Task 1.5: Design Retrieval Mechanisms for FM Augmentation. AIP-C01 Practice Questions (Page 2)

Questions 6–10

1. 6application · medium

   An e-commerce company is using Amazon OpenSearch Service for product search. They have a mix of product descriptions and customer reviews. They want to improve search relevance for queries that contain specific product names (e.g., 'iPhone 15') and also general queries (e.g., 'best budget phone'). They are using Amazon Titan embeddings. What search architecture would best improve relevance?

   • AUse a larger embedding model to improve semantic understanding.
   • BImplement hybrid search combining keyword (BM25) and vector search, using a reranker model to merge results.
   • CUse only keyword-based search with BM25.
   • DUse only semantic search with vector embeddings.
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   A developer is designing a document segmentation strategy for a large corpus of technical manuals. The goal is to improve retrieval performance by preserving the logical structure of the content (e.g., sections, subsections, and paragraphs). Which segmentation approach best aligns with this requirement?

   • ASingle-chunk processing that treats the entire document as one chunk to avoid any information loss.
   • BHierarchical chunking that splits the document based on headings, subheadings, and paragraphs, preserving the content hierarchy.
   • CRandom chunking that selects arbitrary text segments to maximize diversity in the vector index.
   • DFixed-size chunking with a Lambda function that splits text into equal-sized chunks regardless of content boundaries.
   Reveal Solution Discussion Select an answer first
4. 8expert · hard

   A company is building a RAG application that requires low-latency retrieval (<100ms) for a high-traffic website. They are using Amazon OpenSearch Service. They have a large corpus of documents (100GB) and need to support both keyword and semantic search. They are considering using Amazon Bedrock Knowledge Bases with OpenSearch Serverless. What is the best approach to meet the latency requirement?

   • AUse Amazon OpenSearch Service with a provisioned cluster and optimize the index for low latency.
   • BUse Amazon Bedrock Knowledge Bases with OpenSearch Serverless, relying on its automatic scaling.
   • CUse Amazon Kendra for semantic search, which is optimized for enterprise search.
   • DUse Amazon Aurora with pgvector for vector search and a separate OpenSearch cluster for keyword search.
   Reveal Solution Discussion Select an answer first
5. 9foundation · easy

   A developer needs to implement a segmentation strategy for a collection of research papers. The papers have a consistent structure with sections like Abstract, Introduction, Methods, Results, and Conclusion. Which approach is most appropriate to ensure that each chunk represents a meaningful unit of content?

   • AUse hierarchical chunking that splits the document at section headings, creating chunks that correspond to each section.
   • BUse a single chunk for the entire paper to preserve the full context for retrieval.
   • CUse sentence-level chunking that splits the paper into individual sentences.
   • DUse fixed-size chunking with a token limit that is smaller than the average section length.
   Reveal Solution Discussion Select an answer first
6. 10foundation · easy

   A developer wants to improve the relevance of search results by reordering the top candidates retrieved from a vector search. Which Amazon Bedrock feature is designed for this purpose?

   • AAmazon Bedrock Knowledge Bases
   • BAmazon OpenSearch Service's k-NN plugin
   • CAmazon Bedrock reranker models
   • DAmazon Comprehend
   Reveal Solution Discussion Select an answer first