app.py

import streamlit as st
from transformers import pipeline

# Set up Streamlit title and description
st.title("Sentiment Analysis, Audio Feedback, and Response Generation")
st.write("Enter a review and get the sentiment analysis with audio feedback and a professional response.")

# Load the sentiment analysis model
sentiment_analysis = pipeline("text-classification", model="JACOBBBB/CustomModel_JL")

# Load the text-to-speech model
text_to_audio = pipeline("text-to-speech", model="facebook/fastspeech2-en-ljspeech")

# Load the text generation model
response_generator = pipeline("text-generation", model="gpt-2")

# Text input for user to enter the review
review = st.text_area("Enter the review to analyze", "")

# Perform operations when the user clicks the "Analyze and Respond" button
if st.button("Analyze and Respond"):
    # Perform sentiment analysis on the input review
    sentiment_result = sentiment_analysis(review)[0]
    label = sentiment_result['label']
    score = sentiment_result['score']

    # Display the sentiment analysis result
    st.write("Review:", review)
    st.write("Predicted Sentiment:", label)
    st.write("Confidence Score:", f"{score * 100:.2f}%")

    # Generate the response text
    response_text = f"The sentiment of the review is {label} with a confidence of {score * 100:.2f} percent."

    # Generate audio from response text
    audio = text_to_audio(response_text)
    st.audio(audio['audio'], format='audio/wav', sample_rate=audio['sampling_rate'])

    # Generate a professional response based on the review and its sentiment
    prompt = f"The guest has left a {label} review stating: '{review}'. How should the hotel management professionally respond?"
    professional_response = response_generator(prompt, max_length=150, num_return_sequences=1)[0]['generated_text']

    # Display the generated professional response
    st.write("Generated Professional Response:", professional_response)