Mailing-List: contact lucene-dev-help@jakarta.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "Lucene Developers List" <lucene-dev@jakarta.apache.org>
Received-SPF: pass (hermes.apache.org: local policy)
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Subject: RE: Normalized Scoring 
Date: Thu, 21 Oct 2004 00:09:32 -0700
Message-ID: 
 <E3381E0825F1954D953A4E347017BB1C02630DB7@reh001-1.REX001.ExchangeByRegister.com>
Thread-Topic: Normalized Scoring 
Thread-Index: AcS2y0FPEowRC5XCTii7OcxVLwqTkwAb4hAQ
From: "Chuck Williams" <chuck@manawiz.com>
To: "Lucene Developers List" <lucene-dev@jakarta.apache.org>

Thanks Paul.  These places are all mentioned in my proposal to fix this
directly in Lucene.  I don't see a way to accomplish this without
modifying Lucene, e.g. only through subclassing, but will take your
suggestion and look into that possibility.

Chuck

  > -----Original Message-----
  > From: Paul Elschot [mailto:paul.elschot@xs4all.nl]
  > Sent: Wednesday, October 20, 2004 10:40 AM
  > To: lucene-dev@jakarta.apache.org
  > Subject: Re: Normalized Scoring
  >=20
  > Chuck,
  >=20
  > Hits normalizes the final highest score to 1.0, and you
  > can mplement your own HitCollector to suppress
  > that normalisation.
  >=20
  > For the rest have a look at Weight, it can easily be used
  > for your example by having sumOfSquaredWeights()
  > return some sum of the weights, and letting normalize()
  > do the divisions into the weights of the individual TermQuery's.
  >=20
  > Many Lucene queries implement their Weight in an inner class.
  > It is used as an interim step to get from a Query to a Scorer.
  >=20
  > Kind regards,
  > Paul Elschot
  >=20
  > On Wednesday 20 October 2004 07:56, Chuck Williams wrote:
  > ...
  > >
  > > As an example of the simple boost-based normalization, for the
query
  > >   ((a^2 b)^3 (c d^2))
  > > the net boosts are:
  > >   a --> 6
  > >   b --> 3
  > >   c --> 1
  > >   d --> 2
  > >
  > > So if a and b matched, but not c and d, in the top scoring result,
its
  > > score would be 0.75.  The normalizer would be 0.75/(current score
  > except
  > > for the current normalization).  This normalizer would be applied
to
  > all
  > > current scores (minus normalization) to create the normalized
scores.
  > >
  > > For simple query (a b), if only one of the terms matched in the
top
  > > result, then its score would be 0.5, vs. 1.0 or many other
possible
  > > scores today.
  > >
  >=20
  >=20
  >
---------------------------------------------------------------------
  > To unsubscribe, e-mail: lucene-dev-unsubscribe@jakarta.apache.org
  > For additional commands, e-mail: lucene-dev-help@jakarta.apache.org


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