train.py

import streamlit as st
import pandas as pd
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity
import numpy as np

# Load Q&A data
@st.cache
def load_data():
    return pd.read_csv('qna.csv', encoding='utf-8', delimiter=';')

def get_most_similar_question(new_sentence, vectorizer, tfidf_matrix, questions, answers):
    new_tfidf = vectorizer.transform([new_sentence])

    similarities = cosine_similarity(new_tfidf, tfidf_matrix)

    most_similar_index = np.argmax(similarities)

    similarity_percentage = similarities[0, most_similar_index] * 100

    return answers[most_similar_index], similarity_percentage

def answer_the_question(new_sentence, vectorizer, tfidf_matrix, questions, answers):
    most_similar_answer, similarity_percentage = get_most_similar_question(new_sentence, vectorizer, tfidf_matrix, questions, answers)
    if similarity_percentage > 70:
        return most_similar_answer
    else:
        return 'Sorry, I am not aware of this information :('

def main():
    st.title("Q&A Chatbot")

    # Load Q&A data
    data = load_data()
    questions = data['question'].tolist()
    answers = data['answer'].tolist()

    # Create TF-IDF vectorizer
    vectorizer = TfidfVectorizer()
    tfidf_matrix = vectorizer.fit_transform(questions)

    # Get user input
    user_question = st.text_input("Ask me a question:")

    # Process the question when submitted
    if st.button("Submit"):
        if user_question:
            response = answer_the_question(user_question, vectorizer, tfidf_matrix, questions, answers)
            st.write("Answer:")
            st.write(response)
        else:
            st.warning("Please ask a question.")

if __name__ == "__main__":
    main()