Document Structure-Based Chunking

~10 min read

Use a document's own structure — headings, sections, paragraphs — to define chunk boundaries directly, maintaining structural integrity with the document's logical organization.

Document Structure-Based Chunking is a Pro topic

Sign in, then upgrade to Pro or Power to unlock this topic and the full AI Engineering curriculum.

Key points

  • Uses the document's own headings/sections/paragraphs directly as chunk boundaries — no embeddings or size computation needed
  • Maintains structural integrity by aligning exactly with the document's logical organization
  • Works best on genuinely well-structured content (manuals, specs, contracts) — assumes structure that may not exist
  • Resulting chunks can vary wildly in length, sometimes exceeding embedding model token limits
  • Commonly merged with recursive splitting as a second pass, to cap any oversized structural section