lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From dfl <dennis....@usps.gov>
Subject Determining Relevancy/Cutoff
Date Thu, 07 Aug 2014 15:46:31 GMT
What is the best way to determine relevancy and the cutoff of results to
show? 

So the system I'm working on right now involves searching the inventory and
returning the results. Each result must be reviewed by an employee to
determine whether it is a true match. Obviously, we want to minimize the
number of false results we return. 

I've been tweaking boosts and stuff to get it to score better, but we still
have a few problems with determining relevancy. 

An absolute threshold doesn't work because search scores are only meaningful
relative to the results in a given query. So a score of 200 on one query may
not be as relevant as a score of .2 on another. 

The other method I've seen is a score normalized with respect to the top
score of a query. Then we can return all results at are within x% of that
score. However, if there are no good results, then the top result is very
poor, and all the results we return will be poor. 

How can I determine which documents are relevant and which are not? 





--
View this message in context: http://lucene.472066.n3.nabble.com/Determining-Relevancy-Cutoff-tp4151714.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


Mime
View raw message