update

ui/roberta_screen.py

@@ -0,0 +1,125 @@
+"""
+UI components for RoBERTa sentiment analysis screen
+"""
+import gradio as gr
+from processors.roberta_analysis import compare_sentiment_roberta
+
+def create_roberta_screen():
+    """
+    Create the RoBERTa sentiment analysis screen components
+    
+    Returns:
+        tuple: (run_roberta_btn, roberta_output, other components...)
+    """
+    with gr.Column() as roberta_screen:
+        gr.Markdown("## RoBERTa Sentiment Analysis")
+        gr.Markdown("""
+        This tab uses the RoBERTa transformer model to perform sentiment analysis on the LLM responses
+        and compare their emotional tones. RoBERTa was trained on a diverse dataset and can detect subtle
+        differences in sentiment that simpler rule-based classifiers might miss.
+        
+        Click 'Run Sentiment Analysis' to analyze your dataset.
+        """)
+        
+        with gr.Row():
+            with gr.Column(scale=1):
+                # Analysis options (expandable in future)
+                with gr.Accordion("Advanced Options", open=False):
+                    sentence_level = gr.Checkbox(
+                        value=True,
+                        label="Perform sentence-level analysis",
+                        info="Analyze sentiment at the sentence level for more detailed results"
+                    )
+                    
+                    visualization_style = gr.Radio(
+                        choices=["Standard", "Detailed"],
+                        value="Standard",
+                        label="Visualization Style",
+                        info="Standard is more concise, Detailed shows more information"
+                    )
+            
+            with gr.Column(scale=1):
+                # Run button in its own column for better visibility
+                run_roberta_btn = gr.Button("Run Sentiment Analysis", variant="primary", size="large")
+        
+        # Hidden output to store raw analysis results
+        roberta_output = gr.JSON(label="Sentiment Analysis Results", visible=False)
+        
+        # Pre-create visualization components (all initially hidden)
+        visualization_container = gr.Markdown(visible=False)
+        status_message = gr.Markdown(visible=False)
+        
+    return run_roberta_btn, roberta_output, sentence_level, visualization_style, visualization_container, status_message
+
+def process_roberta_request(dataset, sentence_level=True, visualization_style="Standard"):
+    """
+    Process the RoBERTa sentiment analysis request
+    
+    Args:
+        dataset (dict): The input dataset
+        sentence_level (bool): Whether to perform sentence-level analysis
+        visualization_style (str): Visualization style preference
+        
+    Returns:
+        dict: Analysis results
+    """
+    if not dataset or "entries" not in dataset or not dataset["entries"]:
+        return {"error": "No dataset loaded. Please create or load a dataset first."}
+    
+    # Initialize the results structure
+    results = {"analyses": {}}
+    
+    # Get the prompt text from the first entry
+    prompt_text = dataset["entries"][0].get("prompt", "")
+    if not prompt_text:
+        return {"error": "No prompt found in dataset"}
+    
+    # Initialize the analysis container for this prompt
+    results["analyses"][prompt_text] = {}
+    
+    # Get model names and responses
+    model1_name = dataset["entries"][0].get("model", "Model 1")
+    model2_name = dataset["entries"][1].get("model", "Model 2")
+    
+    model1_response = dataset["entries"][0].get("response", "")
+    model2_response = dataset["entries"][1].get("response", "")
+    
+    try:
+        # Perform RoBERTa sentiment analysis
+        print("Starting RoBERTa sentiment analysis...")
+        
+        sentiment_results = compare_sentiment_roberta(
+            texts=[model1_response, model2_response],
+            model_names=[model1_name, model2_name]
+        )
+        
+        # Store the results
+        results["analyses"][prompt_text]["roberta_sentiment"] = sentiment_results
+        
+        # Add metadata about the analysis
+        results["analyses"][prompt_text]["roberta_sentiment"]["metadata"] = {
+            "sentence_level": sentence_level,
+            "visualization_style": visualization_style
+        }
+        
+        return results
+    
+    except Exception as e:
+        import traceback
+        error_msg = f"Error in RoBERTa sentiment analysis: {str(e)}\n{traceback.format_exc()}"
+        print(error_msg)
+        
+        # Return error information
+        return {
+            "error": str(e),
+            "traceback": traceback.format_exc(),
+            "analyses": {
+                prompt_text: {
+                    "roberta_sentiment": {
+                        "error": str(e),
+                        "models": [model1_name, model2_name],
+                        "message": "RoBERTa sentiment analysis failed. Try again or use a different analysis method."
+                    }
+                }
+            }
+        }