preprocess_data.py

import pandas as pd
# from langchain.text_splitter import RecursiveCharacterTextSplitter  # No longer needed
from tqdm import tqdm

# Load the dataset
input_xl = "Final_articles_content_with_hindi.xlsx"  # Updated to the correct input file name
output_csv = "New_chunked_articles.csv"

# Read the input CSV file
data = pd.read_excel(input_xl)

# Prepare a list to store the chunked data
chunked_data = []

# Iterate through the rows and split the content into paragraphs
for idx, row in tqdm(data.iterrows(), total=len(data)):
    title = row["Title"]
    url = row["URL"]
    content = row["Content"]

    # Split the content into paragraphs using double newlines as separators
    paragraphs = content.split('\n\n')

    # Add each paragraph as a new row in the chunked dataset
    for i, paragraph in enumerate(paragraphs):
        if paragraph.strip():  # Only add non-empty paragraphs
            chunked_data.append({
                "URL": url,
                "Title": title,  # Keep the original title as each row represents a part of it
                "Content Chunk": paragraph.strip()  # Store the paragraph content, stripping leading/trailing spaces
            })

# Convert the chunked data into a DataFrame
chunked_df = pd.DataFrame(chunked_data)

# Save the chunked data to a new CSV file
chunked_df.to_csv(output_csv, index=False, encoding="utf-8")

print(f"Chunked data saved to '{output_csv}'")