Questions on Hybrid Search Feature in Upcoming Release
TLDR Chetan inquired about the compatibility and parameter control of hybrid search feature. Jason clarified its functionality, indicating its compatibility with manually created vec field and parameters affecting keyword vs semantic search results.
Jul 24, 2023 (2 months ago)
• is it still compatible with documents indexed with a manually created
vecfield? or do i have to let typesense do the embedding for me from specific field
• is it possible to change the parameter controlling how keyword search results are weighed vs semantic search results?
0.25.0.rc53and later, you can send both a
qparameter and also
vector_queryparameter to do a hybrid search and combine the results together, without having to use built-in embedding generation
This is how results are ranked in hybrid search (it’s called Rank Fusion): https://gist.github.com/jasonbosco/f4187f6b4f585d2dc8902af85408994a#hybrid-search-rank-fusion
So by default keyword search is given more priority over semantic search.
You can influence keyword search ranking and semantic search ranking independently, but not both of them together
Indexed 2764 threads (79% resolved)
Hybrid Search with Keyword and Vector Search
R.A.I.D asked about controlling weights between keyword-based and vector-based search results, with emphasis on personal customization. Kishore Nallan confirmed this option will be possible in the 0.25 release.
Resolving Hybrid Query Error with 'vector_query'
Manish had an error with a hybrid query. Jason suggests using the field name `embedding` instead of `vec` and explains the weightage in hybrid searches.
Hybrid Search and Auto Embedding in Typesense
Md inquired about hybrid search and auto embedding. Jason explained how to use it and clarified the software version compatibility. Md confirmed v25 works well.