MoreLikeThis with document that has not been indexed

[Ben Anhalt <be...@gmail.com>]

Hello,

It is currently possible to use the MoreLikeThis handler to find documents
similar to a given document in the index.

Is there any way to feed the handler a new document in XML or JSON (as one
would do for adding to the index) and have it find similar documents without
indexing the target document?  I understand that it is possible to do a MLT
query using free text, but I want to utilize structured data.

Thanks,

Ben

-- 
Ben Anhalt
ben.anhalt@gmail.com
Mi parolas Esperante.

Re: MoreLikeThis with document that has not been indexed

["David Smiley (@MITRE.org)" <DS...@mitre.org>]

Ben,
It's absolutely possible for MLT to find documents similar to another
indexed document. That's its primary use case.  For externally supplied
data, you will need to supply one blob of text.  You could derive this by
concatenating applicable parts of your structured data before handing to
Solr.  MLT can find similar documents using more than one indexed field,
which you may find useful.  

I have more info on MLT in my book, and then there's the wiki:
http://wiki.apache.org/solr/MoreLikeThis

~ David
Author: https://www.packtpub.com/solr-1-4-enterprise-search-server/book

 
Ben Anhalt wrote:
> 
> Hello,
> 
> It is currently possible to use the MoreLikeThis handler to find documents
> similar to a given document in the index.
> 
> Is there any way to feed the handler a new document in XML or JSON (as one
> would do for adding to the index) and have it find similar documents
> without
> indexing the target document?  I understand that it is possible to do a
> MLT
> query using free text, but I want to utilize structured data.
> 
> Thanks,
> 
> Ben
> 
> -- 
> Ben Anhalt
> ben.anhalt@gmail.com
> Mi parolas Esperante.
> 

-----
 Author: https://www.packtpub.com/solr-1-4-enterprise-search-server/book
--
View this message in context: http://lucene.472066.n3.nabble.com/MoreLikeThis-with-document-that-has-not-been-indexed-tp2754486p2757141.html
Sent from the Solr - User mailing list archive at Nabble.com.