liveBook · Manning

BookRAG

As explained in Retrieval Augmented Generation, The Foundational Ideas

BookRAG is a retrieval method that maps queries and retrieval operations onto a document's inherent table-of-contents hierarchy.

BookRAG maps retrieval onto a document’s inherent table-of-contents hierarchy instead of inducing a synthetic hierarchy through unsupervised clustering. This makes it especially suitable for corpora with clean structural metadata. section 10.7.3

BookRAG’s central design tradeoff is structural fit:

  • Prefer BookRAG when documents already provide reliable chapters, sections, or numbered headings. section 10.7.3
  • Prefer induced approaches such as RAPTOR when the corpus lacks a dependable native hierarchy. section 10.7.3

Structured documents

A textbook with chapters and sections, or an SEC filing with numbered sections, can expose its existing hierarchy directly to retrieval rather than relying on clustering. section 10.7.3

How Retrieval Augmented Generation, The Foundational Ideas covers it

How deep it goes
Defines the structural choice and contrasts it with RAPTOR, but does not detail indexing or query-time implementation.
Examples
Uses textbooks and SEC filings as examples of corpora with strong native hierarchy.
What it assumes
Assumes documents expose a reliable hierarchy such as chapters, sections, or numbered headings.
What you can do after
Apply it by preferring table-of-contents retrieval over unsupervised clustering when clean structural metadata exists.
How it connects
Connects intrinsic hierarchy with RAPTOR, HiRAG, CFT-RAG, and PageIndex.

This explainer was generated by Manning's knowledge graph from the text of Retrieval Augmented Generation, The Foundational Ideas. The section links point to the parts of the book it draws on.