lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Chris Hostetter <hossman_luc...@fucit.org>
Subject Re: Lucene performance bottlenecks
Date Mon, 12 Dec 2005 23:35:54 GMT

: Oh, BTW:  I just found the DisjunctionMaxQuery class, recently added it
: seems. Do you think this query structure could benefit from using it
: instead of the BooleanQuery?

DisjunctionMaxQuery kicks ass (in my opinion), and It certainly seems like
(from your query structure) it's something you might want to consider
using, but I don't know thta it will sove the performance problems you're
having -- I can't think of any situations in which DisjunctionMaxScorer
could skip more docs/terms then DisjuntionSumScorer.

I believe Paul already suggested writting a completely new type of
specialized Query/Scorer for your purposes -- one possibility I can
imaging is a specialized version of BooleanQuery that skips past documents
in the optional scorers once it's gotten a score from at least one other
optional scorer (or N other optional scorers).  Then as long as you're
carefull to list the really usefull sub-clauses (ie: the shorter fields,
with the higher boosts) first, you can save a lot of calculation with
(what i expect would be)  little loss in functionality.

for example, looking at your sample query...

+(url:term1^4.0 anchor:term1^2.0 content:term1 title:term1^1.5 host:term1^2.0)
+(url:term2^4.0 anchor:term2^2.0 content:term2 title:term2^1.5 host:term2^2.0)
url:"term1 term2"~2147483647^4.0
anchor:"term1 term2"~4^2.0
content:"term1 term2"~2147483647
title:"term1 term2"~2147483647^1.5
host:"term1 term2"~2147483647^2.0

...if you reordered those first two clauses like this...

+(url:term1^4.0 anchor:term1^2.0 host:term1^2.0 title:term1^1.5 content:term1)

...once the Scorers from the url term query and the achor term query
have chimed in with a non-zero score for a document, the scores from the
other sub-clauses maynot be very relevant in the final score calculation
... so why not skip them?





-Hoss


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


Mime
View raw message