AI Glossary Term

What does “Parent-Document Retrieval” mean?

Definition

A retrieval pattern that searches small chunks but returns their larger parent documents for richer context.

How Parent-Document Retrieval works

Diagram for Parent-Document Retrieval, steps in order: Split documents into small chunks; then Search the small chunks for the best match; then Return the full parent section; then Model reads richer context.Split documents into small chunksSearch the small chunks for the bestmatchReturn the full parent sectionModel reads richer context

Example

A small paragraph about a refund deadline matches the query precisely. The system returns the whole policy section containing it, so the model also sees the exceptions that follow.

Why it matters

Small chunks match precisely but may lack context; larger ones carry context but match poorly. This approach tries to get both.

Common questions

What is a parent document?

The larger section, page, or file that a small chunk was cut from.

Downside?

Larger returns use more of the context window.

How do I choose sizes?

Test on your own questions. Chunk size should fit how your documents are organized.

Learn more in our free guides

Test yourself

See also