Update app.py

app.py

@@ -770,42 +770,55 @@ if uploaded_file is not None:
         sns.heatmap(pivot, ax=axes[word_pos], cmap='YlOrRd')
         axes[word_pos].set_title(f'Word Position {word_pos+1}')
     st.pyplot(fig)
-
+    
     st.subheader("4. Character Connection Patterns")
-    char_bigrams = Counter()
-    for chars in chars_list:
-        for i in range(len(chars)-1):
-            char_bigrams[tuple(chars[i:i+2])] += 1
+    
+    @st.cache_data
+    def generate_char_network(chars_list):
+        char_bigrams = Counter()
+        for chars in chars_list:
+            for i in range(len(chars)-1):
+                char_bigrams[tuple(chars[i:i+2])] += 1
+        return char_bigrams
+    
+    char_bigrams = generate_char_network(chars_list)
     
     # Create graph with explicit weight handling
     G = nx.Graph()
     edges_with_weights = []
     
-    # Extract top 50 bigrams and their weights
-    for (char1, char2), count in char_bigrams.most_common(50):  # Increased to 50
-        if count > len(chars_list) * 0.01:  # Add frequency threshold
+    # Get the total number of bigrams for percentage calculation
+    total_bigrams = sum(char_bigrams.values())
+    
+    # Extract significant bigrams and their weights
+    for (char1, char2), count in char_bigrams.items():
+        if count > total_bigrams * 0.01:  # Only include if more than 1% of total
             edges_with_weights.append((char1, char2, count))
     
+    # Sort by weight and take top connections
+    edges_with_weights.sort(key=lambda x: x[2], reverse=True)
+    edges_with_weights = edges_with_weights[:50]  # Top 50 connections
+    
     # Add edges to graph
     for char1, char2, weight in edges_with_weights:
         G.add_edge(char1, char2, weight=weight)
     
-    fig4 = plt.figure(figsize=(15, 15))  # Increased figure size
-    pos = nx.spring_layout(G, k=1)  # Adjusted layout spacing
+    fig4 = plt.figure(figsize=(15, 15))
+    pos = nx.spring_layout(G, k=1, seed=42)  # Fixed seed for stable layout
     
-    # Calculate edge widths from explicit weight values
-    weights = [w for _, _, w in edges_with_weights]
+    # Calculate edge widths directly from weights
+    weights = [G[u][v]['weight'] for u,v in G.edges()]
     max_weight = max(weights) if weights else 1
     edge_widths = [w/max_weight * 5 for w in weights]
     
-    # Draw network with pre-calculated widths
+    # Draw network
     nx.draw(G, pos, 
             with_labels=True,
             node_color='lightblue',
-            node_size=2000,  # Increased node size
-            font_size=14,    # Increased font size
+            node_size=2000,
+            font_size=14,
             width=edge_widths)
     
     plt.title("Character Connection Network")
-    st.pyplot(fig4)
-
+    st.pyplot(fig4, clear_figure=True)
+