ebook2text 2.0.3

Convert common book file types to text for machine learning

Convert Ebook File

Overview

This Python script provides functionality for converting various ebook file formats (EPUB, DOCX, PDF, TXT) into a standardized text format. The script processes each file, identifying chapters, and replaces chapter headers with asterisks. It also performs OCR (Optical Character Recognition) for image-based text using GPT-4o and standardizes the text by desmartening punctuation.

Features

• File Format Support: Handles EPUB, DOCX, PDF, and TXT formats.
• Chapter Identification: Detects and marks chapter breaks.
• OCR Capability: Converts text from images using OCR.
• Text Standardization: Replaces smart punctuation with ASCII equivalents.

Requirements

To run this script, you need Python 3.9 or above and the following packages:

• python-docx
• openai
• python-dotenv
• bs4
• pdfminer.six
• pillow

Usage

1. Ensure all dependencies are installed.
2. Set your environment variable for the OpenAI API key.
3. Place your ebook files in a known directory.
4. Run the script with the path to the ebook file and a metadata dictionary with keys of 'title' and 'author' as arguments.

• set save_file to False, if you want a string returned.
• provide a Path object of a file name to be written to, to use a custom output filename.

Functions

• convert_file(file_path: Path, metadata: dict, *, save_file: bool = True, save_path: Optional[Path] = None) -> Union[str, None]: Main function to convert an ebook file to text.

Contributing

Contributions to this project are welcome. Please use Ruff for formatting to ensure that your code follows the existing style for consistency, and follow the ProsePal Open Source Contributor's Code of Contact.

TODO

• Increase test coverage
  • Tests for text converter
  • More edge cases and failure states
• Better handling of ebooklib dependency
• Add additional AI models for OCR as plugins
• Explore additional filetypes
• Other options for determining filetype

License

This project is licensed by ProsePal LLC under the MIT license