(11/03/03 2:54), Mark wrote:
High level overview. We have items and we have sellers. The scoring of our
documents is such that
our boost functions outweight the pure lucene term/query scoring. Our boost
functions basically take
into account how "good" the seller is.
Now for MLT searches we would like to incorporate this same sort of behavior.
Sounds reasonable. You can open an issue (on Lucene, I think).
Just for your infomation, MLT in 3x/trunk debug has more rich information
(assembled queries and explanation):
https://issues.apache.org/jira/browse/SOLR-860
It doesn't solve your problem, but you'd like to try...
Koji
--
http://www.rondhuit.com/en/
