Consider add ranking features for fuzzy query operator

vespa-engine / vespa

AI + Data, online. https://vespa.ai

https://vespa.ai

Apache License 2.0

5.68k stars 592 forks source link

Consider add ranking features for fuzzy query operator #24242

Open jobergum opened 2 years ago

jobergum commented 2 years ago

In https://github.com/vespa-engine/vespa/issues/13814 we added support for fuzzy query with maxEditDistance, for larger values, it would make sense to expose the distance as a rank-feature, for example, rawScore() or a dedicated new rank-feature.

geirst commented 1 year ago

Matches features are missing as well. Example: https://stackoverflow.com/questions/74053567/ranking-in-fuzzy-match-in-vespa.

vekterli commented 4 months ago

I'm resetting the milestone on this issue in light of recent work on fuzzy prefix support (#30720). Fuzzy prefix matching can have a lot more matches than non-prefix fuzzy matching, so exposing the edit distance as a rank feature (rawScore most likely) is all the more useful.