app.py

import re
import nltk
import pickle
import numpy as np
import pandas as pd
import streamlit as st 
from datasets import load_dataset
from sklearn.metrics.pairwise import cosine_similarity

nltk.download('punkt')

dataset = pd.read_csv("Preprocess_LK_Hadith_dataset.csv")

labels = dataset['Arabic_Grade']

# Helper functions
def remove_tashkeel(text):
    tashkeel_pattern = re.compile(r'[\u0617-\u061A\u064B-\u0652]')
    return re.sub(tashkeel_pattern, '', text)

def preprocess_arabic_text(text):
    text = remove_tashkeel(text)
    tokens = nltk.word_tokenize(text)
    cleaned_tokens = [token for token in tokens if token.isalnum()]
    lowercase_tokens = [token.lower() for token in cleaned_tokens]
    return " ".join(lowercase_tokens)

# Function to predict label
def predict_label(input_text, threshold=0.5):
    with open("tfidf_vectorizer.pkl", "rb") as f:
        vectorizer = pickle.load(f)
    with open("cosine_similarity_model.pkl", "rb") as f:
        X = pickle.load(f)

    input_text = preprocess_arabic_text(input_text)
    input_vector = vectorizer.transform([input_text])
    similarities = cosine_similarity(input_vector, X).flatten()

    max_index = np.argmax(similarities)
    max_similarity = similarities[max_index]

    if max_similarity >= threshold:
        return labels.iloc[max_index]
    else:
        return "No similar text found in dataset"

x = st.slider('Enter Hadith') 
st.write(x, 'Hadith Classification', predict_label)