Word Document Parsing

Parse Word documents (.docx) into markdown, JSON, and image artifacts using multi-method extraction.

Usage

Run the parsing script directly:

./scripts/parse_docx.py <path_to_file.docx> <output_dir>

Example:

./scripts/parse_docx.py ~/documents/report.docx ./parsed/

The script uses 4 extraction methods:

• python-docx (basic) - Fast text extraction
• python-docx (detailed) - Full structure with tables
• docx2txt - Simple text-only fallback
• markitdown - Microsoft's markdown converter

Output Structure

output_dir/
├── file.docx/
│   ├── parsing_summary.json
│   ├── python_docx_basic/
│   │   └── content.md
│   ├── python_docx_detailed/
│   │   ├── content.md
│   │   ├── tables.json
│   │   └── images/
│   ├── docx2txt/
│   │   └── content.txt
│   └── markitdown/
│       └── content.md

Script Features

• Self-contained Python script with inline uv metadata
• Handles multiple extraction methods for redundancy
• Creates JSON metadata for tables and document structure
• Extracts images with dimensions and metadata
• Continues on errors (one method failure doesn't stop others)