LlamaIndex
Log in

Find answers from the community

Updated 8 months ago

Scale

mmichal382
·
When combining the retrieval results of the bm25 retriever and a dense retriever is there a way to scale the scores from the bm25 within 0-1 so that they can be compared with cosine similarity scores from the dense retriever?
L
m
2 comments
LLogan M
Yea there's a few ways of doing it.

Check out this blog for a starting point
https://weaviate.io/blog/hybrid-search-fusion-algorithms

We have this built into the query fusion retriever

https://docs.llamaindex.ai/en/stable/examples/retrievers/relative_score_dist_fusion/?h=fusion
mmichal382
thanks will take a look!
Add a reply
Sign up and join the conversation on Discord