Definition
A retrieval architecture that unifies dense vector representations (capturing semantic intent) and sparse keyword-based indices (capturing exact lexical matches) to optimize for both context and precision, albeit at the cost of increased storage overhead and complex result merging.
Combines 'meaning-based' search with 'exact-word' search, rather than referring to a multi-cloud storage strategy.
"A GPS system that uses both satellite coordinates (semantic proximity) and a specific street address (lexical keyword) to locate a destination."
- BM25(Component)
- Vector Embedding(Component)
- Reciprocal Rank Fusion (RRF)(Prerequisite)
- Sparse Vector(Component)
Conceptual Overview
A retrieval architecture that unifies dense vector representations (capturing semantic intent) and sparse keyword-based indices (capturing exact lexical matches) to optimize for both context and precision, albeit at the cost of increased storage overhead and complex result merging.
Disambiguation
Combines 'meaning-based' search with 'exact-word' search, rather than referring to a multi-cloud storage strategy.
Visual Analog
A GPS system that uses both satellite coordinates (semantic proximity) and a specific street address (lexical keyword) to locate a destination.