Explanation

Option B is the correct solution because legal research workloads require both semantic understanding and exact lexical precision, especially for statutes, citations, and domain-specific terminology. A hybrid search architecture directly addresses this need by combining vector similarity search with traditional keyword-based retrieval.

Vector search alone is often insufficient for legal research because exact phrases, citation formats, and jurisdiction-specific terms must be matched precisely. Keyword search ensures high recall and precision for citations and legal terms, while vector search captures deeper semantic relationships between legal concepts, precedents, and arguments. Amazon OpenSearch Service natively supports hybrid search, enabling efficient scoring and ranking without external orchestration.

Key reasons why Option B is correct:

1. Hybrid Search Architecture: Combines vector search (for semantic understanding of legal concepts) with keyword search (for precise matching of citations and legal terminology)
2. Amazon Bedrock Reranker: Optimizes result relevance by re-ranking search results based on semantic relevance
3. Performance Requirements: Meets the need for quick performance and precise results
4. Domain-Specific Needs: Legal research requires both semantic relationships (vector search) and exact lexical matches (keyword search)

Why other options are incorrect:

• Option A: Default vector search configuration may not handle precise citation matching effectively, and query expansion alone may not provide the required precision for legal terminology
• Option C: Amazon Kendra query suggestions focus on helping users formulate queries, not on the core retrieval architecture needed for legal research
• Option D: While using vector search with Titan Embeddings is good, the custom Lambda functions and RDS database approach adds unnecessary complexity compared to OpenSearch's native hybrid search capabilities