Concept guide · RAG foundations
How do overlap and parent-child retrieval help?
A slower explanation for when the short interview answer is not enough. Start with the intuition, then connect it to the technical idea and the way an interviewer expects you to reason about it.
ML area
Generative AI → Overlap and parent-child retrieval
Interview format
Core · Concept
Question
GENAI-052
Start with the idea
In simple terms
Overlap repeats a little text between neighbouring chunks so a fact near a boundary is not lost. Parent-child retrieval searches small precise pieces but returns a larger parent section when the model needs context.
Make it concrete
Search a paragraph-sized child that contains “eligibility,” then return its full policy subsection so the answer also sees the exception and effective date.
Build the answer
How to reason about it
- 1Use overlap only when boundary misses are demonstrated.
- 2Retrieve with a precise child and expand to a bounded parent context.
- 3Control duplicate evidence, token cost, and citation granularity.
Go one level deeper
The technical explanation
Overlap can improve boundary recall but increases duplicates, index size, and prompt noise. Parent-child designs separate the unit used for ranking from the unit used for reading. Store the relationship explicitly, deduplicate parent sections, and evaluate whether the added context improves grounded answers rather than assuming that larger context is automatically helpful.
Say this in the interview
Overlap repeats boundary text so a fact is less likely to be split away from its context, but too much overlap creates duplicates and cost. Parent-child retrieval indexes small children for precision while returning a larger parent section for context. Both need evaluation because more text is not always better.
Keep preparing