Score per Term

[Andrzej Bialecki <ab...@getopt.org>]

Hello,

Is there any simple way to get the information from the search results 
on which of the query terms contributed the most to the document's 
score? I'm working on an application which could use this sort of 
information to give a hint to the user why particular document scores 
the way it does against particular query...

Any suggestions?

-- 
Best regards,
Andrzej Bialecki

-------------------------------------------------
Software Architect, System Integration Specialist
-------------------------------------------------
FreeBSD developer (http://www.freebsd.org)



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

Re: Score per Term

[Doug Cutting <cu...@lucene.com>]

Andrzej Bialecki wrote:
> Do you think it would be possible/feasible to modify the 
> searching classes so that they create Explanations at the same time I'm 
> running the query?

That's not feasable because it would slow down query execution too much.

Doug


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

Re: Score per Term

[Andrzej Bialecki <ab...@getopt.org>]

Doug Cutting wrote:
> Check out the new Explanation API in the latest CVS sources.  It permits 
> one to get a detailed explanation of how a query was scored against a 
> document.  Note that these explanations are designed for user perusal, 
> not for further computation, and are as expensive to construct as 
> re-running the query, but it sounds like that fits your needs.

Wow, that was fast... :-) Yes, it seems this is the functionality I'm 
looking for. Do you think it would be possible/feasible to modify the 
searching classes so that they create Explanations at the same time I'm 
running the query? Should I expect a similar cost for that as creating 
Explanations separately?

BTW: I tried to contact you regarding some help in a commercial project. 
Is <do...@lucene.com> the right way to do it?

Thanks!

> Andrzej Bialecki wrote:
> 
>> Hello,
>>
>> Is there any simple way to get the information from the search results 
>> on which of the query terms contributed the most to the document's 
>> score? I'm working on an application which could use this sort of 
>> information to give a hint to the user why particular document scores 
>> the way it does against particular query...
>>
>> Any suggestions?

-- 
Best regards,
Andrzej Bialecki

-------------------------------------------------
Software Architect, System Integration Specialist
-------------------------------------------------
FreeBSD developer (http://www.freebsd.org)



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

Re: Score per Term

[Doug Cutting <cu...@lucene.com>]

Check out the new Explanation API in the latest CVS sources.  It permits 
one to get a detailed explanation of how a query was scored against a 
document.  Note that these explanations are designed for user perusal, 
not for further computation, and are as expensive to construct as 
re-running the query, but it sounds like that fits your needs.

Doug

Andrzej Bialecki wrote:
> Hello,
> 
> Is there any simple way to get the information from the search results 
> on which of the query terms contributed the most to the document's 
> score? I'm working on an application which could use this sort of 
> information to give a hint to the user why particular document scores 
> the way it does against particular query...
> 
> Any suggestions?
> 


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