app.py

# Codes that passed test

'''
import streamlit as st
from transformers import pipeline

# Load model only once using caching
@st.cache_resource  # Use Streamlit's caching to avoid reloading the model
def load_sentiment_pipeline():
    return pipeline("sentiment-analysis", model="Rocky080808/finetuned-roberta-base")

# Main application logic
def main():
    # Load the sentiment analysis pipeline only once
    sentiment_pipeline = load_sentiment_pipeline()

    st.title("Final Project Demonstration for Group 8")
    st.write("This is an application for customer comments sentiment analysis for an e-commerce company.")
    st.write("Please input the customer comments for analysis below:")

    user_input = st.text_input("Enter customer comments here:")

    # Define a mapping from label to English descriptions
    label_to_text = {
        0: "Very dissatisfied, immediate follow-up is required.",
        1: "Dissatisfied, please arrange follow-up.",
        2: "Neutral sentiment, further case analysis is needed.",
        3: "Satisfied, the customer may return for a purchase.",
        4: "Very satisfied, the customer is very likely to return and recommend."
    }

    if user_input:
        # Call the preloaded pipeline to analyze sentiment
        result = sentiment_pipeline(user_input)
        label_str = result[0]["label"]  # Get the label as a string, e.g., "LABEL_0"
        label = int(label_str.split("_")[-1])  # Extract the numeric part of the label
        confidence = result[0]["score"]

        # Get the corresponding text description based on the label
        sentiment_text = label_to_text.get(label, "Unrecognized sentiment")

        st.write(f"Sentiment Analysis Result: {sentiment_text}")
        # Hide the confidence score, no need to show to the users
        # st.write(f"Confidence Score: {confidence:.2f}")

if __name__ == "__main__":
    main()
'''

# New codes to be tested

import streamlit as st
from transformers import pipeline
from langdetect import detect

# Load translation pipeline for multiple languages
@st.cache_resource  # Cache the model to avoid reloading it
def load_translation_pipeline():
    return pipeline("translation", model="facebook/m2m100_418M")

# Load sentiment analysis pipeline
@st.cache_resource  # Cache the sentiment analysis model
def load_sentiment_pipeline():
    return pipeline("sentiment-analysis", model="Rocky080808/finetuned-roberta-base")

# Function to detect language and translate to English
def translate_to_english(text, translation_pipeline):
    # Detect the language of the input text
    detected_language = detect(text)
    
    # Supported languages: Chinese, Japanese, German, Spanish, French
    language_map = {
        'zh': "zh",  # Chinese
        'ja': "ja",  # Japanese
        'de': "de",  # German
        'es': "es",  # Spanish
        'fr': "fr"   # French
    }
    
    if detected_language not in language_map:
        return None, "Unsupported language"
    
    # Translate the text to English using the detected language
    translated_text = translation_pipeline(text, src_lang=language_map[detected_language], tgt_lang="en")
    
    return translated_text[0]['translation_text'], detected_language

# Main application logic
def main():
    # Load the translation and sentiment pipelines
    translation_pipeline = load_translation_pipeline()
    sentiment_pipeline = load_sentiment_pipeline()

    st.title("Final Project Demonstration for Group 8")
    st.write("This application supports customer comments sentiment analysis for an e-commerce company.")
    st.write("You can input text in Chinese, Japanese, German, Spanish, or French. The text will be translated to English for sentiment analysis.")

    user_input = st.text_input("Enter customer comments in supported languages:")

    # Define a mapping from label to English descriptions
    label_to_text = {
        0: "Very dissatisfied, immediate follow-up is required.",
        1: "Dissatisfied, please arrange follow-up.",
        2: "Neutral sentiment, further case analysis is needed.",
        3: "Satisfied, the customer may return for a purchase.",
        4: "Very satisfied, the customer is very likely to return and recommend."
    }

    if user_input:
        # Step 1: Translate the input text to English
        translated_text, detected_language = translate_to_english(user_input, translation_pipeline)
        
        if detected_language == "Unsupported language":
            st.write("The input language is not supported. Please use Chinese, Japanese, German, Spanish, or French.")
        else:
            # Display the translated text
            st.write(f"Detected language: {detected_language}")
            st.write(f"Translated Text: {translated_text}")

            # Step 2: Perform sentiment analysis on the translated text
            result = sentiment_pipeline(translated_text)
            label_str = result[0]["label"]  # Get the label as a string, e.g., "LABEL_0"
            label = int(label_str.split("_")[-1])  # Extract the numeric part of the label
            confidence = result[0]["score"]

            # Get the corresponding text description based on the label
            sentiment_text = label_to_text.get(label, "Unrecognized sentiment")

            st.write(f"Sentiment Analysis Result: {sentiment_text}")
            st.write(f"Confidence Score: {confidence:.2f}")

if __name__ == "__main__":
    main()