Spaces:
Sleeping
Sleeping
| import streamlit as st | |
| from transformers import pipeline | |
| from transformers import AutoModelForQuestionAnswering, AutoTokenizer | |
| st.set_page_config(page_title="Automated Question Answering System") # set page title | |
| # heading | |
| st.markdown("<h2 style='text-align: center;'>Question Answering on Academic Essays</h2>", unsafe_allow_html=True) | |
| # description | |
| st.markdown("<h3 style='text-align: left; color:#F63366; font-size:18px;'><b>What is extractive question answering about?<b></h3>", unsafe_allow_html=True) | |
| st.write("Extractive question answering is a Natural Language Processing task where text is provided for a model so that the model can refer to it and make predictions about where the answer to a question is.") | |
| # store the model in cache resources to enhance efficiency (ref: https://docs.streamlit.io/library/advanced-features/caching) | |
| def question_model(): | |
| # call my model for question answering | |
| model_name = "deepset/deberta-v3-base-squad2" | |
| tokenizer = AutoTokenizer.from_pretrained(model_name) | |
| model = AutoModelForQuestionAnswering.from_pretrained(model_name) | |
| question_answerer = pipeline("question-answering", model=model, tokenizer=tokenizer) | |
| return question_answerer | |
| # get the answer by passing the context & question to the model | |
| def question_answering(context, question): | |
| with st.spinner(text="Loading question model..."): | |
| question_answerer = question_model() | |
| with st.spinner(text="Getting answer..."): | |
| answer = question_answerer(context=context, question=question) | |
| answer_score = str(answer["score"]) | |
| answer = answer["answer"] | |
| # display the result in container | |
| container = st.container(border=True) | |
| container.write("<h5><b>Answer:</b></h5>"+answer+"<p><small>(F1 score: "+answer_score+")</small></p><br>", unsafe_allow_html=True) | |
| # choose the source with different tabs | |
| tab1, tab2 = st.tabs(["Input text", "Upload File"]) | |
| #---------- input text ---------- | |
| # if type the text as input | |
| with tab1: | |
| # set the example | |
| sample_question = "What is NLP?" | |
| with open("sample.txt", "r") as text_file: | |
| sample_text = text_file.read() | |
| # Get the initial values of context and question | |
| context = st.session_state.get("contextInput", "") | |
| question = st.session_state.get("questionInput", "") | |
| # Button to try the example | |
| example = st.button("Try example") | |
| # Update the values if the "Try example" button is clicked | |
| if example: | |
| context = sample_text | |
| question = sample_question | |
| # Display the text area and text input with the updated or default values | |
| context = st.text_area("Enter the essay below:", value=context, key="contextInput", height=330) | |
| question = st.text_input(label="Enter the question: ", value=question, key="questionInput") | |
| # perform question answering when "get answer" button clicked | |
| button = st.button("Get answer", key="textInput") | |
| if button: | |
| if context=="" or question=="": | |
| st.error ("Please enter BOTH the context and the question", icon="🚨") | |
| else: | |
| question_answering(context, question) | |
| # ---------- upload file ---------- | |
| # if upload file as input | |
| with tab2: | |
| # provide upload place | |
| uploaded_file = st.file_uploader("Choose a .txt file to upload", type=["txt"]) | |
| # transfer file to context and allow ask question, then perform question answering | |
| if uploaded_file is not None: | |
| raw_text = str(uploaded_file.read(),"utf-8") | |
| context = st.text_area("Your essay context: ", value=raw_text, height=330) | |
| question = st.text_input(label="Enter your question", value="Enter question here") | |
| # perform question answering when "get answer" button clicked | |
| button2 = st.button("Get answer", key="fileInput") | |
| if button2: | |
| if context=="" or question=="": | |
| st.error ("Please enter BOTH the context and the question", icon="🚨") | |
| else: | |
| question_answering(context, question) | |
| st.markdown("<br><br><br><br><br>", unsafe_allow_html=True) | |