changed question wording to improve performance

app.py

@@ -17,7 +17,7 @@ def load_pipeline(model_cp, tokenizer_cp):
 
 
 def choose_model():
-    with st.sidebar: # streamlit doesn't know how to has this by default?
+    with st.sidebar:
         st.write("# Model Selection")
         model_cp = st.selectbox('Select model for inference',
                             ('deepset/roberta-base-squad2',
@@ -53,7 +53,7 @@ if uploaded_file is not None:
 
     # Run inference on first example
     first_example = df['review'][0]
-    question = "how to improve this recipe?"
+    question = "How was the recipe improved?"
     answer = question_answer(question=question, context=first_example)
 
     # Present results