Definition
A retrieval architecture that executes parallel searches using high-dimensional sparse vectors (e.g., BM25) for keyword-exact matching and dense embeddings for semantic context, merging the results via fusion algorithms. It optimizes for both lexical precision and conceptual breadth, though it adds complexity in index management and result reranking.
Combines keyword matching with semantic meaning, rather than mixing data formats like images and text.
"A library search that simultaneously uses a word-for-word index and a map of conceptual relationships."
Conceptual Overview
A retrieval architecture that executes parallel searches using high-dimensional sparse vectors (e.g., BM25) for keyword-exact matching and dense embeddings for semantic context, merging the results via fusion algorithms. It optimizes for both lexical precision and conceptual breadth, though it adds complexity in index management and result reranking.
Disambiguation
Combines keyword matching with semantic meaning, rather than mixing data formats like images and text.
Visual Analog
A library search that simultaneously uses a word-for-word index and a map of conceptual relationships.