lucene-java-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ype Kingma <ykin...@xs4all.nl>
Subject Re: Query Boosting
Date Sat, 02 Nov 2002 14:48:48 GMT
On Saturday 02 November 2002 05:09, Harpreet S Walia wrote:
> Hi ,
>
> I am expecting my index to go into millions(5-10) of documnets very soon ,
> so would using a limiting filter pose any problems for such kind of a
> index.

Apart from java object overhead once, a limit filter takes one (1) bit per 
document, ie. 125 bytes per 1000 docs, in your case 1.25Mbytes per filter.
You can reuse search filters as long as the document numbers in the index
do not change.
A LRU cache for filters can be worthwhile. Java 1.4 has a LinkedHashMap for 
that.

Have fun,
Ype


>
> Thanks And Regards,
> Harpreet
>
> ----- Original Message -----
> From: "Ype Kingma" <ykingma@xs4all.nl>
> To: "Lucene Users List" <lucene-user@jakarta.apache.org>;
> <harpreet@sansuisoftware.com>
> Sent: Friday, November 01, 2002 12:05 AM
> Subject: Re: Query Boosting
>
> > On Thursday 31 October 2002 18:45, harpreet@sansuisoftware.com wrote:
> > > Hi,
> > >
> > > My application requires a facility to have security build into the
> > > documents so that when i search for a given word depending on the
>
> security
>
> > > credentials stored in a field in the document the results are filtered
> > > .
> > >
> > > Now the problem i am facing is that the score of such results includes
> > > these security credentials in the query in addition to the query
> > > entered
>
> by
>
> > > the user. So the relevancy according to the actual search word entered
>
> by
>
> > > the use is affected .
> > >
> > > One way that i see to avoid this is by boosting the words entered by
> > > the user and reducing the boost for the security credentials.
> > >
> > > I tried something like :
> > > <User Query>^1 AND <Security Credentials>^0.1
> > >
> > > This seems to work for the current set of data which i have .
> > >
> > > Is this the right approach or is there any other better away to handle
> > > this.
> >
> > It's ok.
> > You can also use a limit filter to limit the results. This does not
> > affect scoring at all. See java docs on IndexSearcher and Filter in
> > package lucene.search.
> >
> > Regards,
> > Ype
> >
> >
> > --
> > To unsubscribe, e-mail:
>
> <mailto:lucene-user-unsubscribe@jakarta.apache.org>
>
> > For additional commands, e-mail:
>
> <mailto:lucene-user-help@jakarta.apache.org>

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


Mime
View raw message