Re: problems with concise bounded descriptions

* Patrick.Stickler@nokia.com <Patrick.Stickler@nokia.com> [2004-10-01 14:27+0300]
> > > We do not assert that it is perfect, either for any particular
> > > application, or for even for a majority of applications. 
> > 
> > I'm happy to see the idea written up, and I think it'll find 
> > a niche in
> > certain applications.
> > 
> > Re 'perfect', http://www.w3.org/Submission/2004/SUBM-CBD-20040930/
> > does say (in the abstact), 
> > 
> > 	This document defines a concise bounded description of 
> > a resource in
> > 	terms of an RDF graph, as an optimal unit of specific 
> > knowledge about
> > 	that resource to be utilized by, and/or interchanged 
> > between, semantic
> > 	web agents.
> > 
> > ...where 'optimal' suggests a certain comfort with the design, on my
> > reading of http://dictionary.reference.com/search?q=optimal
> 
> I *do* assert that CBDs are *an* optimal unit of specific knowledge.
> 
> I do *not* assert that CBDs are either *the* optimal unit of knowledge,
> or a *perfect* unit of knowledge.
> 
> There may be equally optimal units of knowledge for similar applications.
> 
> There are surely ways that the definition of CBDs can be improved.

OK I'm not going to quibble over that one word, beyond noting that
'optimal' to me suggests 'the' rather than 'a'. 
> 
> > The point about foaf:maker/foaf:made (and depicts/depiction) was just
> > that there is an asymmetry in the design of the RDF syntax, since it
> > projects directedness of RDF arcs on to nestedness of XML 
> > elements. This should
> > be of no consequence to those working with the RDF model, in theory.
> > 
> > However in practice, we find that designers of RDF vocabs feel the
> > likely RDF/XML encoding of instance data using their properties is a
> > (perhaps minor) design constraint on their property naming 
> > choices. CBD
> > has a similar asymmetry, treating a graph built from 'depicts'
> > differently from another couched in terms of 'depiction', 
> > despite their
> > being true description of the world under pretty much(*) the same
> > circumstances. 
> 
> I consider inference to be the better way to address issues
> of symmetry of knowledge (possibly implicit) versus asymmetry
> of presentation/expression.
> 
> Thus, an asymmetrical presentation can neverhtheless reflect
> a (subset of) fully symettrical knowledge.
> 
> It's been my experience that resource-as-subject-only (asymmetrical)
> knowledge is far easier to work with when directing particular
> tasks/processes than having to work with resource-as-subject-or-object
> (symmetrical knowledge).
> 
> Of course, that's just my own experience, and I don't at all mean
> to suggest that it constitutes the usual experience.
>  
> > My concern then, was just that CBD would introduce yet
> > another factor into RDFS vocab design, actually a very similar bias to
> > that already associated with the RDF/XML syntax: vocab designers would
> > have to think more carefully about the direction in which they name
> > their RDF properties, even though the pure RDF graph view of this
> > suggests they shouldn't have to.
> 
> Actually, if the designers simply capture the symmetric relations
> between properties, then using inference, one can decide freely
> whether a symmetric or asymmetric perspective is most useful.

OK, I can see that working for tools that deal with inverses. Running
CBD over system that understands inverses could be differently optimal
to using it over a raw RDF graph store...

Dan

Received on Friday, 1 October 2004 11:34:31 UTC