You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Le Zhao (JIRA)" <ji...@apache.org> on 2016/02/11 08:48:18 UTC
[jira] [Comment Edited] (SOLR-6600) configurable relevance impact
of phrases for edismax
[ https://issues.apache.org/jira/browse/SOLR-6600?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15142363#comment-15142363 ]
Le Zhao edited comment on SOLR-6600 at 2/11/16 7:47 AM:
--------------------------------------------------------
Am I missing something here?
This issue is the exact opposite (or revert) of SOLR-6062, not a duplicate?
The SOLR-6062 behavior (summing from all fields instead of max or tie break) is not very desirable because phrase scores can increase disproportionately to unigram scores (which are limited by max or tie break), making it very hard to control/limit the influence of these phrases. Spurious bigram matches can easily bring false positives to the top of the rank.
was (Author: lezhao):
Am I missing something here?
This issue is the exact opposite (or revert) of SOLR-6062, not a duplicate?
The SOLR-6062 behavior (summing from all fields instead of max or tie break) is not very desirable because phrase scores can increase disproportionately to unigram weights (being controlled by max or tie break), making it very hard to control/limit the influence of these phrases. Spurious bigram matches can easily bring false positives to the top of the rank.
> configurable relevance impact of phrases for edismax
> ----------------------------------------------------
>
> Key: SOLR-6600
> URL: https://issues.apache.org/jira/browse/SOLR-6600
> Project: Solr
> Issue Type: Improvement
> Components: query parsers
> Affects Versions: 4.9
> Reporter: Alexey Kozhemiakin
> Labels: edismax
>
> Currently solr has a tie breaker parameter which control how to aggregate relevance score for search hits.
> But score for fields (pf, pf2, pf3) are always summed up.
> The goal of the patch is to wrap phrase clauses into single dismax clause instead of multipe ones
> Before patch
> +(
> DisjunctionMaxQuery((Body:james | Title:james)~tie_breaker)
> DisjunctionMaxQuery((Body:kirk | Title:kirk)~tie_breaker))
> )
> DisjunctionMaxQuery((Body:"james kirk")~tie_breaker)
> DisjunctionMaxQuery((Title:"james kirk")~tie_breaker)
> after patch
> +(
> DisjunctionMaxQuery((Body:james | Title:james)~tie_breaker)
> DisjunctionMaxQuery((Body:kirk | Title:kirk)~tie_breaker))
> )
> DisjunctionMaxQuery((Body:"james kirk" | Title:"james kirk") ~tie_breaker)
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org