Update app.py

app.py

@@ -36,23 +36,6 @@ def load_models():
 
 model, tfidf_vectorizer, word2vec_model = load_models()
 
-def calculate_combined_similarity(paraphrasing_similarity, direct_text_comparison_similarity, summarization_similarity):
-    try:
-        # Adjusted weights
-        weighted_paraphrasing = paraphrasing_similarity * 0.6
-        weighted_direct_comparison = direct_text_comparison_similarity * 0.3
-        weighted_summarization = summarization_similarity * 0.1
-        
-        # Combined similarity calculation with robustness to outliers
-        combined_similarity = (weighted_paraphrasing + weighted_direct_comparison + weighted_summarization)
-        
-        # Ensure the score is between 0 and 100%
-        combined_similarity = max(0, min(100, combined_similarity))
-        return combined_similarity
-    except Exception as e:
-        st.error(f"Error calculating combined similarity: {e}")
-        return 0
-
 # Initialize session state for results table if not already present
 if 'results_df' not in st.session_state:
     st.session_state.results_df = pd.DataFrame(columns=[
@@ -120,13 +103,8 @@ def calculate_word_similarity_ratio(text1, text2):
 
 @st.cache_data
 def calculate_bleu_score(reference, candidate):
-    try:
-        from nltk.translate.bleu_score import sentence_bleu
-        return sentence_bleu([reference.split()], candidate.split()) * 100
-    except Exception as e:
-        st.error(f"Error calculating BLEU score: {e}")
-        return 0
-
+    from nltk.translate.bleu_score import sentence_bleu
+    return sentence_bleu([reference.split()], candidate.split())
 
 @st.cache_data
 def calculate_rouge_l_score(reference, candidate):
@@ -216,9 +194,8 @@ def calculate_direct_text_comparison_similarity(text1, text2):
 @st.cache_data
 def calculate_summarization_similarity(text1, text2):
     try:
-        wmd = calculate_wmd(text1, text2)  # Placeholder for actual WMD calculation
-        normalized_wmd = max(0, min(1, wmd))
-        return (1 - normalized_wmd) * 100  # Scale to percentage
+        wmd = calculate_wmd(text1, text2)
+        return (1 - wmd) * 100
     except Exception as e:
         st.error(f"Error calculating summarization similarity: {e}")
         return 0
@@ -261,9 +238,15 @@ with col1:
             paraphrasing_similarity = calculate_paraphrasing_similarity(text_input_1, text_input_2)
             direct_text_comparison_similarity = calculate_direct_text_comparison_similarity(text_input_1, text_input_2)
             summarization_similarity = calculate_summarization_similarity(text_input_1, text_input_2)
-
+            if summarization_similarity<0:
+                summarization_similarity=0
+            if direct_text_comparison_similarity<0:
+                direct_text_comparison_similarity=0                
+                        
             # Combine all metrics into a single similarity score
-            total_similarity = calculate_combined_similarity(paraphrasing_similarity, direct_text_comparison_similarity, summarization_similarity)
+            total_similarity = (paraphrasing_similarity * 0.6 +  # High weight
+                                direct_text_comparison_similarity * 0.3 +  # Moderate weight
+                                summarization_similarity * 0.1)  # Low weight
 
             # Update results table in session state
             new_row = pd.Series({
@@ -304,59 +287,92 @@ with col1:
                     }
                 </style>
             """
+            good_case = """
+                <style>
+                    .good {
+                        font-size: 48px;
+                        font-weight: bold;
+                        color: #4CAF50; /* Green color for positive emphasis */
+                        background-color: #f0f0f5;
+                        padding: 20px;
+                        border-radius: 15px;
+                        text-align: center;
+                        margin-top: 30px;
+                        box-shadow: 2px 2px 12px rgba(0, 0, 0, 0.1);
+                    }
+                </style>
+            """
+            bad_case = """
+                <style>
+                    .bad {
+                        font-size: 48px;
+                        font-weight: bold;
+                        color: #FF0000; /* Red color for negative emphasis */
+                        background-color: #f0f0f5;
+                        padding: 20px;
+                        border-radius: 15px;
+                        text-align: center;
+                        margin-top: 30px;
+                        box-shadow: 2px 2px 12px rgba(0, 0, 0, 0.1);
+                    }
+                </style>
+            """
 
             # Apply the style
             st.markdown(combined_score_style, unsafe_allow_html=True)
-
+            st.markdown(good_case, unsafe_allow_html=True)
+            st.markdown(bad_case, unsafe_allow_html=True)
             # Display the combined similarity score
             st.markdown(f'<div class="combined-score">Combined Similarity Score: {total_similarity:.2f}%</div>', unsafe_allow_html=True)
+            # Calculate context-words difference
+            context_words_diff = int(paraphrasing_similarity) - int(direct_text_comparison_similarity)
+
+            # Display distinguishing factor
+            if total_similarity >= 100:
+                st.markdown(f'<div class="bad">Similar Responses</div>', unsafe_allow_html=True)
+            elif total_similarity >= 55:
+                if context_words_diff >= 42 and context_words_diff < 57.08:
+                    st.markdown(f'<div class="bad">Similar Responses</div>', unsafe_allow_html=True)
+                elif context_words_diff > 35:
+                    st.markdown(f'<div class="good">Response 2 is better.</div>', unsafe_allow_html=True)                
+                else:
+                    st.markdown(f'<div class="bad">Similar Responses</div>', unsafe_allow_html=True)
+            else:
+                st.markdown(f'<div class="bad">Similar Responses</div>', unsafe_allow_html=True)
+
 
-def plot_radar_chart(data, labels):
-    num_vars = len(labels)
-    angles = np.linspace(0, 2 * np.pi, num_vars, endpoint=False).tolist()
-    angles += angles[:1]
-
-    fig, ax = plt.subplots(figsize=(6, 6), subplot_kw=dict(polar=True))
-    for idx, entry in enumerate(data):
-        values = [
-            entry.get(label, 0) for label in labels
-        ]
-        values += values[:1]
-        ax.fill(angles, values, color=sns.color_palette("husl", len(data))[idx], alpha=0.25, label=entry["name"])
-        ax.plot(angles, values, color=sns.color_palette("husl", len(data))[idx], linewidth=2, linestyle='solid')
-
-    ax.set_yticklabels([])
-    ax.set_xticks(angles[:-1])
-    ax.set_xticklabels(labels)
-    plt.title("Radar Chart of Similarity Metrics")
-    plt.legend(loc='upper right', bbox_to_anchor=(1.3, 1.1))
-    return fig
-    
 with col2:
 
     # Display radar chart
     if st.session_state.radar_chart_data:
         st.subheader("Metrics Comparison")
         st.markdown("*Larger area = More similarity of responses.*")
-
-        labels = ["paraphrasing_similarity", "direct_text_comparison_similarity", "summarization_similarity"]
-        fig = plot_radar_chart(st.session_state.radar_chart_data, labels)
+        labels = ["Context similarity", "Words Similarity", "Summarization Similarity"]
+        num_vars = len(labels)
+        angles = np.linspace(0, 2 * np.pi, num_vars, endpoint=False).tolist()
+        angles += angles[:1]
+
+        fig, ax = plt.subplots(figsize=(6, 6), subplot_kw=dict(polar=True))
+
+        # Plot each response with a different color
+        color_palette = sns.color_palette("husl", len(st.session_state.radar_chart_data))
+        for idx, data in enumerate(st.session_state.radar_chart_data):
+            values = [
+                data["paraphrasing_similarity"],
+                data["direct_text_comparison_similarity"],
+                data["summarization_similarity"]
+            ]
+            values += values[:1]
+            ax.fill(angles, values, color=color_palette[idx], alpha=0.25, label=data["name"])
+            ax.plot(angles, values, color=color_palette[idx], linewidth=2, linestyle='solid')
+
+        ax.set_yticklabels([])
+        ax.set_xticks(angles[:-1])
+        ax.set_xticklabels(labels)
+        plt.title("Radar Chart of Similarity Metrics")
+        plt.legend(loc='upper right', bbox_to_anchor=(1.3, 1.1))
         st.pyplot(fig)
 
-        # Display metrics sliders beside the radar chart
-        st.subheader("Similarity Factors")
-        st.markdown("*100 being the best case*")
-        metrics = st.session_state.radar_chart_data[-1]
-        for metric_name in ["paraphrasing_similarity", "direct_text_comparison_similarity", "summarization_similarity"]:
-            st.slider(
-                slider_labels[metric_name],
-                0, 100,
-                int(metrics[metric_name]),
-                key=metric_name,
-                disabled=True,  # Make the slider non-editable
-                format="%.0f"  # Format the slider value to be an integer
-            )
-
     # Display metrics sliders beside the radar chart
     if st.session_state.radar_chart_data:
         st.subheader("Similarity Factors")
@@ -434,4 +450,4 @@ if st.sidebar.button("About"):
     - Scikit-learn
     - NLTK
     - Gensim
-    """)
+    """)