Conquering PDFs: document understanding beyond plain text
summary
A practical wrapper integrates Dockling with spaCy to extract text plus layout spans and bounding boxes from PDFs, feed that Dockling document into existing NLP pipelines, and produce annotated spaCy docs for named entities, categories, and tables (tableformer output becomes a pandas DataFrame). A Prodigy scriptable workflow captures corrected annotations for transfer learning, OCR is pluggable, and choices between layout-aware models and simple text-first pipelines balance speed and accuracy — unified structured Dockling document, convert PDF tables to pandas DataFrame, treat PDFs as input not source of truth.
AI-written summary · about
chapters (10)
- 0:00Introduction and background
- 1:39The challenges of PDF data
- 5:09Integrating Dockling and spaCy
- 7:08Processing PDFs with spaCy
- 9:24Working with tabular data
- 11:32Workflow design and modularity
- 15:19Practical annotation workflows
- 17:04From training to production
- 19:09Key takeaways and conclusion
- 22:22Q&A session
related talks
29mConquering PDFs- document understanding beyond plain text
31mDocling: Get Your Documents Ready for Gen AI
26mAdvanced Document Parsing for AI Workflows
01h 07mMeet Docling: The “Pandas” for Document AI
01h 00mComplex Data Ingestion with Open Source AI
39mEnterprise-ready document processing for AI
01h 25mLearn to Unlock Document Intelligence with Open-Source AI
39mAdventures in Understanding Documents