Task 1.5: Design Retrieval Mechanisms for FM Augmentation. AIP-C01 Practice Questions (Page 6)

Questions 26–26

1. 26application · medium

   A legal tech company is building a document Q&A system using Amazon Bedrock. Their documents are long contracts with clear section headings and subheadings. They need to retrieve precise clauses (e.g., indemnification) and also support broad questions like 'What are the payment terms?'. They are using Amazon Titan Text Embeddings and OpenSearch Serverless. What is the most effective segmentation strategy to balance precision and recall?

   • AUse hierarchical chunking based on document structure, storing both parent and child chunks with metadata linking them.
   • BUse fixed-size chunking with a small chunk size (e.g., 200 tokens) and no overlap.
   • CUse a single chunk per document to preserve full context.
   • DUse fixed-size chunking with a large chunk size (e.g., 2000 tokens) and 10% overlap.
   Reveal Solution Discussion Select an answer first