app.py

import pandas as pd
import streamlit as st
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# Load the pre-trained model and tokenizer
tokenizer = AutoTokenizer.from_pretrained("textattack/roberta-base-imdb")
model = AutoModelForSequenceClassification.from_pretrained("textattack/roberta-base-imdb")

def analyze_text(text):
    # Preprocess the text
    text = text.lower()

    # Encode the text
    encoded_text = tokenizer(text, truncation=True, padding=True, return_tensors='pt')

    # Classify the text
    with torch.no_grad():
        output = model(**encoded_text)
        predictions = output.logits.argmax(-1).item()

    if predictions == 1:  # For IMDb sentiment analysis, 1 indicates positive sentiment
        return "Job Related"
    else:
        return "Not Job Related"

def count_job_related_messages(data):
    job_related_count = 0
    not_job_related_count = 0

    for message in data["message"]:
        result = analyze_text(message)
        if result == "Job Related":
            job_related_count += 1
        else:
            not_job_related_count += 1

    return job_related_count, not_job_related_count

# Streamlit application
st.title("Job Related Message Analyzer")

uploaded_file = st.file_uploader("Upload CSV file")
user_input = st.text_input("Enter text")

if uploaded_file:
    # Read the CSV file
    data = pd.read_csv(uploaded_file)

    # Analyze messages
    results = []
    for message in data["message"]:
        result = analyze_text(message)
        results.append(result)

    data["Job Related"] = results

    # Count job-related messages
    job_related_count, not_job_related_count = count_job_related_messages(data)

    st.dataframe(data)
    st.write(f"Job Related Messages: {job_related_count}")
    st.write(f"Not Job Related Messages: {not_job_related_count}")
elif user_input:
    # Analyze user-input text
    result = analyze_text(user_input)
    st.write(f"Message Classification: {result}")
else:
    st.write("Please upload a CSV file or enter text to analyze.")