fix(search): enable true semantic recall for hybrid/multi ranking - #419
Open
Mardi123 wants to merge 1 commit into
Open
fix(search): enable true semantic recall for hybrid/multi ranking#419Mardi123 wants to merge 1 commit into
Mardi123 wants to merge 1 commit into
Conversation
Semantic ranking could only re-rank BM25 keyword hits: a primitive that was embedding-similar but shared no query token could never be retrieved, because the scoring candidate set was seeded only from BM25 term matches. Seed every filtered candidate when the query carries embeddings so the dense signal drives recall, not just re-ordering. Also: - Clamp cosine contributions to [0, 1] so an unrelated (negative-cosine) embedding cannot penalise a hit below its BM25 score. - Unify -ies/-ied plurals with their -y singular in the stemmer (queries<->query, studied<->study), which previously never matched. Adds focused tests for pure-semantic recall and stemmer unification.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Semantic ranking could only re-rank BM25 keyword hits: a primitive that was embedding-similar but shared no query token could never be retrieved, because the scoring candidate set was seeded only from BM25 term matches. Seed every filtered candidate when the query carries embeddings so the dense signal drives recall, not just re-ordering.
Also:
Adds focused tests for pure-semantic recall and stemmer unification.
Description
Type of Change
Related Issues
Closes #
Fixes #
Relates to #
Changes Made
Testing
Test Coverage
Manual Testing Steps
Tested On
macOS
Windows
Linux
VS Code Stable
VS Code Insiders
Screenshots
Checklist
Documentation
Additional Notes
Reviewer Guidelines
Please pay special attention to:
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache License 2.0.