[ 
https://issues.apache.org/jira/browse/SOLR-1493?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12762783#action_12762783
 ] 

Shalin Shekhar Mangar commented on SOLR-1493:
---------------------------------------------

Jason, yes SpellingQueryConvertor blindly splits by whitespace. But if you use 
spellcheck.q for your query, the corresponding Solr field's analyzer will be 
used. So I think your use-case can also be solved by creating your dictionary 
from a Solr field which has a KeywordTokenizer and using spellcheck.q parameter.

> Provide a non delimiting SpellingQueryConverter
> -----------------------------------------------
>
>                 Key: SOLR-1493
>                 URL: https://issues.apache.org/jira/browse/SOLR-1493
>             Project: Solr
>          Issue Type: Improvement
>    Affects Versions: 1.3
>            Reporter: Jason Falk
>            Priority: Minor
>             Fix For: 1.4
>
>         Attachments: SOLR-1493.patch
>
>
> The current SpellingQueryConverter spell checks individual words only, but in 
> the case of products or names, it is sometimes better to spell check groups 
> of words together.  For example if you are searching a person's name, you 
> might want to compute the edit distance against the whole person's name, not 
> individual words.  For example:
> If I search for Jonny Cash with the current SpellingQueryConverter, it won't 
> suggest any fixes cause we have both an artist with the name Jonny in it and 
> of course the artist who we really want, Johnny Cash.  If we don't delimit 
> the words, it will realize Jonny Cash as a whole doesn't exist and will 
> return Johnny Cash instead as a did you mean.  
> The other advantage of this is that it gets rid of the possibility of the did 
> you mean suggesting a spell correction for one of the two (or more) words 
> that also doesn't exist.  Let's say hypothetically we searched for Jonny Cash 
> again, the did you mean might currently suggest Jinny Cash, who also doesn't 
> exist.  If we don't delimit the words going into the spellchecker, this 
> shouldn't happen.

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

Reply via email to