Hybrid vector and full-text search in LanceDB
A table-level retrieval recipe for combining vector similarity and full-text results.
Published by LanceDB
How to use it
Create a full-text index on the passage column, retain the embedding field, and retrieve candidates using both modes. Keep a stable ID so any combined result points to the original passage.
Evaluate queries with rare identifiers and semantic paraphrases separately. Compare the hybrid ranking to the individual search modes under the same filters.
- Compatibility
- Python, LanceDB
- Components
- Hybrid retrieval, Full-text indexing, Vector search
- Access
- Local LanceDB recipe with full-text indexing and an embedding model.