[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [help-GIFT] algorithms--weighting functions-methods
From: |
Wolfgang Müller |
Subject: |
Re: [help-GIFT] algorithms--weighting functions-methods |
Date: |
Mon, 30 Apr 2007 11:12:04 +0200 |
User-agent: |
KMail/1.9.1 |
Hi,
> firstly, When you say change the algorithm, you mean that change the
> weighting function , right?
Depends on the exact context that I am lacking now. For doing runtime indexing
changing the weighting function is unrelated. If you want to index text, then
probably all you want is to add features in the feature extraction step.
> I think there are 3 place to be changed when some one want to apply
> a new algorithm, they are:
>
> 1, double CWeightingFunction::subApply(const double
> inDocumentFrequency,
> const double inNormalizingFactor)
>
> 2, void CScoreBoard::operator()(const CWeightingFunction&
> inWF,
>
> const CDocumentFrequencyElement& inDFE)
>
> 3, double CWeightingFunction::getTermFrequency()
>
>
>
> Am I right?
If you want to slightly change the current main algorithm of Viper, these are
probably the places.
> Secondly , in gift-config.mrml, there are algorithm such as "Separate
> Normalisation" ,"Perl link" . But I cant find the code for these
> algorithms, Did Viper implement them?
I the GIFT package you unpacked and installed at your place should contain
folders such as libGIFTAcPerl and libGIFTQuPerl. This is the code that links
GIFT to perl.
The separate normalization is done by specifying several "sub-algorithms" that
are "vanilla" viper algorithms. The results of these then are combined into
one result.
> Thirdly, what is the relationship between algorithm and weighting function
> ? And what's the difference? In mrml tag <algorithm>, there are
> information for both of them.
Weighting function is one parameter that influences the weighting function in
inverted-file-based query processing.
The <algorithm> tag specifies the query engine that is to be used. The query
engine will be told a collection then will seek to open the necessary
accessor given the collection.
> The last, what tag in mrml does the Server use to determine which query
> engine to use ? Can one algorithm be used by two query engine?
<algorithm>
In a running GIFT there is one instantiated query engine per session.
Potentially several query engines share one accessor.
Best,
Wolfgang
--
Dr. Wolfgang Müller
LS Medieninformatik
Universität Bamberg
Check out the SIG MM web site http://www.sigmm.org