lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dan Miller <dan.mil...@goddard-associates.com>
Subject morelikethis - boosting on other field matches
Date Mon, 03 Jan 2011 21:12:17 GMT

This is a question about manipulating the ranking of morelikethis results.

For some original document, I am running a morelikethis query against a text
field in my documents. Another field contains an alphanumeric category code
(e.g. "A9Z"). I would like to boost the morelikethis score of documents
whose code matches the code in the original document.

I tried adding the code field as an mlt query field (and adding term vectors
at index time). But the code field never makes it into the list of
"interesting terms" so I don't get any boost there. 

Any suggestions?

If the above is possible, I'd eventually like to boost on the "closeness"
(in terms of edit distance) of the original field's code to the mlt results'
codes. Is this possible? (I could probably convert the codes to numbers if
that was helpful).

Finally, is there any documentation on how Solr/Lucene chooses "interesting
terms" in this context? If I could force the code field contents to be
"interesting" this would presumably do the job. The number different codes
is large but not infinite - maybe thousands or tens of thousands.

Thanks very much.
-- 
View this message in context: http://lucene.472066.n3.nabble.com/morelikethis-boosting-on-other-field-matches-tp2187671p2187671.html
Sent from the Solr - User mailing list archive at Nabble.com.

Mime
View raw message