Question 18
Domain 2: Fundamentals of Generative AIWhat is the purpose of chunking in Retrieval Augmented Generation (RAG)?
Correct answer: C
Explanation
Chunking splits source documents into smaller pieces so the retriever can return the most relevant passages instead of entire files. In RAG, this improves “the contextual relevancy of results retrieved from the vector index” because embeddings are created and searched at the chunk level, not the whole document level.
Why each option is right or wrong
A. To avoid database storage limitations for large text documents by storing parts or chunks of the text
Chunking supports retrieval and embeddings, not database storage limits.
B. To improve efficiency by avoiding the need to convert large text into vector embeddings
Chunking does not replace embeddings; Knowledge Bases generates vector embeddings for chunks.
C. To improve the contextual relevancy of results retrieved from the vector index
AWS Knowledge Bases for Amazon Bedrock explicitly performs document chunking before embedding and retrieval, and the chunking settings are part of the retrieval pipeline described in the Bedrock documentation. By splitting source material into smaller segments, the vector search can return the most relevant passage-level matches at inference time instead of dragging back an entire document, which improves the relevance of retrieved context for the generator.
D. To decrease the cost of storage by storing parts or chunks of the text
Chunking improves retrieval quality, not storage cost reduction.