app.py

import gradio as gr
from transformers import pipeline
from difflib import SequenceMatcher

with open("my_document.txt", "r") as f:
    document_text = f.read()

question_input = gr.inputs.Textbox(label="Inserisci la tua domanda qui")

qa_model = pipeline("question-answering", model="timpal0l/mdeberta-v3-base-squad2")

def similar_questions(question, question_list):
    max_ratio = 0
    best_question = ""
    for q in question_list:
        ratio = SequenceMatcher(None, question.lower(), q.lower()).ratio()
        if ratio > max_ratio:
            max_ratio = ratio
            best_question = q
    return best_question if max_ratio >= 0.7 else None

def answer_question(question):
    ground_truth_questions = ["Qual è la capitale dell'Italia?", "Chi ha scritto la Divina Commedia?", "Qual è il fiume più lungo del mondo?"]
    similar_question = similar_questions(question, ground_truth_questions)
    if similar_question:
        if similar_question == "Qual è la capitale dell'Italia?":
            return "Roma"
        elif similar_question == "Chi ha scritto la Divina Commedia?":
            return "Dante Alighieri"
        elif similar_question == "Qual è il fiume più lungo del mondo?":
            return "Nilo"
    else:
        answer = qa_model(question=question, context=document_text, max_answer_len=200)
        return answer["answer"]

gr.Interface(fn=answer_question, inputs=question_input, outputs="text").launch()