 
-                                        max_length=130, 
-                                        min_length=30, 
-                                        do_sample=False)
+                with st.spinner("Loading summarization model..."):
+                    summary = generate_summary(combined_text)
 
                 # Show summary
                 st.markdown("**📝 Review Summary:**")
-                st.info(summary[0]['summary_text'])
+                st.info(summary)
 
                 # Generate downloadable file
                 with tempfile.NamedTemporaryFile(delete=False, suffix=".csv") as tmp: