Spaces:
Sleeping
Sleeping
| 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 | |
| ) |