Update app.py

app.py

@@ -33,7 +33,7 @@ max_target_length = 200
 # Load your dataset
 dataset = datasets.load_dataset("minh21/COVID-QA-Chunk-64-testset-biencoder-data-90_10", split="train")
 dataset = dataset.shuffle()
-dataset = dataset.select(range(10))
+dataset = dataset.select([6, 18, 24, 156, 650, 19, 31, 97, 133, 183])
 
 # Context chunking
 min_sentences_per_chunk = 3
@@ -145,6 +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))
     context = retrieve_context(question, contexts)
+    ground_in_context = ground in context
     
     # Combine question and context
     input_text = f"question: {question} context: {context}"
@@ -182,7 +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, context, paraphrased_answer, pretrained_answer, pretrained_paraphrased_answer
+    return generated_answer, paraphrased_answer, ground_in_context, pretrained_answer, pretrained_paraphrased_answer, context
 
 
 # Define a function to list examples from the dataset
@@ -209,10 +210,11 @@ iface = gr.Interface(
     ],
     outputs=[
         Textbox(label="Generated Answer"),
-        Textbox(label="Retrieved Context"),
         Textbox(label="Natural Answer"),
+        Checkbox(label="Ground truth in the retrieved context"),
         Textbox(label="Pretrained Model's Answer"),
-        Textbox(label="Pretrained Model's Natural Answer")
+        Textbox(label="Pretrained Model's Natural Answer"),
+        Textbox(label="Retrieved Context")
     ],
     examples=list_examples(),
     examples_per_page=1,