update

.idea/workspace.xml

@@ -5,13 +5,29 @@
   </component>
   <component name="ChangeListManager">
     <list default="true" id="8e67814c-7f04-433c-ab7a-2b65a1106d4c" name="Changes" comment="">
-      <change beforePath="$PROJECT_DIR$/processors/topic_modeling.py" beforeDir="false" afterPath="$PROJECT_DIR$/processors/topic_modeling.py" afterDir="false" />
+      <change afterPath="$PROJECT_DIR$/analysis_runner.py" afterDir="false" />
+      <change afterPath="$PROJECT_DIR$/data_handler.py" afterDir="false" />
+      <change afterPath="$PROJECT_DIR$/processors/bias_processor.py" afterDir="false" />
+      <change afterPath="$PROJECT_DIR$/processors/bow_processor.py" afterDir="false" />
+      <change afterPath="$PROJECT_DIR$/processors/classifier_processor.py" afterDir="false" />
+      <change afterPath="$PROJECT_DIR$/processors/ngram_processor.py" afterDir="false" />
+      <change afterPath="$PROJECT_DIR$/processors/topic_processor.py" afterDir="false" />
+      <change afterPath="$PROJECT_DIR$/visualization_handler.py" afterDir="false" />
+      <change beforePath="$PROJECT_DIR$/.idea/workspace.xml" beforeDir="false" afterPath="$PROJECT_DIR$/.idea/workspace.xml" afterDir="false" />
+      <change beforePath="$PROJECT_DIR$/app.py" beforeDir="false" afterPath="$PROJECT_DIR$/app.py" afterDir="false" />
     </list>
     <option name="SHOW_DIALOG" value="false" />
     <option name="HIGHLIGHT_CONFLICTS" value="true" />
     <option name="HIGHLIGHT_NON_ACTIVE_CHANGELIST" value="false" />
     <option name="LAST_RESOLUTION" value="IGNORE" />
   </component>
+  <component name="FileTemplateManagerImpl">
+    <option name="RECENT_TEMPLATES">
+      <list>
+        <option value="Jupyter Notebook" />
+      </list>
+    </option>
+  </component>
   <component name="Git.Settings">
     <option name="RECENT_GIT_ROOT_PATH" value="$PROJECT_DIR$" />
   </component>
@@ -55,7 +71,7 @@
       <option name="presentableId" value="Default" />
       <updated>1745170754325</updated>
       <workItem from="1745170755404" duration="245000" />
-      <workItem from="1745172030020" duration="11940000" />
+      <workItem from="1745172030020" duration="14852000" />
     </task>
     <servers />
   </component>

analysis_runner.py

@@ -0,0 +1,196 @@
+import logging
+import traceback
+import gradio as gr
+from processors.bow_processor import process_bow_analysis
+from processors.ngram_processor import process_ngram_analysis
+from processors.topic_processor import process_topic_modeling
+from processors.classifier_processor import process_classifier_analysis
+from processors.bias_processor import process_bias_detection
+
+# Set up logging
+logger = logging.getLogger('gradio_app.analysis_runner')
+
+# Try to use the improved version of process_analysis_request if available
+try:
+    from improved_analysis_handler import process_analysis_request
+
+    logger.info("Using improved analysis handler")
+except ImportError:
+    logger.info("Using original analysis handler")
+    from ui.analysis_screen import process_analysis_request
+
+
+def run_analysis(dataset, selected_analysis, ngram_n, ngram_top, topic_count):
+    """
+    Run the selected analysis on the provided dataset and return visualization components
+
+    Args:
+        dataset (dict): The dataset to analyze
+        selected_analysis (str): The type of analysis to run
+        ngram_n (int): N-gram size for N-gram analysis
+        ngram_top (int): Number of top N-grams to display
+        topic_count (int): Number of topics for topic modeling
+
+    Returns:
+        tuple: Updated visualization components
+    """
+    try:
+        # Create parameter dictionary
+        parameters = {
+            "ngram_n": ngram_n,
+            "ngram_top": ngram_top,
+            "topic_count": topic_count
+        }
+
+        logger.info(f"Running analysis with selected type: {selected_analysis}")
+        logger.info(f"Parameters: {parameters}")
+
+        if not dataset or "entries" not in dataset or not dataset["entries"]:
+            return default_no_dataset()
+
+        # Process the analysis request - passing selected_analysis as a string
+        analysis_results, _ = process_analysis_request(dataset, selected_analysis, parameters)
+
+        # If there's an error or no results
+        if not analysis_results or "analyses" not in analysis_results or not analysis_results["analyses"]:
+            return default_no_results()
+
+        # Extract information to display in components
+        prompt = list(analysis_results["analyses"].keys())[0]
+        analyses = analysis_results["analyses"][prompt]
+
+        # Check for messages from placeholder analyses
+        if "message" in analyses:
+            return default_message_response(analyses["message"])
+
+        # Route to the appropriate processor based on analysis type
+        if selected_analysis == "Bag of Words" and "bag_of_words" in analyses:
+            return process_bow_analysis(analysis_results, prompt, analyses)
+
+        elif selected_analysis == "N-gram Analysis" and "ngram_analysis" in analyses:
+            return process_ngram_analysis(analysis_results, prompt, analyses)
+
+        elif selected_analysis == "Topic Modeling" and "topic_modeling" in analyses:
+            return process_topic_modeling(analysis_results, prompt, analyses)
+
+        elif selected_analysis == "Classifier" and "classifier" in analyses:
+            return process_classifier_analysis(analysis_results, prompt, analyses)
+
+        elif selected_analysis == "Bias Detection" and "bias_detection" in analyses:
+            return process_bias_detection(analysis_results, prompt, analyses)
+
+        # If we don't have visualization data from any analysis
+        return default_no_visualization(analysis_results)
+
+    except Exception as e:
+        error_msg = f"Error in analysis: {str(e)}\n{traceback.format_exc()}"
+        logger.error(error_msg)
+
+        return default_error_response(error_msg, str(e))
+
+
+def default_no_dataset():
+    """Return default component updates when no dataset is loaded"""
+    return (
+        {},  # analysis_results_state
+        False,  # analysis_output visibility
+        False,  # visualization_area_visible
+        gr.update(visible=False),  # analysis_title
+        gr.update(visible=False),  # prompt_title
+        gr.update(visible=False),  # models_compared
+        gr.update(visible=False),  # model1_title
+        gr.update(visible=False),  # model1_words
+        gr.update(visible=False),  # model2_title
+        gr.update(visible=False),  # model2_words
+        gr.update(visible=False),  # similarity_metrics_title
+        gr.update(visible=False),  # similarity_metrics
+        True,  # status_message_visible
+        gr.update(visible=True, value="❌ **Error:** No dataset loaded. Please create or load a dataset first."),
+        # status_message
+        gr.update(visible=False)  # bias_visualizations
+    )
+
+
+def default_no_results():
+    """Return default component updates when no analysis results are found"""
+    return (
+        {},  # analysis_results_state
+        False,  # analysis_output visibility
+        False,  # visualization_area_visible
+        gr.update(visible=False),  # analysis_title
+        gr.update(visible=False),  # prompt_title
+        gr.update(visible=False),  # models_compared
+        gr.update(visible=False),  # model1_title
+        gr.update(visible=False),  # model1_words
+        gr.update(visible=False),  # model2_title
+        gr.update(visible=False),  # model2_words
+        gr.update(visible=False),  # similarity_metrics_title
+        gr.update(visible=False),  # similarity_metrics
+        True,  # status_message_visible
+        gr.update(visible=True, value="❌ **No results found.** Try a different analysis option."),  # status_message
+        gr.update(visible=False)  # bias_visualizations
+    )
+
+
+def default_message_response(message):
+    """Return default component updates for a simple message response"""
+    return (
+        {},  # analysis_results_state
+        False,  # analysis_output visibility
+        False,  # visualization_area_visible
+        gr.update(visible=False),  # analysis_title
+        gr.update(visible=False),  # prompt_title
+        gr.update(visible=False),  # models_compared
+        gr.update(visible=False),  # model1_title
+        gr.update(visible=False),  # model1_words
+        gr.update(visible=False),  # model2_title
+        gr.update(visible=False),  # model2_words
+        gr.update(visible=False),  # similarity_metrics_title
+        gr.update(visible=False),  # similarity_metrics
+        True,  # status_message_visible
+        gr.update(visible=True, value=f"ℹ️ **{message}**"),  # status_message
+        gr.update(visible=False)  # bias_visualizations
+    )
+
+
+def default_no_visualization(analysis_results):
+    """Return a default set of component updates when no visualization can be shown"""
+    return (
+        analysis_results,
+        False,
+        False,
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        True,
+        gr.update(visible=True,
+                  value="❌ **No visualization data found.** Make sure to select a valid analysis option."),
+        gr.update(visible=False)  # bias_visualizations - Hide it
+    )
+
+
+def default_error_response(error_msg, error_summary):
+    """Return default component updates for an error response"""
+    return (
+        {"error": error_msg},  # analysis_results_state
+        True,  # analysis_output visibility (show raw JSON for debugging)
+        False,  # visualization_area_visible
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        gr.update(visible=False),
+        True,  # status_message_visible
+        gr.update(visible=True, value=f"❌ **Error during analysis:**\n\n```\n{error_summary}\n```"),  # status_message
+        gr.update(visible=False)  # bias_visualizations - Hide it during errors
+    )

app.py

@@ -1,29 +1,10 @@
 import gradio as gr
+import logging
+from data_handler import download_nltk_resources
+from analysis_runner import run_analysis
+from visualization_handler import create_visualization_components
 from ui.dataset_input import create_dataset_input, load_example_dataset
 from ui.analysis_screen import create_analysis_screen
-from visualization.bow_visualizer import process_and_visualize_analysis
-from visualization.bias_visualizer import process_and_visualize_bias_analysis
-import nltk
-import os
-import logging
-import sys
-import traceback
-
-def create_bias_visualization_tab(analysis_results):
-    """
-    Create a specialized tab for bias visualization
-    
-    Args:
-        analysis_results (dict): The analysis results
-        
-    Returns:
-        gradio.Tab: A tab containing the bias visualization
-    """
-    with gr.Tab("Bias Analysis") as tab:
-        # Let the bias visualizer create all needed components
-        components = process_and_visualize_bias_analysis(analysis_results)
-        
-    return tab
 
 # Set up logging
 logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
@@ -33,58 +14,17 @@ logger = logging.getLogger('gradio_app')
 # Try to use the improved version if available, otherwise use original
 try:
     from improved_analysis_handler import process_analysis_request
+
     logger.info("Using improved analysis handler")
 except ImportError:
     logger.info("Using original analysis handler")
     from ui.analysis_screen import process_analysis_request
 
-# Download necessary NLTK resources
-def download_nltk_resources():
-    """Download required NLTK resources if not already downloaded"""
-    try:
-        # Create nltk_data directory in the user's home directory if it doesn't exist
-        nltk_data_path = os.path.expanduser("~/nltk_data")
-        os.makedirs(nltk_data_path, exist_ok=True)
-        
-        # Add this path to NLTK's data path
-        nltk.data.path.append(nltk_data_path)
-        
-        # Download required resources
-        resources = ['punkt', 'wordnet', 'stopwords', 'vader_lexicon']
-        for resource in resources:
-            try:
-                # Different resources can be in different directories in NLTK
-                locations = [
-                    f'tokenizers/{resource}',
-                    f'corpora/{resource}',
-                    f'taggers/{resource}',
-                    f'{resource}'
-                ]
-                
-                found = False
-                for location in locations:
-                    try:
-                        nltk.data.find(location)
-                        logger.info(f"Resource {resource} already downloaded")
-                        found = True
-                        break
-                    except LookupError:
-                        continue
-                
-                if not found:
-                    logger.info(f"Downloading {resource}...")
-                    nltk.download(resource, quiet=True)
-            except Exception as e:
-                logger.error(f"Error with resource {resource}: {e}")
-        
-        logger.info("NLTK resources check completed")
-    except Exception as e:
-        logger.error(f"Error downloading NLTK resources: {e}")
 
 def create_app():
     """
-    Create a streamlined Gradio app for dataset input and analysis. 
-    
+    Create a streamlined Gradio app for dataset input and analysis.
+
     Returns:
         gr.Blocks: The Gradio application
     """
@@ -92,14 +32,14 @@ def create_app():
         # Application state to share data between tabs
         dataset_state = gr.State({})
         analysis_results_state = gr.State({})
-        
+
         # Dataset Input Tab
         with gr.Tab("Dataset Input"):
             dataset_inputs, example_dropdown, load_example_btn, create_btn, prompt, response1, model1, response2, model2 = create_dataset_input()
-            
+
             # Add status indicator to show when dataset is created
             dataset_status = gr.Markdown("*No dataset loaded*")
-            
+
             # Load example dataset
             load_example_btn.click(
                 fn=load_example_dataset,
@@ -111,7 +51,7 @@ def create_app():
             def create_dataset(p, r1, m1, r2, m2):
                 if not p or not r1 or not r2:
                     return {}, "❌ **Error:** Please fill in at least the prompt and both responses"
-                
+
                 dataset = {
                     "entries": [
                         {"prompt": p, "response": r1, "model": m1 or "Model 1"},
@@ -119,16 +59,16 @@ def create_app():
                     ]
                 }
                 return dataset, "✅ **Dataset created successfully!** You can now go to the Analysis tab"
-                
+
             create_btn.click(
                 fn=create_dataset,
                 inputs=[prompt, response1, model1, response2, model2],
                 outputs=[dataset_state, dataset_status]
             )
-        
+
         # Analysis Tab
         with gr.Tab("Analysis"):
-            # Fix the value unpacking to match the actual return values from create_analysis_screen()
+            # Create analysis screen
             analysis_components = create_analysis_screen()
             analysis_options = analysis_components[0]
             analysis_params = analysis_components[1]
@@ -137,555 +77,27 @@ def create_app():
             ngram_n = analysis_components[4]
             ngram_top = analysis_components[5]
             topic_count = analysis_components[6]
-            
-            # Pre-create visualization components (initially hidden)
-            visualization_area_visible = gr.Checkbox(value=False, visible=False, label="Visualization Visible")
-            analysis_title = gr.Markdown("## Analysis Results", visible=False)
-            prompt_title = gr.Markdown(visible=False)
-            models_compared = gr.Markdown(visible=False)
-            
-            # Container for model 1 words
-            model1_title = gr.Markdown(visible=False)
-            model1_words = gr.Markdown(visible=False)
-            
-            # Container for model 2 words
-            model2_title = gr.Markdown(visible=False)
-            model2_words = gr.Markdown(visible=False)
-            
-            # Similarity metrics
-            similarity_metrics_title = gr.Markdown("### Similarity Metrics", visible=False)
-            similarity_metrics = gr.Markdown(visible=False)
-            
-            # Status or error message area
-            status_message_visible = gr.Checkbox(value=False, visible=False, label="Status Message Visible")
-            status_message = gr.Markdown(visible=False)
-
-            # Create bias visualization container (initially hidden)
-            with gr.Column(visible=False) as bias_visualizations:
-                gr.Markdown("### Bias Analysis Visualizations")
-                # This will be populated dynamically
-            
-            # Define a helper function to extract parameter values and run the analysis
-            def run_analysis(dataset, selected_analysis, ngram_n, ngram_top, topic_count):
-                try:
-                    # Create parameter dictionary
-                    parameters = {
-                        "ngram_n": ngram_n,
-                        "ngram_top": ngram_top,
-                        "topic_count": topic_count
-                    }
-                    
-                    logger.info(f"Running analysis with selected type: {selected_analysis}")
-                    logger.info(f"Parameters: {parameters}")
-                    
-                    if not dataset or "entries" not in dataset or not dataset["entries"]:
-                        return (
-                            {},  # analysis_results_state
-                            False,  # analysis_output visibility
-                            False,  # visualization_area_visible
-                            gr.update(visible=False),  # analysis_title
-                            gr.update(visible=False),  # prompt_title
-                            gr.update(visible=False),  # models_compared
-                            gr.update(visible=False),  # model1_title
-                            gr.update(visible=False),  # model1_words
-                            gr.update(visible=False),  # model2_title
-                            gr.update(visible=False),  # model2_words
-                            gr.update(visible=False),  # similarity_metrics_title
-                            gr.update(visible=False),  # similarity_metrics
-                            True,  # status_message_visible
-                            gr.update(visible=True, value="❌ **Error:** No dataset loaded. Please create or load a dataset first."),  # status_message
-                            gr.update(visible=False)  # bias_visualizations
-                        )
-                    
-                    # Process the analysis request - passing selected_analysis as a string
-                    analysis_results, _ = process_analysis_request(dataset, selected_analysis, parameters)
-                    
-                    # If there's an error or no results
-                    if not analysis_results or "analyses" not in analysis_results or not analysis_results["analyses"]:
-                        return (
-                            analysis_results,
-                            False,
-                            False,
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            True,
-                            gr.update(visible=True, value="❌ **No results found.** Try a different analysis option."),
-                            gr.update(visible=False)  # bias_visualizations
-                        )
-                    
-                    # Extract information to display in components
-                    prompt = list(analysis_results["analyses"].keys())[0]
-                    analyses = analysis_results["analyses"][prompt]
-                    
-                    # Initialize visualization components visibilities and contents
-                    visualization_area_visible = False
-                    prompt_title_visible = False
-                    prompt_title_value = ""
-                    models_compared_visible = False
-                    models_compared_value = ""
-                    
-                    model1_title_visible = False
-                    model1_title_value = ""
-                    model1_words_visible = False
-                    model1_words_value = ""
-                    
-                    model2_title_visible = False
-                    model2_title_value = ""
-                    model2_words_visible = False
-                    model2_words_value = ""
-                    
-                    similarity_title_visible = False
-                    similarity_metrics_visible = False
-                    similarity_metrics_value = ""
-                    
-                    # Check for messages from placeholder analyses
-                    if "message" in analyses:
-                        return (
-                            analysis_results,
-                            False,
-                            False,
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            True,
-                            gr.update(visible=True, value=f"ℹ️ **{analyses['message']}**"),
-                            gr.update(visible=False)  # bias_visualizations
-                        )
-                    
-                    # Process based on the selected analysis type
-                    if selected_analysis == "Bag of Words" and "bag_of_words" in analyses:
-                        visualization_area_visible = True
-                        bow_results = analyses["bag_of_words"]
-                        models = bow_results.get("models", [])
-                        
-                        if len(models) >= 2:
-                            prompt_title_visible = True
-                            prompt_title_value = f"## Analysis of Prompt: \"{prompt[:100]}...\""
-                            
-                            models_compared_visible = True
-                            models_compared_value = f"### Comparing responses from {models[0]} and {models[1]}"
-                            
-                            # Extract and format information for display
-                            model1_name = models[0]
-                            model2_name = models[1]
-                            
-                            # Format important words for each model
-                            important_words = bow_results.get("important_words", {})
-                            
-                            if model1_name in important_words:
-                                model1_title_visible = True
-                                model1_title_value = f"#### Top Words Used by {model1_name}"
-                                
-                                word_list = [f"**{item['word']}** ({item['count']})" for item in important_words[model1_name][:10]]
-                                model1_words_visible = True
-                                model1_words_value = ", ".join(word_list)
-                            
-                            if model2_name in important_words:
-                                model2_title_visible = True
-                                model2_title_value = f"#### Top Words Used by {model2_name}"
-                                
-                                word_list = [f"**{item['word']}** ({item['count']})" for item in important_words[model2_name][:10]]
-                                model2_words_visible = True
-                                model2_words_value = ", ".join(word_list)
-                            
-                            # Format similarity metrics
-                            comparisons = bow_results.get("comparisons", {})
-                            comparison_key = f"{model1_name} vs {model2_name}"
 
-                            if comparison_key in comparisons:
-                                metrics = comparisons[comparison_key]
-                                cosine = metrics.get("cosine_similarity", 0)
-                                jaccard = metrics.get("jaccard_similarity", 0)
-                                semantic = metrics.get("semantic_similarity", 0)
-                                common_words = metrics.get("common_word_count", 0)
-                                
-                                similarity_title_visible = True
-                                similarity_metrics_visible = True
-                                similarity_metrics_value = f"""
-                                - **Cosine Similarity**: {cosine:.2f} (higher means more similar word frequency patterns)
-                                - **Jaccard Similarity**: {jaccard:.2f} (higher means more word overlap)
-                                - **Semantic Similarity**: {semantic:.2f} (higher means more similar meaning)
-                                - **Common Words**: {common_words} words appear in both responses
-                                """
-                                
-                    # Check for N-gram analysis
-                    elif selected_analysis == "N-gram Analysis" and "ngram_analysis" in analyses:
-                        visualization_area_visible = True
-                        ngram_results = analyses["ngram_analysis"]
-                        models = ngram_results.get("models", [])
-                        ngram_size = ngram_results.get("ngram_size", 2)
-                        size_name = "Unigrams" if ngram_size == 1 else f"{ngram_size}-grams"
-                        
-                        if len(models) >= 2:
-                            prompt_title_visible = True
-                            prompt_title_value = f"## Analysis of Prompt: \"{prompt[:100]}...\""
-                            
-                            models_compared_visible = True
-                            models_compared_value = f"### {size_name} Analysis: Comparing responses from {models[0]} and {models[1]}"
-                            
-                            # Extract and format information for display
-                            model1_name = models[0]
-                            model2_name = models[1]
-                            
-                            # Format important n-grams for each model
-                            important_ngrams = ngram_results.get("important_ngrams", {})
-                            
-                            if model1_name in important_ngrams:
-                                model1_title_visible = True
-                                model1_title_value = f"#### Top {size_name} Used by {model1_name}"
-                                
-                                ngram_list = [f"**{item['ngram']}** ({item['count']})" for item in important_ngrams[model1_name][:10]]
-                                model1_words_visible = True
-                                model1_words_value = ", ".join(ngram_list)
-                            
-                            if model2_name in important_ngrams:
-                                model2_title_visible = True
-                                model2_title_value = f"#### Top {size_name} Used by {model2_name}"
-                                
-                                ngram_list = [f"**{item['ngram']}** ({item['count']})" for item in important_ngrams[model2_name][:10]]
-                                model2_words_visible = True
-                                model2_words_value = ", ".join(ngram_list)
-                            
-                            # Format similarity metrics if available
-                            if "comparisons" in ngram_results:
-                                comparison_key = f"{model1_name} vs {model2_name}"
-                                
-                                if comparison_key in ngram_results["comparisons"]:
-                                    metrics = ngram_results["comparisons"][comparison_key]
-                                    common_count = metrics.get("common_ngram_count", 0)
-                                    
-                                    similarity_title_visible = True
-                                    similarity_metrics_visible = True
-                                    similarity_metrics_value = f"""
-                                    - **Common {size_name}**: {common_count} {size_name.lower()} appear in both responses
-                                    """
-                    
-                    # Check for Topic Modeling analysis - IMPROVED HANDLING
-                    elif selected_analysis == "Topic Modeling" and "topic_modeling" in analyses:
-                        visualization_area_visible = True
-                        topic_results = analyses["topic_modeling"]
-                        
-                        # Check for errors in topic modeling
-                        if "error" in topic_results:
-                            return (
-                                analysis_results,
-                                False,  # Don't show raw JSON
-                                False,  # Don't show visualization area
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                True,  # Show status message
-                                gr.update(visible=True, value=f"❌ **Topic modeling error:** {topic_results['error']}"),
-                                gr.update(visible=False)  # bias_visualizations
-                            )
-                        
-                        models = topic_results.get("models", [])
-                        method = topic_results.get("method", "lda").upper()
-                        n_topics = topic_results.get("n_topics", 3)
-                        
-                        if len(models) >= 2:
-                            prompt_title_visible = True
-                            prompt_title_value = f"## Analysis of Prompt: \"{prompt[:100]}...\""
-                            
-                            models_compared_visible = True
-                            models_compared_value = f"### Topic Modeling Analysis ({method}, {n_topics} topics)"
-                            
-                            # Extract and format topic information
-                            topics = topic_results.get("topics", [])
-                            
-                            if topics:
-                                # Format topic info for display
-                                topic_info = []
-                                for topic in topics[:5]:  # Show first 5 topics
-                                    topic_id = topic.get("id", 0)
-                                    words = topic.get("words", [])[:5]  # Top 5 words per topic
-                                    
-                                    if words:
-                                        topic_info.append(f"**Topic {topic_id+1}**: {', '.join(words)}")
-                                
-                                if topic_info:
-                                    model1_title_visible = True
-                                    model1_title_value = "#### Discovered Topics"
-                                    model1_words_visible = True
-                                    model1_words_value = "\n".join(topic_info)
-                            
-                            # Get topic distributions for models
-                            model_topics = topic_results.get("model_topics", {})
-                            
-                            if model_topics:
-                                model1_name = models[0]
-                                model2_name = models[1]
-                                
-                                # Format topic distribution info
-                                if model1_name in model_topics and model2_name in model_topics:
-                                    model2_title_visible = True
-                                    model2_title_value = "#### Topic Distribution"
-                                    model2_words_visible = True
-                                    
-                                    # Simple distribution display
-                                    dist1 = model_topics[model1_name]
-                                    dist2 = model_topics[model2_name]
-                                    
-                                    model2_words_value = f"""
-                                    **{model1_name}**: {', '.join([f"Topic {i+1}: {v:.2f}" for i, v in enumerate(dist1[:5])])}
-                                    
-                                    **{model2_name}**: {', '.join([f"Topic {i+1}: {v:.2f}" for i, v in enumerate(dist2[:5])])}
-                                    """
-                                    
-                            # Add similarity metrics if available
-                            comparisons = topic_results.get("comparisons", {})
-                            if comparisons:
-                                comparison_key = f"{model1_name} vs {model2_name}"
-                                
-                                if comparison_key in comparisons:
-                                    metrics = comparisons[comparison_key]
-                                    js_div = metrics.get("js_divergence", 0)
-                                    
-                                    # Add interpretation
-                                    similarity_text = ""
-                                    if js_div < 0.2:
-                                        similarity_text = "very similar"
-                                    elif js_div < 0.4:
-                                        similarity_text = "somewhat similar"
-                                    elif js_div < 0.6:
-                                        similarity_text = "moderately different"
-                                    else:
-                                        similarity_text = "very different"
-                                    
-                                    similarity_title_visible = True
-                                    similarity_metrics_visible = True
-                                    similarity_metrics_value = f"""
-                                    - **Topic Distribution Divergence**: {js_div:.4f}
-                                    - The topic distributions between models are **{similarity_text}**
-                                    - *Lower divergence values indicate more similar topic distributions*
-                                    """
-                    
-                    # Check for Classifier analysis
-                    elif selected_analysis == "Classifier" and "classifier" in analyses:
-                        visualization_area_visible = True
-                        classifier_results = analyses["classifier"]
-                        models = classifier_results.get("models", [])
-                        
-                        if len(models) >= 2:
-                            prompt_title_visible = True
-                            prompt_title_value = f"## Analysis of Prompt: \"{prompt[:100]}...\""
-                            
-                            models_compared_visible = True
-                            models_compared_value = f"### Classifier Analysis for {models[0]} and {models[1]}"
-                            
-                            # Extract and format classifier information
-                            model1_name = models[0]
-                            model2_name = models[1]
-                            
-                            # Display classifications for each model
-                            classifications = classifier_results.get("classifications", {})
-                            
-                            if classifications:
-                                model1_title_visible = True
-                                model1_title_value = f"#### Classification Results"
-                                model1_words_visible = True
-                                
-                                model1_results = classifications.get(model1_name, {})
-                                model2_results = classifications.get(model2_name, {})
-                                
-                                model1_words_value = f"""
-                                **{model1_name}**:
-                                - Formality: {model1_results.get('formality', 'N/A')}
-                                - Sentiment: {model1_results.get('sentiment', 'N/A')}
-                                - Complexity: {model1_results.get('complexity', 'N/A')}
-                                
-                                **{model2_name}**:
-                                - Formality: {model2_results.get('formality', 'N/A')}
-                                - Sentiment: {model2_results.get('sentiment', 'N/A')}
-                                - Complexity: {model2_results.get('complexity', 'N/A')}
-                                """
-                                
-                                # Show comparison
-                                model2_title_visible = True
-                                model2_title_value = f"#### Classification Comparison"
-                                model2_words_visible = True
-                                
-                                differences = classifier_results.get("differences", {})
-                                model2_words_value = "\n".join([
-                                    f"- **{category}**: {diff}" 
-                                    for category, diff in differences.items()
-                                ])
-                    
-                    # Add visualization handling for Bias Detection
-                    elif selected_analysis == "Bias Detection" and "bias_detection" in analyses:
-                        logger.info("Processing Bias Detection visualization")
-                        
-                        models = analyses["bias_detection"].get("models", ["Model 1", "Model 2"])
-                        logger.info(f"Bias models: {models}")
-                        
-                        try:
-                            # Initialize bias_visualizations container
-                            bias_visualizations.clear()
-                            
-                            # Create bias visualization components
-                            from visualization.bias_visualizer import process_and_visualize_bias_analysis
-                            bias_components = process_and_visualize_bias_analysis(analysis_results)
-                            
-                            logger.info(f"Created {len(bias_components)} bias visualization components")
-                            
-                            # Add components to container
-                            for component in bias_components:
-                                bias_visualizations.append(component)
-                            
-                            logger.info("Added components to bias_visualizations")
-                            
-                            return (
-                                analysis_results,  # analysis_results_state
-                                False,  # analysis_output visibility  
-                                True,  # visualization_area_visible
-                                gr.update(visible=True),  # analysis_title
-                                gr.update(visible=True, value=f"## Analysis of Prompt: \"{prompt[:100]}...\""),  # prompt_title
-                                gr.update(visible=True, value=f"### Comparing responses from {models[0]} and {models[1]}"),  # models_compared
-                                gr.update(visible=True, value="#### Bias detection visualization is available below"),  # model1_title
-                                gr.update(visible=True, value="The detailed bias analysis includes sentiment analysis, partisan term detection, and framing analysis."),  # model1_words
-                                gr.update(visible=False),  # model2_title
-                                gr.update(visible=False),  # model2_words
-                                gr.update(visible=False),  # similarity_metrics_title
-                                gr.update(visible=False),  # similarity_metrics
-                                False,  # status_message_visible
-                                gr.update(visible=False),  # status_message
-                                gr.update(visible=True)  # bias_visualizations - Make this visible
-                            )
-                        except Exception as e:
-                            import traceback
-                            logger.error(f"Error generating bias visualization: {str(e)}\n{traceback.format_exc()}")
-                            
-                            return (
-                                analysis_results,
-                                True,  # Show raw JSON for debugging
-                                False,
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                gr.update(visible=False),
-                                True,
-                                gr.update(visible=True, value=f"❌ **Error generating bias visualization:** {str(e)}"),
-                                gr.update(visible=False)  # bias_visualizations
-                            )
+            # Create visualization components
+            visualization_components = create_visualization_components()
 
-                    # If we don't have visualization data from any analysis
-                    if not visualization_area_visible:
-                        return (
-                            analysis_results,
-                            False,
-                            False,
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            gr.update(visible=False),
-                            True,
-                            gr.update(visible=True, value="❌ **No visualization data found.** Make sure to select a valid analysis option."),
-                            gr.update(visible=False)  # bias_visualizations - Hide it
-                        )
-
-                    # For all other analysis types, make sure bias_visualizations is not visible
-                    return (
-                        analysis_results,  # analysis_results_state
-                        False,  # analysis_output visibility
-                        True,   # visualization_area_visible
-                        gr.update(visible=True),  # analysis_title
-                        gr.update(visible=prompt_title_visible, value=prompt_title_value),  # prompt_title
-                        gr.update(visible=models_compared_visible, value=models_compared_value),  # models_compared
-                        gr.update(visible=model1_title_visible, value=model1_title_value),  # model1_title
-                        gr.update(visible=model1_words_visible, value=model1_words_value),  # model1_words
-                        gr.update(visible=model2_title_visible, value=model2_title_value),  # model2_title
-                        gr.update(visible=model2_words_visible, value=model2_words_value),  # model2_words
-                        gr.update(visible=similarity_title_visible),  # similarity_metrics_title
-                        gr.update(visible=similarity_metrics_visible, value=similarity_metrics_value),  # similarity_metrics
-                        False,  # status_message_visible
-                        gr.update(visible=False),  # status_message
-                        gr.update(visible=False)  # bias_visualizations - Not visible for other analyses
-                    )
-                
-                except Exception as e:
-                    import traceback
-                    error_msg = f"Error in analysis: {str(e)}\n{traceback.format_exc()}"
-                    logger.error(error_msg)
-                    
-                    return (
-                        {"error": error_msg},  # analysis_results_state
-                        True,  # analysis_output visibility (show raw JSON for debugging)
-                        False,  # visualization_area_visible
-                        gr.update(visible=False),
-                        gr.update(visible=False),
-                        gr.update(visible=False),
-                        gr.update(visible=False),
-                        gr.update(visible=False),
-                        gr.update(visible=False),
-                        gr.update(visible=False),
-                        gr.update(visible=False),
-                        gr.update(visible=False),
-                        True,  # status_message_visible
-                        gr.update(visible=True, value=f"❌ **Error during analysis:**\n\n```\n{str(e)}\n```"),  # status_message
-                        gr.update(visible=False)  # bias_visualizations - Hide it during errors
-                    )
-        
-        # Connect the run button to the analysis function 
+            # Connect the run button to the analysis function
             run_analysis_btn.click(
                 fn=run_analysis,
                 inputs=[dataset_state, analysis_options, ngram_n, ngram_top, topic_count],
-                outputs=[
-                    analysis_results_state,
-                    analysis_output,
-                    visualization_area_visible,
-                    analysis_title,
-                    prompt_title,
-                    models_compared,
-                    model1_title,
-                    model1_words,
-                    model2_title,
-                    model2_words,
-                    similarity_metrics_title,
-                    similarity_metrics,
-                    status_message_visible,
-                    status_message,
-                    bias_visualizations  # Add this output for bias visualization
-                ]
+                outputs=visualization_components
             )
-    
+
     return app
 
+
 if __name__ == "__main__":
     # Download required NLTK resources before launching the app
     download_nltk_resources()
-    
+
     logger.info("Starting LLM Response Comparator application")
     logger.info("===== Application Startup =====")
-    
+
     # Create and launch the application
     app = create_app()
     app.launch()

data_handler.py

@@ -0,0 +1,49 @@
+import os
+import nltk
+import logging
+
+# Set up logging
+logger = logging.getLogger('gradio_app.data_handler')
+
+
+def download_nltk_resources():
+    """Download required NLTK resources if not already downloaded"""
+    try:
+        # Create nltk_data directory in the user's home directory if it doesn't exist
+        nltk_data_path = os.path.expanduser("~/nltk_data")
+        os.makedirs(nltk_data_path, exist_ok=True)
+
+        # Add this path to NLTK's data path
+        nltk.data.path.append(nltk_data_path)
+
+        # Download required resources
+        resources = ['punkt', 'wordnet', 'stopwords', 'vader_lexicon']
+        for resource in resources:
+            try:
+                # Different resources can be in different directories in NLTK
+                locations = [
+                    f'tokenizers/{resource}',
+                    f'corpora/{resource}',
+                    f'taggers/{resource}',
+                    f'{resource}'
+                ]
+
+                found = False
+                for location in locations:
+                    try:
+                        nltk.data.find(location)
+                        logger.info(f"Resource {resource} already downloaded")
+                        found = True
+                        break
+                    except LookupError:
+                        continue
+
+                if not found:
+                    logger.info(f"Downloading {resource}...")
+                    nltk.download(resource, quiet=True)
+            except Exception as e:
+                logger.error(f"Error with resource {resource}: {e}")
+
+        logger.info("NLTK resources check completed")
+    except Exception as e:
+        logger.error(f"Error downloading NLTK resources: {e}")

processors/bias_processor.py

@@ -0,0 +1,72 @@
+import gradio as gr
+import logging
+import traceback
+
+# Set up logging
+logger = logging.getLogger('gradio_app.processors.bias')
+
+
+def process_bias_detection(analysis_results, prompt, analyses):
+    """
+    Process Bias Detection analysis and return UI updates
+
+    Args:
+        analysis_results (dict): Complete analysis results
+        prompt (str): The prompt being analyzed
+        analyses (dict): Analysis data for the prompt
+
+    Returns:
+        tuple: UI component updates
+    """
+    logger.info("Processing Bias Detection visualization")
+
+    models = analyses["bias_detection"].get("models", ["Model 1", "Model 2"])
+    logger.info(f"Bias models: {models}")
+
+    try:
+        # Create bias visualization components
+        from visualization.bias_visualizer import process_and_visualize_bias_analysis
+        bias_components = process_and_visualize_bias_analysis(analysis_results)
+
+        logger.info(f"Created {len(bias_components)} bias visualization components")
+
+        return (
+            analysis_results,  # analysis_results_state
+            False,  # analysis_output visibility
+            True,  # visualization_area_visible
+            gr.update(visible=True),  # analysis_title
+            gr.update(visible=True, value=f"## Analysis of Prompt: \"{prompt[:100]}...\""),  # prompt_title
+            gr.update(visible=True, value=f"### Comparing responses from {models[0]} and {models[1]}"),
+            # models_compared
+            gr.update(visible=True, value="#### Bias detection visualization is available below"),  # model1_title
+            gr.update(visible=True,
+                      value="The detailed bias analysis includes sentiment analysis, partisan term detection, and framing analysis."),
+            # model1_words
+            gr.update(visible=False),  # model2_title
+            gr.update(visible=False),  # model2_words
+            gr.update(visible=False),  # similarity_metrics_title
+            gr.update(visible=False),  # similarity_metrics
+            False,  # status_message_visible
+            gr.update(visible=False),  # status_message
+            gr.update(visible=True)  # bias_visualizations - Make this visible
+        )
+    except Exception as e:
+        logger.error(f"Error generating bias visualization: {str(e)}\n{traceback.format_exc()}")
+
+        return (
+            analysis_results,
+            True,  # Show raw JSON for debugging
+            False,
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            True,
+            gr.update(visible=True, value=f"❌ **Error generating bias visualization:** {str(e)}"),
+            gr.update(visible=False)  # bias_visualizations
+        )

processors/bow_processor.py

@@ -0,0 +1,107 @@
+import gradio as gr
+import logging
+
+# Set up logging
+logger = logging.getLogger('gradio_app.processors.bow')
+
+
+def process_bow_analysis(analysis_results, prompt, analyses):
+    """
+    Process Bag of Words analysis and return UI updates
+
+    Args:
+        analysis_results (dict): Complete analysis results
+        prompt (str): The prompt being analyzed
+        analyses (dict): Analysis data for the prompt
+
+    Returns:
+        tuple: UI component updates
+    """
+    visualization_area_visible = True
+    bow_results = analyses["bag_of_words"]
+    models = bow_results.get("models", [])
+
+    if len(models) < 2:
+        from analysis_runner import default_no_visualization
+        return default_no_visualization(analysis_results)
+
+    prompt_title_visible = True
+    prompt_title_value = f"## Analysis of Prompt: \"{prompt[:100]}...\""
+
+    models_compared_visible = True
+    models_compared_value = f"### Comparing responses from {models[0]} and {models[1]}"
+
+    # Extract and format information for display
+    model1_name = models[0]
+    model2_name = models[1]
+
+    # Format important words for each model
+    important_words = bow_results.get("important_words", {})
+
+    model1_title_visible = False
+    model1_title_value = ""
+    model1_words_visible = False
+    model1_words_value = ""
+
+    if model1_name in important_words:
+        model1_title_visible = True
+        model1_title_value = f"#### Top Words Used by {model1_name}"
+
+        word_list = [f"**{item['word']}** ({item['count']})" for item in important_words[model1_name][:10]]
+        model1_words_visible = True
+        model1_words_value = ", ".join(word_list)
+
+    model2_title_visible = False
+    model2_title_value = ""
+    model2_words_visible = False
+    model2_words_value = ""
+
+    if model2_name in important_words:
+        model2_title_visible = True
+        model2_title_value = f"#### Top Words Used by {model2_name}"
+
+        word_list = [f"**{item['word']}** ({item['count']})" for item in important_words[model2_name][:10]]
+        model2_words_visible = True
+        model2_words_value = ", ".join(word_list)
+
+    similarity_title_visible = False
+    similarity_metrics_visible = False
+    similarity_metrics_value = ""
+
+    # Format similarity metrics
+    comparisons = bow_results.get("comparisons", {})
+    comparison_key = f"{model1_name} vs {model2_name}"
+
+    if comparison_key in comparisons:
+        metrics = comparisons[comparison_key]
+        cosine = metrics.get("cosine_similarity", 0)
+        jaccard = metrics.get("jaccard_similarity", 0)
+        semantic = metrics.get("semantic_similarity", 0)
+        common_words = metrics.get("common_word_count", 0)
+
+        similarity_title_visible = True
+        similarity_metrics_visible = True
+        similarity_metrics_value = f"""
+        - **Cosine Similarity**: {cosine:.2f} (higher means more similar word frequency patterns)
+        - **Jaccard Similarity**: {jaccard:.2f} (higher means more word overlap)
+        - **Semantic Similarity**: {semantic:.2f} (higher means more similar meaning)
+        - **Common Words**: {common_words} words appear in both responses
+        """
+
+    return (
+        analysis_results,  # analysis_results_state
+        False,  # analysis_output visibility
+        True,  # visualization_area_visible
+        gr.update(visible=True),  # analysis_title
+        gr.update(visible=prompt_title_visible, value=prompt_title_value),  # prompt_title
+        gr.update(visible=models_compared_visible, value=models_compared_value),  # models_compared
+        gr.update(visible=model1_title_visible, value=model1_title_value),  # model1_title
+        gr.update(visible=model1_words_visible, value=model1_words_value),  # model1_words
+        gr.update(visible=model2_title_visible, value=model2_title_value),  # model2_title
+        gr.update(visible=model2_words_visible, value=model2_words_value),  # model2_words
+        gr.update(visible=similarity_title_visible),  # similarity_metrics_title
+        gr.update(visible=similarity_metrics_visible, value=similarity_metrics_value),  # similarity_metrics
+        False,  # status_message_visible
+        gr.update(visible=False),  # status_message
+        gr.update(visible=False)  # bias_visualizations - Not visible for BoW analysis
+    )

processors/classifier_processor.py

@@ -0,0 +1,99 @@
+import gradio as gr
+import logging
+
+# Set up logging
+logger = logging.getLogger('gradio_app.processors.classifier')
+
+
+def process_classifier_analysis(analysis_results, prompt, analyses):
+    """
+    Process Classifier analysis and return UI updates
+
+    Args:
+        analysis_results (dict): Complete analysis results
+        prompt (str): The prompt being analyzed
+        analyses (dict): Analysis data for the prompt
+
+    Returns:
+        tuple: UI component updates
+    """
+    visualization_area_visible = True
+    classifier_results = analyses["classifier"]
+    models = classifier_results.get("models", [])
+
+    if len(models) < 2:
+        from analysis_runner import default_no_visualization
+        return default_no_visualization(analysis_results)
+
+    prompt_title_visible = True
+    prompt_title_value = f"## Analysis of Prompt: \"{prompt[:100]}...\""
+
+    models_compared_visible = True
+    models_compared_value = f"### Classifier Analysis for {models[0]} and {models[1]}"
+
+    # Extract and format classifier information
+    model1_name = models[0]
+    model2_name = models[1]
+
+    # Display classifications for each model
+    classifications = classifier_results.get("classifications", {})
+
+    model1_title_visible = False
+    model1_title_value = ""
+    model1_words_visible = False
+    model1_words_value = ""
+
+    if classifications:
+        model1_title_visible = True
+        model1_title_value = f"#### Classification Results"
+        model1_words_visible = True
+
+        model1_results = classifications.get(model1_name, {})
+        model2_results = classifications.get(model2_name, {})
+
+        model1_words_value = f"""
+        **{model1_name}**:
+        - Formality: {model1_results.get('formality', 'N/A')}
+        - Sentiment: {model1_results.get('sentiment', 'N/A')}
+        - Complexity: {model1_results.get('complexity', 'N/A')}
+
+        **{model2_name}**:
+        - Formality: {model2_results.get('formality', 'N/A')}
+        - Sentiment: {model2_results.get('sentiment', 'N/A')}
+        - Complexity: {model2_results.get('complexity', 'N/A')}
+        """
+
+    # Show comparison
+    model2_title_visible = False
+    model2_title_value = ""
+    model2_words_visible = False
+    model2_words_value = ""
+
+    differences = classifier_results.get("differences", {})
+    if differences:
+        model2_title_visible = True
+        model2_title_value = f"#### Classification Comparison"
+        model2_words_visible = True
+
+        model2_words_value = "\n".join([
+            f"- **{category}**: {diff}"
+            for category, diff in differences.items()
+        ])
+
+    return (
+        analysis_results,  # analysis_results_state
+        False,  # analysis_output visibility
+        True,  # visualization_area_visible
+        gr.update(visible=True),  # analysis_title
+        gr.update(visible=prompt_title_visible, value=prompt_title_value),  # prompt_title
+        gr.update(visible=models_compared_visible, value=models_compared_value),  # models_compared
+        gr.update(visible=model1_title_visible, value=model1_title_value),  # model1_title
+        gr.update(visible=model1_words_visible, value=model1_words_value),  # model1_words
+        gr.update(visible=model2_title_visible, value=model2_title_value),  # model2_title
+        gr.update(visible=model2_words_visible, value=model2_words_value),  # model2_words
+        gr.update(visible=False),  # similarity_metrics_title
+        gr.update(visible=False),  # similarity_metrics
+        False,  # status_message_visible
+        gr.update(visible=False),  # status_message
+        gr.update(visible=False)  # bias_visualizations - Not visible for Classifier analysis
+    )

processors/ngram_processor.py

@@ -0,0 +1,103 @@
+import gradio as gr
+import logging
+
+# Set up logging
+logger = logging.getLogger('gradio_app.processors.ngram')
+
+
+def process_ngram_analysis(analysis_results, prompt, analyses):
+    """
+    Process N-gram analysis and return UI updates
+
+    Args:
+        analysis_results (dict): Complete analysis results
+        prompt (str): The prompt being analyzed
+        analyses (dict): Analysis data for the prompt
+
+    Returns:
+        tuple: UI component updates
+    """
+    visualization_area_visible = True
+    ngram_results = analyses["ngram_analysis"]
+    models = ngram_results.get("models", [])
+    ngram_size = ngram_results.get("ngram_size", 2)
+    size_name = "Unigrams" if ngram_size == 1 else f"{ngram_size}-grams"
+
+    if len(models) < 2:
+        from analysis_runner import default_no_visualization
+        return default_no_visualization(analysis_results)
+
+    prompt_title_visible = True
+    prompt_title_value = f"## Analysis of Prompt: \"{prompt[:100]}...\""
+
+    models_compared_visible = True
+    models_compared_value = f"### {size_name} Analysis: Comparing responses from {models[0]} and {models[1]}"
+
+    # Extract and format information for display
+    model1_name = models[0]
+    model2_name = models[1]
+
+    # Format important n-grams for each model
+    important_ngrams = ngram_results.get("important_ngrams", {})
+
+    model1_title_visible = False
+    model1_title_value = ""
+    model1_words_visible = False
+    model1_words_value = ""
+
+    if model1_name in important_ngrams:
+        model1_title_visible = True
+        model1_title_value = f"#### Top {size_name} Used by {model1_name}"
+
+        ngram_list = [f"**{item['ngram']}** ({item['count']})" for item in important_ngrams[model1_name][:10]]
+        model1_words_visible = True
+        model1_words_value = ", ".join(ngram_list)
+
+    model2_title_visible = False
+    model2_title_value = ""
+    model2_words_visible = False
+    model2_words_value = ""
+
+    if model2_name in important_ngrams:
+        model2_title_visible = True
+        model2_title_value = f"#### Top {size_name} Used by {model2_name}"
+
+        ngram_list = [f"**{item['ngram']}** ({item['count']})" for item in important_ngrams[model2_name][:10]]
+        model2_words_visible = True
+        model2_words_value = ", ".join(ngram_list)
+
+    similarity_title_visible = False
+    similarity_metrics_visible = False
+    similarity_metrics_value = ""
+
+    # Format similarity metrics if available
+    if "comparisons" in ngram_results:
+        comparison_key = f"{model1_name} vs {model2_name}"
+
+        if comparison_key in ngram_results["comparisons"]:
+            metrics = ngram_results["comparisons"][comparison_key]
+            common_count = metrics.get("common_ngram_count", 0)
+
+            similarity_title_visible = True
+            similarity_metrics_visible = True
+            similarity_metrics_value = f"""
+            - **Common {size_name}**: {common_count} {size_name.lower()} appear in both responses
+            """
+
+    return (
+        analysis_results,  # analysis_results_state
+        False,  # analysis_output visibility
+        True,  # visualization_area_visible
+        gr.update(visible=True),  # analysis_title
+        gr.update(visible=prompt_title_visible, value=prompt_title_value),  # prompt_title
+        gr.update(visible=models_compared_visible, value=models_compared_value),  # models_compared
+        gr.update(visible=model1_title_visible, value=model1_title_value),  # model1_title
+        gr.update(visible=model1_words_visible, value=model1_words_value),  # model1_words
+        gr.update(visible=model2_title_visible, value=model2_title_value),  # model2_title
+        gr.update(visible=model2_words_visible, value=model2_words_value),  # model2_words
+        gr.update(visible=similarity_title_visible),  # similarity_metrics_title
+        gr.update(visible=similarity_metrics_visible, value=similarity_metrics_value),  # similarity_metrics
+        False,  # status_message_visible
+        gr.update(visible=False),  # status_message
+        gr.update(visible=False)  # bias_visualizations - Not visible for N-gram analysis
+    )

processors/topic_processor.py

@@ -0,0 +1,156 @@
+import gradio as gr
+import logging
+
+# Set up logging
+logger = logging.getLogger('gradio_app.processors.topic')
+
+
+def process_topic_modeling(analysis_results, prompt, analyses):
+    """
+    Process Topic Modeling analysis and return UI updates
+
+    Args:
+        analysis_results (dict): Complete analysis results
+        prompt (str): The prompt being analyzed
+        analyses (dict): Analysis data for the prompt
+
+    Returns:
+        tuple: UI component updates
+    """
+    topic_results = analyses["topic_modeling"]
+
+    # Check for errors in topic modeling
+    if "error" in topic_results:
+        return (
+            analysis_results,
+            False,  # Don't show raw JSON
+            False,  # Don't show visualization area
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            gr.update(visible=False),
+            True,  # Show status message
+            gr.update(visible=True, value=f"❌ **Topic modeling error:** {topic_results['error']}"),
+            gr.update(visible=False)  # bias_visualizations
+        )
+
+    visualization_area_visible = True
+    models = topic_results.get("models", [])
+    method = topic_results.get("method", "lda").upper()
+    n_topics = topic_results.get("n_topics", 3)
+
+    if len(models) < 2:
+        from analysis_runner import default_no_visualization
+        return default_no_visualization(analysis_results)
+
+    prompt_title_visible = True
+    prompt_title_value = f"## Analysis of Prompt: \"{prompt[:100]}...\""
+
+    models_compared_visible = True
+    models_compared_value = f"### Topic Modeling Analysis ({method}, {n_topics} topics)"
+
+    # Initialize component visibility and values
+    model1_title_visible = False
+    model1_title_value = ""
+    model1_words_visible = False
+    model1_words_value = ""
+    model2_title_visible = False
+    model2_title_value = ""
+    model2_words_visible = False
+    model2_words_value = ""
+    similarity_title_visible = False
+    similarity_metrics_visible = False
+    similarity_metrics_value = ""
+
+    # Extract and format topic information
+    topics = topic_results.get("topics", [])
+
+    if topics:
+        # Format topic info for display
+        topic_info = []
+        for topic in topics[:5]:  # Show first 5 topics
+            topic_id = topic.get("id", 0)
+            words = topic.get("words", [])[:5]  # Top 5 words per topic
+
+            if words:
+                topic_info.append(f"**Topic {topic_id + 1}**: {', '.join(words)}")
+
+        if topic_info:
+            model1_title_visible = True
+            model1_title_value = "#### Discovered Topics"
+            model1_words_visible = True
+            model1_words_value = "\n".join(topic_info)
+
+    # Get topic distributions for models
+    model_topics = topic_results.get("model_topics", {})
+
+    if model_topics:
+        model1_name = models[0]
+        model2_name = models[1]
+
+        # Format topic distribution info
+        if model1_name in model_topics and model2_name in model_topics:
+            model2_title_visible = True
+            model2_title_value = "#### Topic Distribution"
+            model2_words_visible = True
+
+            # Simple distribution display
+            dist1 = model_topics[model1_name]
+            dist2 = model_topics[model2_name]
+
+            model2_words_value = f"""
+            **{model1_name}**: {', '.join([f"Topic {i + 1}: {v:.2f}" for i, v in enumerate(dist1[:5])])}
+
+            **{model2_name}**: {', '.join([f"Topic {i + 1}: {v:.2f}" for i, v in enumerate(dist2[:5])])}
+            """
+
+    # Add similarity metrics if available
+    comparisons = topic_results.get("comparisons", {})
+    if comparisons:
+        comparison_key = f"{model1_name} vs {model2_name}"
+
+        if comparison_key in comparisons:
+            metrics = comparisons[comparison_key]
+            js_div = metrics.get("js_divergence", 0)
+
+            # Add interpretation
+            similarity_text = ""
+            if js_div < 0.2:
+                similarity_text = "very similar"
+            elif js_div < 0.4:
+                similarity_text = "somewhat similar"
+            elif js_div < 0.6:
+                similarity_text = "moderately different"
+            else:
+                similarity_text = "very different"
+
+            similarity_title_visible = True
+            similarity_metrics_visible = True
+            similarity_metrics_value = f"""
+            - **Topic Distribution Divergence**: {js_div:.4f}
+            - The topic distributions between models are **{similarity_text}**
+            - *Lower divergence values indicate more similar topic distributions*
+            """
+
+    return (
+        analysis_results,  # analysis_results_state
+        False,  # analysis_output visibility
+        True,  # visualization_area_visible
+        gr.update(visible=True),  # analysis_title
+        gr.update(visible=prompt_title_visible, value=prompt_title_value),  # prompt_title
+        gr.update(visible=models_compared_visible, value=models_compared_value),  # models_compared
+        gr.update(visible=model1_title_visible, value=model1_title_value),  # model1_title
+        gr.update(visible=model1_words_visible, value=model1_words_value),  # model1_words
+        gr.update(visible=model2_title_visible, value=model2_title_value),  # model2_title
+        gr.update(visible=model2_words_visible, value=model2_words_value),  # model2_words
+        gr.update(visible=similarity_title_visible),  # similarity_metrics_title
+        gr.update(visible=similarity_metrics_visible, value=similarity_metrics_value),  # similarity_metrics
+        False,  # status_message_visible
+        gr.update(visible=False),  # status_message
+        gr.update(visible=False)  # bias_visualizations - Not visible for Topic Modeling
+    )

visualization_handler.py

@@ -0,0 +1,73 @@
+import gradio as gr
+import logging
+
+# Set up logging
+logger = logging.getLogger('gradio_app.visualization_handler')
+
+
+def create_visualization_components():
+    """
+    Creates all the visualization components used in the analysis tab
+
+    Returns:
+        list: A list of all gradio components for visualization
+    """
+    # Pre-create visualization components (initially hidden)
+    visualization_area_visible = gr.Checkbox(value=False, visible=False, label="Visualization Visible")
+    analysis_title = gr.Markdown("## Analysis Results", visible=False)
+    prompt_title = gr.Markdown(visible=False)
+    models_compared = gr.Markdown(visible=False)
+
+    # Container for model 1 words
+    model1_title = gr.Markdown(visible=False)
+    model1_words = gr.Markdown(visible=False)
+
+    # Container for model 2 words
+    model2_title = gr.Markdown(visible=False)
+    model2_words = gr.Markdown(visible=False)
+
+    # Similarity metrics
+    similarity_metrics_title = gr.Markdown("### Similarity Metrics", visible=False)
+    similarity_metrics = gr.Markdown(visible=False)
+
+    # Status or error message area
+    status_message_visible = gr.Checkbox(value=False, visible=False, label="Status Message Visible")
+    status_message = gr.Markdown(visible=False)
+
+    # Create bias visualization container (initially hidden)
+    with gr.Column(visible=False) as bias_visualizations:
+        gr.Markdown("### Bias Analysis Visualizations")
+        # This will be populated dynamically
+
+    # Return all components as a list
+    return [
+        analysis_results_state := gr.State({}),
+        analysis_output := gr.JSON(visible=False),
+        visualization_area_visible,
+        analysis_title,
+        prompt_title,
+        models_compared,
+        model1_title,
+        model1_words,
+        model2_title,
+        model2_words,
+        similarity_metrics_title,
+        similarity_metrics,
+        status_message_visible,
+        status_message,
+        bias_visualizations
+    ]
+
+
+def process_and_visualize_bias_analysis(analysis_results):
+    """
+    Wrapper for bias visualization function from visualization.bias_visualizer
+
+    Args:
+        analysis_results (dict): The analysis results
+
+    Returns:
+        list: Components for bias visualization
+    """
+    from visualization.bias_visualizer import process_and_visualize_bias_analysis
+    return process_and_visualize_bias_analysis(analysis_results)