[
https://issues.apache.org/jira/browse/SOLR-5787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14118747#comment-14118747
]
Nalini Kartha commented on SOLR-5787:
-------------------------------------
[~jdyer] Not quite - we want to allow for some terms (mis-spelled and correctly
spelled) not occurring in the documents so the collation hit counts with mm set
to 100 would be misleading or not helpful.
Say for eg. the query is 'classic stor' - the term 'classic' does not exist in
the index at all but the term 'store' exists in 1000 documents in the index,
100 of which satisfy the fq params and the term 'story' exists in 2000
documents in the index, 10 of which satisfy the fq params.
The collations 'classic store' or 'classic story' with mm=100 would have 0 hits
because 'classic' doesn't exist in the index. In this eg. we want to know that
'store' is more frequent in the index compared to 'story' - when taking into
account the fq params - so that we can choose 'store' as the correction. But
the frequencies in the extended results and the collation hits don't give us
the information we're looking for.
I realize that our use case is a bit strange but in general, having the
frequencies returned in extended results reflect the fq params seems useful and
more accurate. What do you think?
> Get spellcheck frequency relatively to current query
> ----------------------------------------------------
>
> Key: SOLR-5787
> URL: https://issues.apache.org/jira/browse/SOLR-5787
> Project: Solr
> Issue Type: Improvement
> Components: spellchecker
> Affects Versions: 4.6
> Environment: Solr deployed on Jetty 9 Servlet container
> Reporter: Hakim
> Priority: Minor
> Labels: features, newbie
>
> I guess that this functionnality isn't implemented yet. I'll begin by an
> example to explain what I'm requesting:
> I have a lucene query that get articles satisfying a certain query. With this
> same command, I'm getting at the same time suggestions if this query doesnt
> return any article (so far, nothing unusual).
> The frequency (count) associated with these suggestions is relative to all
> index (it counts all occurences of the suggestion in the whole index). What I
> want is that it counts only suggestion occurences satisfying current lucene
> query.
> P.S: I'm using solr's spellcheck component (solr.DirectSolrSpellChecker).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
