Mailing-List: contact lucene-user-help@jakarta.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "Lucene Users List" <lucene-user@jakarta.apache.org>
Received-SPF: pass (hermes.apache.org: domain of mark@weballistics.com
 designates 82.68.192.116 as permitted sender)
Subject: Re: query boosting using a word list.
From: Mark Page <mark@weballistics.com>
To: Lucene Users List <lucene-user@jakarta.apache.org>
In-Reply-To: <20041106175225.79711.qmail@web12701.mail.yahoo.com>
References: <20041106175225.79711.qmail@web12701.mail.yahoo.com>
Content-Type: text/plain
Organization: WEBallistics
Message-Id: <1099770657.6724.59.camel@dev.weballistics.co.uk>
Mime-Version: 1.0
Date: Sat, 06 Nov 2004 19:50:58 +0000
Content-Transfer-Encoding: 7bit

Hi Otis,

thanks for the reply. I'll have to read up the javadocs a bit more
before I fully understand your answer. ;)

One thing I had noticed in the docs (and was half expecting to get
pointed towards) is the WordlistLoader class. I was thinking that I
could maybe use this and create a kind of 'reverse' of the stop-word
list (boosting instead of removing), or am I barking up the wrong tree?

regards, -Mark.

On Sat, 2004-11-06 at 17:52, Otis Gospodnetic wrote:
> Hello Mark,
> 
> It sounds like you could extend QueryParser and override one of the
> Query get***Query methods (getFieldQuery?), perhaps first calling the
> super method, and then adding a boost based on the words, which you
> would look up in your implementation of the getFieldQuery method.
> 
> Otis
> 
> 
> --- Mark Page <mark@weballistics.com> wrote:
> 
> > Hi,
> > 
> > I have a database table of text flattend out and indexed.
> > 
> > Although searching with fuzzy query works well in most instances, on
> > occasions however the target record appears way down the list of
> > matching records.
> > 
> > This is because the query text may contain lots of irrelevant terms
> > (in
> > the context of the app) because the data is pulled from another
> > source.
> > 
> > To solve this I need to create a word list, so that the terms that
> > are
> > important to the app are boosted in the search. as an example...
> > 
> > word list contains car manufacturers and models:-
> > ...
> > volkswagon
> > golf
> > polo
> > ...
> > 
> > query text = "gleaming white 2-door volkswagon golf"
> > 
> > search = "gleaming white 2 door volkswagon^9 golf^9"
> > 
> > I can use regexes to massage the raw query text, but was wondering if
> > there is a more elegant solution available within the Lucene API.
> > 
> > As a Lucene newbie any pointers or suggestion to solve what must be
> > quite a common scenario appreciated. 
> > 
> > Regards, -Mark. 
> > 
> > 
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: lucene-user-unsubscribe@jakarta.apache.org
> > For additional commands, e-mail: lucene-user-help@jakarta.apache.org
> > 
> > 
> 
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: lucene-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: lucene-user-help@jakarta.apache.org
> 
-- 
Mark Page
WEBallistics
tel/fax: +44(0)20 7704 9885


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