Update app.py

app.py

@@ -73,7 +73,27 @@ def plot_radar_chart(ragas_results):
     plt.close(fig)
     return chart_path
 
-def generate_word_report(science_goal, ragas_results, radar_chart_path):
+def interpret_ragas_results_with_gpt(formatted_scores: list, llm) -> str:
+    
+    if not formatted_scores or not isinstance(formatted_scores[0], dict):
+        return "Invalid RAGAS scores provided."
+
+    score_text = "\n".join([f"{k}: {v}" for k, v in formatted_scores[0].items()])
+
+    prompt = f"""
+You are an expert in RAGAS evaluation metrics to evaluate AI-generated content. Based on the following RAGAS evaluation scores, provide a concise interpretation of each of the metric for the evaluation of AI-generated text. Write in a professional, clear, and objective tone.
+
+RAGAS Scores:
+{score_text}
+
+Provide a paragraph-style interpretation.
+"""
+
+    response = llm.invoke(prompt)
+    return response.content.strip()
+
+
+def generate_word_report(science_goal, ragas_results, radar_chart_path, interpretation):
     doc = Document()
     doc.add_heading("SCDD Evaluation Report", 0)
 
@@ -87,6 +107,9 @@ def generate_word_report(science_goal, ragas_results, radar_chart_path):
     doc.add_heading("RAGAS Metrics Chart", level=1)
     doc.add_picture(radar_chart_path, width=Inches(5))
 
+    doc.add_heading("GPT Interpretation", level=1)
+    doc.add_paragraph(interpretation)
+    
     output_path = "SCDD_Evaluation_Report.docx"
     doc.save(output_path)
     return output_path
@@ -130,9 +153,10 @@ def evaluate_scdd(ai_scdd_file, human_scdd_file, user_input):
     
     formatted_scores = format_ragas_results(ragas_result)
     radar_chart_path = plot_radar_chart(ragas_result)
-    word_report_path = generate_word_report(user_input, ragas_result, radar_chart_path)
+    interpretation = interpret_ragas_results_with_gpt(formatted_scores, llm)
+    word_report_path = generate_word_report(user_input, ragas_result, radar_chart_path, interpretation)
 
-    return formatted_scores, radar_chart_path, word_report_path
+    return formatted_scores, radar_chart_path, interpretation, word_report_path
 
 
 # ----- Gradio Interface -----
@@ -147,6 +171,7 @@ interface = gr.Interface(
     outputs=[
         gr.JSON(label="RAGAS Scores"),
         gr.Image(label="RAGAS Metrics Radar Chart"),
+        gr.Textbox(label="GPT Interpretation of RAGAS Results"),
         gr.File(label="Download Word Report")
     ],
     title="RAGAS Evaluation: AI vs Human SCDD",