RE: [OEP] "Classes as values": comments on draft $swbp

I like the first proposal. The second one may be exactly appropriate in some circumstances, but I have a concern about it.  SKOS seems to be an ontology representing the important things needed to represent thesauri (i.e. it creates a sub-language for representing specific thesauri.)  That is to say, the classes in SKOS represent language constructs used to create thesauri, RATHER than representing things in the world that are of interest (e.g. lions).  This distinction is evident by using the name "LionConcept" rather than "Lion".  This use of RDFS is as a meta-language, used to define a thesaurus language.

So the concern I have is mixing up use of RDFS/OWL as an ontology language for representing domains directly, vs. a meta-language for representing other representation languages/notations. 

A close analogy of this would be to say, use OWL (or RDFS) to represent another ontology language. For example, we could create an OWL ontology with classes that referred to Flogic language constructs. e.g.  OWL classes named: 'FlogicCLass' and 'FlogicAttribute'...  This is a very different use of OWL, than representing domains directly.  It is certainly not the main intended use if OWL, and really is using OWL as a meta-language.


 -----Original Message-----
From: 	Miles, AJ (Alistair)  [] 
Sent:	Thursday, April 29, 2004 7:37 AM
To:	Uschold, Michael F; Jeremy Carroll;
Subject:	RE: [OEP] "Classes as values": comments on draft

Sorry, resending this correcting some N3 syntax mistakes ...

I believe the best way to express the fact that a particular image depicts a
thing which is a member of the class of Lions would be to say (this is the
FOAF model):

	a	AnimalImage;
	foaf:depicts	[a	Lion].

	a	owl:Class;
	subClassOf	Mammal.

Mammal	a	owl:Class.
AnimalImage	a	owl:Class.

The alternative way of expressing similar information is to use the
dc:subject property along with the SKOS model [2] for describing concepts
that are intended to act as 'subjects' or 'topics' for information

	a	AnimalImage;
	dc:subject	LionConcept.

	a	skos:Concept;
	skos:prefLabel	'Lions';
	skos:broader	MammalConcept.

	a	skos:Concept;
	skos:prefLabel	'Mammals';
	skos:narrower	LionConcept.

The SKOS vocab already defines a class 'Concept' and a set of properties for
organising concepts into a hierarchy, without demanding that the hierarchy
implies a subclass relationship.  I refer the WG to the document [2] which
outlines the SKOS-Core schema, although you should currently ignore the
final section on 'using SKOS-Core with DC and FOAF' as this will change
shortly to be in line with the model of usage that I have briefly described




Received on Thursday, 29 April 2004 16:23:07 UTC