TOP and SCORE
Patrick Dowler
patrick.dowler at nrc-cnrc.gc.ca
Tue Jul 26 15:22:14 PDT 2005
Doesn't it make more sense to have the query parameter TOP=5 and not
use SCORE at all? The user wants to limit the number of results to the
N best results and the service decides which ones are the best using
some implementation detail. As Doug mentioned, since score values themselves
have no meaning outside the service, it isn't really useful to even output
them...
Thus, if one has to pick between SCORE and TOP as the query keyword
for limiting the number of output entries, TOP seems clear.
Patrick
PS-To be complete, we could also have BOTTOM, because sometimes you
want to get the worst results :-)
On 26.7.2005 12:38, Arnold Rots wrote:
> What I proposed from the start was that SCORE would be input as well
> as output parameter - and that's the way we have implemented it.
> SCORE can have two values on input: "ALL" (default) and "TOP".
> What I was suggesting does not require normalization, since
> SCORE=5 would be short for SCORE="TOP 5".
>
> And my objection remains. There is no other case where the name of
> the parameter is different in input and output - and SCORE and TOP
> refer to the same parameter.
>
> - Arnold
>
> Doug Tody wrote:
> > Hi Arnold -
> >
> > I am not sure where you got this from: the mention of "score" in
> connection
> > with SIA referred to this as a service capability not as a parameter
> > (along with logical names, compression, etc.). The query parameter
> > proposed in both cases is TOP. TOP=N returns the top scoring N
> records.
> > To have a query parameter SCORE could also be useful but this would
> only
> > work if the score were normalized in some fashion. In all the
> discussions
> > we have had of the SCORE heuristic the consensus has always been that
> > we don't want to try to normalize the score, mainly because the
> scoring
> > heuristic used is not defined except in general terms (intentionally,
> to
> > allow smart heuristics to be developed), and would vary from one
> service
> > implementation to the next, making it difficult to compare score
> values.
> >
> > - Doug
> >
> >
> >
> > On Tue, 26 Jul 2005, Arnold Rots wrote:
> >
> > > Doug,
> > >
> > > I noticed in your presentation at the NVO team meeting that there
> are
> > > two query parameters that overlap between SIAP and SSAP: SCORE and
> TOP.
> > > They represent the same concept; to amke it worse, "TOP" is an
> allowed
> > > value of SCORE. I think that is confusing and undesirable.
> > >
> > > I would suggest that TOP be dropped and SCORE be used in both cases.
> > > If it is considered important to allow a service to return a number
> of
> > > records other than 1 or ALL, one might consider allowing an integer
> > > value for SCORE (as for TOP), in addition to "ALL" and "TOP".
> > >
> > > Cheers,
> > >
> > > - Arnold
> > >
> > >
> ------------------------------------------------------------------------
> --
> > > Arnold H. Rots Chandra X-ray Science
> Center
> > > Smithsonian Astrophysical Observatory tel: +1 617
> 496 7701
> > > 60 Garden Street, MS 67 fax: +1 617
> 495 7356
> > > Cambridge, MA 02138
> arots at head.cfa.harvard.edu
> > > USA
> http://hea-www.harvard.edu/~arots/
> > >
> ------------------------------------------------------------------------
> --
> > >
> > >
> >
> ------------------------------------------------------------------------
> --
> Arnold H. Rots Chandra X-ray Science
> Center
> Smithsonian Astrophysical Observatory tel: +1 617 496
> 7701
> 60 Garden Street, MS 67 fax: +1 617 495
> 7356
> Cambridge, MA 02138
> arots at head.cfa.harvard.edu
> USA
> http://hea-www.harvard.edu/~arots/
> ------------------------------------------------------------------------
> --
>
--
Patrick Dowler
Tel/Tél: (250) 363-6914 | fax/télécopieur: (250) 363-0045
Canadian Astronomy Data Centre | Centre canadien de donnees astronomiques
National Research Council Canada | Conseil national de recherches Canada
Government of Canada | Gouvernement du Canada
5071 West Saanich Road | 5071, chemin West Saanich
Victoria, BC | Victoria (C.-B.)
More information about the dal
mailing list