RE: RE : How to use notations from classification schemes in SKOS

Dear Lars,

> >
> > Notations (such as 25F) are indeed the only things that are supposed
> > to be used to classify the documents. There is a text attached to
> > them (e.g., 'animals'), but this text, though very important (it
> > gives the natural meaning of the subject) will never be used as the
> > real descriptor for images. We had therefore something very
> > comparable to the prefLabel/altLabel distinction.
> Can you explain exactly how you solved the problem?

Simply
Notations -> prefLabel
Text (called 'textual correlates') -> altLabel
Actually some textual correlates (btw Stella do you approve of this
vocabulary ;-)) are quite complex ('David, from the roof (or balcony) of
his palace, sees Bathsheba bathing') and could be interpreted as tending
towards definitions or scope note. But many were only interpretable as
simple labels. altLabel was thus the only generic solution, and
performed OK in the following of our experiments.
 
> >
> >>>>>  - The value should be a typed literal, where the datatype
> >>>>> defines both
> > the scope of reference, and the lexical space of allowed values.
> >
> > Ok, and that could be a nice added value of such a property.
> > But beware that people do not waste thinking too much about
> > specifying the acceptable notations: IconClass for example allows
> > only for one or two letters in a notation, in a way that is not
> > easily specifiable even using regular formulas.
> > The effect would be even worse if designers realize at the end that
> > that was not really useful, or worse, is now preventing them to use
> > standard tools that will work with prefLabel. Perhaps this is
similar
> > to Stella's final argument in [4].
> 
> Summing the discussion up, I'd say that Alistair's proposal [1]
(thanks
> a lot, Alistair! You described it the way I would have done it) would
> probably not reach the necessary consensus. The question of course
> remains, how to handle this. What would you all say to the invention a
> new property skos:notation as a subproperty of both skos:prefLabel and
> dc:identifier? This way we could keep the prefLabel unique as stated
in
> the skos vocabulary specification [2]. 

I'm not enthusiast about it, for it raises the problem of relying on DC
specifications which may be not as normative as you want them to be. DC
website says 'Recommended best practice is to identify the resource by
means of a string or number conforming to a formal identification
system.'. You just have a best practice, and in theory lexical
prefLabels used as unique identifiers (which can be the case) could
therefore be interpreted as dc:identifiers, but not skos:notation.

Cheers,

Antoine

Received on Wednesday, 15 February 2006 11:00:37 UTC