Re: rdfs:subClassOf on Annotation Properties for Dublin Core in OWL

This creates a weird situation that the WG should carefully consider (IMHO).

The ultimate idea is that the QName "dc:title" or "dc:source" should map 
to a URI reference that presumably contains a definition, we hope in 
OWL, but also perhaps a RDDL that points to an OWL.

So what are they? AnnotationProperties or regular Properties? I am not 
sure that "dc:source" should always be an annotation property. What if 
this property is attached to something other than an OWL class? Are you 
saying that we should have multiple Dublic Core ontologies specific to 
each class of things that they are being attached to? Perhaps, well then 
we really do need a RDDL between a namespace URI and an OWL definition 
so that we can switch between possibly 3 different ontologies (Lite, DL, 
Full) for each QName/URIref.

I am worried that this is going to cause a whole host of 
interoperability problems, and be downright confusing ( /me again 
following the rule of thumb that if I am confused, other folks will 
probably also be confused)

Perhaps we should forget AnnotationProperties and just say that OWL 
DL/Lite ignores things attached to owl:Classes that it doesn't 
understand -- i.e. just treat all unknown properties attached to an OWL 
DL/Lite class as an annotation....

Raphael Volz wrote:

>Hi -
>the dublin core elements enjoy a wide popularity as annotation properties.
>Since the RDF syntax for Dublin Core
>is based on a RDF Schema, all ontologies making use of that vocabulary are
>per definition
>OWL Full.
>I have edited the original definitions and created some OWL (Lite/DL, Full)
>ontologies that
>allow to use Dublin Core as annotation properties within OWL Lite/DL and
>Full ontologies without
>breaking the species through use of untyped properties and wrongly typed
>While doing this some unclarities with respect to rdfs:subPropertyOf arises,
>which is used
>in DC between properties.
>According to my understanding of the abstract syntax, it is not possible to
>super properties of annotation properities.
>Why is this so, what is the intention behind this decision ?
>Since annotation properties are pairwise disjoint from object and datatype
>properties allowing
>a super property axiom (which could be used/useful in OWL Full) appears to
>be no problem, right ?
>Btw., the ontologies are available here:
>OWL Lite/DL Dublin Core:
>OWL Full Dublin Core:
>I have decided to factor out all statements that make the DC declarations
>OWL Full into
>separate ontologies and include the OWL Lite/DL version via owl:imports.
>This appears
>to be a good design pattern for designing OWL Full ontologies since maximum
>reuse for
>non-Full "agents" is guaranteed.

Received on Tuesday, 2 September 2003 16:36:24 UTC