Master 2.5 GB of unstructured specification documents with ease
summary
2.5 GB of messy .doc/.docx become structured JSON and a local browser UI that loads and full‑text searches everything in about 6.5 seconds. The pipeline uses pywin32 to drive Word exports to HTML, falls back to mammoth/python‑docx where possible, scrubs HTML with BeautifulSoup, extracts headings/tuples, analyzes co‑occurrence with pandas, builds graphs with networkx + py2cytoscape/cytoscape, and serves a serverless JS frontend with endless scrolling, recommendations and typo detection — it works and scales.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 0:56Outline
- 1:33Background
- 2:20Key challenge
- 5:18Target areas
- 6:53Solution overview
- 9:54Existing frameworks
- 12:38MS Word
- 13:59Identify relationships
- 15:48Identify relationships across documents
- 17:53User interface
- 19:43Call graph
- 23:08Conclusion
- 25:33Case study
- 27:23QA
related talks
01h 07mMeet Docling: The “Pandas” for Document AI
35mDocling: Get your documents ready for gen AI
31mDocling: Get Your Documents Ready for Gen AI
24mLightning Talks
27mWoosh: a fast pure-Python search engine library
30mAutoDocish: Automated-ish Dataset Documentation
02h 59mDocumenting your code from docstrings to automated builds
23mMkDocs: Documenting projects with Markdown