app.py

import gradio as gr
import pandas as pd
import re
import os
import fitz
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

tokenizer = AutoTokenizer.from_pretrained("potsawee/t5-large-generation-squad-QuestionAnswer")
model = AutoModelForSeq2SeqLM.from_pretrained("potsawee/t5-large-generation-squad-QuestionAnswer")

def extract_text_from_pdf(pdf_file_path):
    doc = fitz.open(pdf_file_path) 
    text = ""
    for page in doc:
        text+=page.get_text() 

    return text

def generate_question_answer_pairs(pdf_file):
    if pdf_file is None:
        return "Please upload a PDF file"

    d = {'Question':[],'Answer':[]}
    df = pd.DataFrame(data=d)

    pdf_text = extract_text_from_pdf(pdf_file.name)

    sentences = re.split(r'(?<=[.!?])', pdf_text)
    question_answer_pairs = []

    for sentence in sentences:
        input_ids = tokenizer.encode(sentence, return_tensors="pt")
        outputs = model.generate(input_ids, max_length=100, num_return_sequences=1)
        question_answer = tokenizer.decode(outputs[0], skip_special_tokens=True)
        question_answer_pairs.append(question_answer)

    result = ''

    for question_answer in question_answer_pairs:
        qa_parts = question_answer.split("?")
        if len(qa_parts) >= 2:
            question_part = qa_parts[0] + "?"
            answer_part = qa_parts[1].strip()
            new_data = {'Question': [question_part], 'Answer': [answer_part]}
            df = pd.concat([df, pd.DataFrame(new_data)], ignore_index=True)
            result += f"Question: {question_part}\nAnswer: {answer_part}\n\n"
            
    df.to_csv("QAPairs.csv")
    return result, "QAPairs.csv"

title = "Question-Answer Pairs Generation"
input_file = gr.File(label="Upload a PDF file")
output_file = gr.File(label="Download as csv")
output_text = gr.Textbox()

interface = gr.Interface(
    fn=generate_question_answer_pairs,
    inputs=input_file,
    outputs=[output_text, output_file],
    title=title,
)
interface.launch()