Structure-aware chunking with Docling
A document-preparation recipe for balancing document structure with an embedding model’s token budget.
Published by Docling Project
How to use it
Convert a document to Docling’s structured representation, then use the hybrid chunker with the tokenizer appropriate to the embedding model. Inspect headings and metadata alongside the chunk text.
Compare a small collection against a simple fixed-length split. Check whether the retriever returns intact facts and usable citation locations, especially for tables or sections that span pages.
- Compatibility
- Python, Docling, PDF
- Components
- Structure-aware chunking, Document parsing, Token budgets
- Access
- Local Python recipe. Tokenizer or model artifacts may need downloading.