You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Alexandre Rafalovitch (JIRA)" <ji...@apache.org> on 2018/07/23 18:36:00 UTC
[jira] [Commented] (SOLR-12581) add a "min_popularity" option to
relatedness() aggregation that forces scores to -Inf if fg/bg pops don't
meet a threshold
[ https://issues.apache.org/jira/browse/SOLR-12581?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16553241#comment-16553241 ]
Alexandre Rafalovitch commented on SOLR-12581:
----------------------------------------------
Sort of a side-question, but this work seems to overlap/compliment the significantTerms work done for streaming/QueryParser: [http://lucene.apache.org/solr/guide/7_4/stream-source-reference.html#significantterms]
Are we saying SignificantTerms is for simpler use cases (as fore/back queries are corpus-wide) and then go into relatedness() for more complex analysis?
Should the options be roughly compatible where it makes sense and/or similarly named?
Just wondering because I could see this confusing newbies trying to see when to use which option.
> add a "min_popularity" option to relatedness() aggregation that forces scores to -Inf if fg/bg pops don't meet a threshold
> --------------------------------------------------------------------------------------------------------------------------
>
> Key: SOLR-12581
> URL: https://issues.apache.org/jira/browse/SOLR-12581
> Project: Solr
> Issue Type: Improvement
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Hoss Man
> Assignee: Hoss Man
> Priority: Major
> Attachments: SOLR-12581.patch
>
>
> as discussed in SOLR-9480 and noted in TODO comments, the original SKG code base had a "min_pop" option that would completely ignore "terms" if the fg/bg popularities weren't above a user specified threshold. with the implementation of SKG as a {{relatedness()}} aggregation function, we need to leave any actual filtering of buckets by an aggregation result to a future generalized JSON facet enhancement, but what we can do today is implement an optional {{min_popularity}} option on {{relatedness()}} that forces the {{relatedness}} score to -Infinity so buckets that don't meat the threshold at least score "as low as possible"
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org