[Python] How to extract page-wise HTML content from DOCX files using docling library by...

I have successfully implemented page-wise HTML extraction for PDF files using docling and PyPDF2. Here's what my current code does for PDFs:

1. Uses PyPDF2 to split PDF into individual pages
2. Converts each page to HTML using docling's DocumentConverter
3. Extracts HTML content with embedded images
4. Adds metadata (page number, document ID, filename)
5. Saves everything to a JSON structure

Important Note: The reason I'm splitting PDF into individual pages first is because docling's save_as_html() and export_to_html() functions work on complete document objects, not on individual pages. To get page-wise HTML content, I need to create temporary single-page PDFs and convert each one separately.

Here's the sample JSON structure I get for each page: { "page": "Page 1", "content": "", "metadata": { "documentId": "uuid", "filename": "document.pdf", "page_number": 1, "total_pages": total } }

Now I need to implement the same functionality for DOCX files.According to me, DOCX files contain elements like headers, footers, and page breaks, we could use these page breaks to split the content into pages.

I'm using the docling library for conversion (docx to HTML), but I'm unable to identify or detect page breaks in the DOCX file. Since docling's HTML conversion works on complete documents, I need to first split the DOCX content based on page breaks, similar to how I'm handling PDFs.

Questions:

1. How can I detect page breaks in a DOCX file using docling?
2. Is there a way to split DOCX content based on these page breaks to create separate document objects?
3. If docling doesn't support this directly, are there other Python libraries I should use alongside docling to detect and split at page breaks?

I've tried looking through the docling documentation but couldn't find information about handling page breaks in DOCX files.

Any help or guidance would be appreciated!

Related libraries I'm using:

• docling
• python-docx (if needed)

Continue reading...