-
Notifications
You must be signed in to change notification settings - Fork 688
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Adding KNN scorer explanation #1914
Conversation
f386d79
to
efaeacb
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks fine to me.
@metonymic-smokey would you add a sample snippet of the "explain score" output to the commit message here.
Also, I'm glad you've set the base branch to MB-59102
so there'd be no conflicts, but I think we should merge this into unstable after the base PR has been so we don't squash this into it and create a separate commit entry for this.
efaeacb
to
3b9f58d
Compare
Done, pls let me know what you think. |
Includes: * dd26ea1 Thejas-bhat | command line tool support for vector search (#175) * d292ef9 Thejas-bhat | replacing the flat index with the IVF family index. (#180) * f5e149a Thejas-bhat | bug fix: accounting the vecIDs approriately while merging (#179) * 0296d71 Thejas-bhat | optimization: avoiding storing of bitmaps when vector is in a single document (#181) * 6b9b047 Abhi Dangeti | MB-59692: Respect choice of similarity metric, [email protected] (#182) * e1fe4fb Abhi Dangeti | MB-59569: Upgrade to blevesearch/[email protected] (#183)
# Jira [MB-59102](https://issues.couchbase.com/browse/MB-59102) ## Description When performing a MultiSearch across an alias representing a partitioned index, we need special logic to merge the KNN results together into the final search result. --------- Co-authored-by: Abhinav Dangeti <[email protected]>
This PR allows a user to specify a KNN Operator to determine if KNN hits are conjuncted/disjuncted with filter query hits.
Sample explanation: { "value": 0.125, "message": "weight(desc:query Vector^1.000000 in one), product of:", "children": [ { "value": 0.5, "message": "queryWeight(desc:query Vector^1.000000), product of:", "children": [ { "value": 1, "message": "boost" }, { "value": 0.5, "message": "queryNorm" } ] }, { "value": 0.25, "message": "fieldWeight(desc in doc one), score of:", "children": [ { "value": 0.25, "message": "vector(field(desc:one) with similarity_metric(dot_product)=0.250000" } ] } ] }
3b9f58d
to
6475b83
Compare
Pls review this PR instead - #1899 |
This PR adds a score Explanation to the KNN scorer, along with a unit test.
Also, contains a minor refactor of scoring related code.