Why split documents into chunks for retrieval?
Why C is the answer
So that searches return focused, relevant passages
Chunking lets retrieval find the relevant part rather than a whole document.
In detail
- Chunks let a search return focused passages.
- Chunks that are too small lose context.
Use case
An illustrative scenario, not a real incident.
A 200-page manual cannot fit a focused answer. Chunks let retrieval return just the page about resetting a device.
See it in different ways
sequenceDiagram participant U as User participant K as Knowledge base participant M as Foundation model U->>K: Question K-->>M: Relevant chunks with the question M-->>U: Grounded answer
sequenceDiagram participant U as User participant K as Knowledge Bases participant S as Vector store participant M as Foundation model U->>K: Question K->>S: Finds similar chunks S-->>K: Passages K->>M: Prompt with passages M-->>U: Answer
flowchart LR S0["Document"] --> S1["Chunks"] --> S2["Embeddings"] --> S3["Relevant chunk returned"]
Key point: Chunks make retrieval focused.
Why the other options are wrong
- A. So that the model can be trained more quickly
- Chunking is not for training.
- B. So that the documents can be translated automatically
- It does not translate.
- D. So that documents take up less space in the model
- Chunks are not stored in the model.