Update app.py

app.py

@@ -145,7 +145,7 @@ def retrieve_context(question, contexts):
 def generate_answer(question, context, ground, do_pretrained, do_natural, do_pretrained_natural):
     contexts = chunk_splitter(clean_data(context))
     retrieved_context = retrieve_context(question, contexts)
-    ground_in_context = ground in retrieved_context
+    ground_in_context = (retrieved_context.find(ground) != -1)
     
     # Combine question and context
     input_text = f"question: {question} context: {retrieved_context}"
@@ -183,7 +183,7 @@ def generate_answer(question, context, ground, do_pretrained, do_natural, do_pre
     if do_pretrained_natural:
         pretrained_paraphrased_answer = paraphrase_answer(question, generated_answer, True)
 
-    return generated_answer, paraphrased_answer, ground_in_context, pretrained_answer, pretrained_paraphrased_answer, retrieved_context
+    return generated_answer, paraphrased_answer, ("Yes" if ground_in_context else "No"), pretrained_answer, pretrained_paraphrased_answer, retrieved_context
 
 
 # Define a function to list examples from the dataset
@@ -211,7 +211,7 @@ iface = gr.Interface(
     outputs=[
         Textbox(label="Generated Answer"),
         Textbox(label="Natural Answer"),
-        Checkbox(label="Ground truth in the retrieved context"),
+        Textbox(label="Ground truth in the retrieved context"),
         Textbox(label="Pretrained Model's Answer"),
         Textbox(label="Pretrained Model's Natural Answer"),
         Textbox(label="Retrieved Context")