lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Yonik Seeley (JIRA)" <>
Subject [jira] Commented: (LUCENE-2617) coord should still apply to missing terms/clauses
Date Tue, 24 Aug 2010 14:26:16 GMT


Yonik Seeley commented on LUCENE-2617:

bq. Are we sure that non-existent terms/phrases should contribute to coord?

Yep :-)

Adding an unrelated document and having scores radically change on other docs isn't a good
It also makes testing and relevancy tuning hard (hoss & I were really scratching our heads
last night trying to figure out why coord wasn't being applied).  idf is currently the only
similarity factor that varies with index size + content... adding coord to that list wouldn't
be good.

Actually, I just thought of something potentially *much* worse.  With per-segment searching,
doesn't this make coord vary per-segment?  That means we're adding up scores that aren't comparable...
and all of a sudden this is looking like a much worse bug.  A match in a very small segment
is likely to contribute much more to a score than a match in a large segment.

> coord should still apply to missing terms/clauses
> -------------------------------------------------
>                 Key: LUCENE-2617
>                 URL:
>             Project: Lucene - Java
>          Issue Type: Bug
>    Affects Versions: 4.0
>            Reporter: Yonik Seeley
>         Attachments: LUCENE-2617.patch
> Missing terms in a boolean query "disappear" (i.e. they don't even affect the coord factor).

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message