Different similarity for different types of fields

[maximd <ma...@softlibsw.com>]

Hello,
In my scenario I want to categorize document fields by field types and treat
each field type differently in the meaning of score calculation. To be more
specific, I want to categorize the fields to important fields and
less-important fields, my query will contain both important fields and
less-important ones but I want to boost those documents containing high
coverage of important fields. In addition, for important fields I want to
provide special idf function. Obviously I'll boost an important fields with
the boost factor, but I'm wondering if I could also rewrite Similarity.coord
and Similarity.idf to better reflect my situation. Unfortunately, I don't
see how just implementing custom Similarity would help as there is no
relevant information about field there. Is there any other approach I could
use? Is it possible to have 2 similarities (both for indexing and searching)
so I'll split my search to first search in important fields and then search
in less important fields and combine the results?

Thanks in advance,
Maxim
-- 
View this message in context: http://old.nabble.com/Different-similarity-for-different-types-of-fields-tp26863215p26863215.html
Sent from the Lucene - Java Users mailing list archive at Nabble.com.


---------------------------------------------------------------------
To unsubscribe, e-mail: java-user-unsubscribe@lucene.apache.org
For additional commands, e-mail: java-user-help@lucene.apache.org