msokolov commented on issue #13611: URL: https://github.com/apache/lucene/issues/13611#issuecomment-3346709689
These approximate algorithms generally report recall (at K) as the proportion of the (true) top K documents that are returned by the algorithm (averaged over a number of queries). Typical numbers are in 0.9-0.95 -- you can tune for higher recall, but it can be challenging to achieve very high recall. So you should expect some expected documents to go missing. It's hard to be more precise without better statistical data - a single document missing doesn't really tell us much about whether there is a bug or not. Having said that I'd be curious if you can check with your data how the "regular" KNN query performs as compared to the parent-join version. Does it also fail to find document #29? Also -- what rank is it among all the child documents (as opposed to among all the parents after joining)? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
