[
https://issues.apache.org/jira/browse/LUCENE-7996?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16208509#comment-16208509
]
Robert Muir commented on LUCENE-7996:
-------------------------------------
+1 we try to fix similarities that produce negative scores because it hurts
relevance, too. Classic example is an unmodified bm25 IDF, goes negative for
stopword-like terms.
In some cases formulas can not so easily be "hacked" to avoid these problems,
it results in a less robust ranking algorithm. For example in the case above,
if someone's stopword list isn't perfect for their collection, then queries
will suffer.
Currently the problematic algorithms just have big javadocs warnings. But I
think its ok to look at putting them in the sandbox or deprecatinoulg or
removing like that instead.
At the end of the day, we should at least add better tests, so we know about
the problems. Thats step 1 I think. I have some ideas.
> Should we require positive scores?
> ----------------------------------
>
> Key: LUCENE-7996
> URL: https://issues.apache.org/jira/browse/LUCENE-7996
> Project: Lucene - Core
> Issue Type: Wish
> Reporter: Adrien Grand
> Priority: Minor
>
> Having worked on MAXSCORE recently, things would be simpler if we required
> that scores are positive. Practically, this would mean
> - forbidding/fixing similarities that may produce negative scores (we have
> some of them)
> - forbidding things like negative boosts
> - fixing the scoring formula of some queries like {{BoostingQuery}} (which
> subtracts a score to another score) so that the end result may never be
> negative
> So I'd be curious to have opinions whether this would be a sane requirement
> or whether we need to be able to cope with negative scores eg. because some
> similarities that we want to support produce negative scores by design.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
