A discharge-instructions PDF describes wound care across two consecutive paragraphs that build on each other. Why might a Data 360 admin configure some overlap between adjacent chunks during indexing?
Select an answer to reveal the explanation.
Short Explanation
Without a little overlap, a sentence can get chopped right in half at a chunk boundary — a slice of shared context on each side keeps the idea from falling apart when it's retrieved.
Full Explanation
Configuring some overlap between adjacent chunks means the end of one chunk repeats a bit of content at the start of the next, which helps preserve continuity for ideas that span a chunk boundary — like wound-care guidance that builds across two paragraphs — so a retrieved chunk is less likely to feel cut off mid-thought. This is a deliberate quality trade-off, not a performance shortcut, so claiming it doubles indexing speed misrepresents its purpose and effect. Overlap is a configurable choice made by whoever sets up chunking, not a mandatory, non-optional platform requirement — different content types may reasonably use different overlap settings. Overlap also has nothing to do with which search type a retriever uses; chunking strategy and search type (keyword, vector, hybrid) are independent configuration decisions that happen at different stages of the grounding pipeline. For the exam, remember chunk overlap as a technique for preserving context continuity, distinct from chunk size and distinct from search type selection.