app.py

import streamlit as st
import numpy as np
import tensorflow as tf
from tensorflow.keras.preprocessing.text import Tokenizer
from tensorflow.keras.preprocessing.sequence import pad_sequences
from tensorflow.keras.models import load_model
import pickle
import joblib

# Load model and tokenizer
model = load_model('rnn_lstm_final.h5')
loaded_model = joblib.load("my_rnn_model.joblib")

with open("tokenizer_and_sequences.pkl", "rb") as f:
    tokenizer, data = pickle.load(f)

# Define helper functions
def is_valid_punjabi_text(text):
    english_alphabet = set("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ")
    numbers = set("0123456789")
    punctuation = set("!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~")

    for char in text:
        if char in english_alphabet or char in numbers or char in punctuation:
            return False
    return True

def process_input(text):
    a = [text]
    a = tokenizer.texts_to_sequences(a)
    a = np.array(a)
    a = pad_sequences(a, padding='post', maxlen=100)
    a = a.reshape((a.shape[0], a.shape[1], 1))
    prediction = model.predict(np.array(a))
    for row in prediction:
        element1 = row[0]
        element2 = row[1]
        return "Negative" if element1 > element2 else "Positive"

# Streamlit app
st.title("Indic Sentence Summarization & Sentiment Analysis")
st.header("Insightful Echoes: Crafting Summaries with Sentiments (for ਪੰਜਾਬੀ Text)")

summarize_before_sentiment = st.checkbox("Summarize before analyzing sentiment")
user_input = st.text_area("Enter some text here")

if st.button("Analyze Sentiment"):
    if not is_valid_punjabi_text(user_input):
        st.warning("Please enter valid Punjabi text.")
    else:
        sentiment_output = process_input(user_input)
        st.text_area("Sentiment Output", sentiment_output, height=200)