Store document embeddings and retrieve documents close to a query vector.
Use vector distance together with metadata filters rather than keyword matching alone.
Which data store is most appropriate?
Use Amazon OpenSearch vector search with an approximate nearest-neighbor index and metadata filters.
Detailed explanation
Object-key order does not calculate semantic vector distance.
Object-key order does not calculate semantic vector distance.
A message ordering group is not a vector-search index.
A message ordering group is not a vector-search index.
LIKE matches text patterns and does not perform nearest-neighbor vector search.
LIKE matches text patterns and does not perform nearest-neighbor vector search.
OpenSearch supports vector fields, approximate nearest-neighbor search, and metadata filters.
OpenSearch supports vector fields, approximate nearest-neighbor search, and metadata filters.
Check official documentation
Compare the following points with the official documentation.
Amazon OpenSearch Serviceのk-NN・ベクトルフィールド・検索クエリの公式説明を確認する。
What to confirm
全文・LIKE・キュー順序とベクトル近傍検索の違いを説明できる。
Key points
- Vector search
- Embedding
- k-NN
- Metadata filter
Notes
- Environment: AWS公式DEA-C01試験ガイドと各サービスの公式ドキュメントの確認
- Confirm current AWS service behavior in the official DEA-C01 documentation.
Foundation review
Semantic versus lexical search
Vector distance represents similarity while keyword indexes match terms.
Recall and latency
Approximate search trades some exactness for lower latency and scalable indexing.