- From: Matthew Horridge <matthew.horridge@cs.man.ac.uk>
- Date: Thu, 26 Jul 2007 14:35:59 +0100
- To: Owl Dev <public-owl-dev@w3.org>
On 26 Jul 2007, at 14:21, Evren Sirin wrote: > > On 7/25/07 10:13 AM, Matthew Horridge wrote: >> >> Revisiting the issue of declaring and typing, because it is >> causing problems - in particular backwards compatibility with OWL 1.0 >> >> Below is a message to the list from Evren Sirin >> >>> On 26 Jan 2007, at 20:22, Evren Sirin wrote: >>> >>> In OWL 1.0, there is not really a difference between declarations >>> and typing. Having a triple <p, rdf:type, owl:ObjectProperty> >>> constitutes its declaration (as on object property in this case). >>> I agree that requiring declaration for every resource is not a >>> good idea. OWL-DL requires every resource to be typed and it >>> turns out that many ontologies out on the Web fall into OWL-DL >>> expressivity but do not meet this requirement. But now are we >>> separating declarations from typing and say that declarations are >>> not required but typing still is? >>> >>> >>> And if I understand the mapping from RDF graphs to OWL 1.1 >>> correctly, an ontology that has just the above triple (or any any >>> number of rdf:type triples where the object is one of >>> owl:ObjectProperty, owl:DatatypeProperty, or owl:Class) will be >>> mapped to an empty OWL 1.1 ontology. I don't think this is a >>> desired result. >> >> I am in complete agreement with the last point. There are plenty >> (enough to cause problems) of ontologies that just consist of >> rdf:type triples such as A rdf:type owl:Class. As Evren points >> out, when parsed, these documents result in empty ontologies, >> which is less than desirable - users of tools such as editors find >> this confusing and don't expect it. Does anyone have any >> suggestions about how to resolve this? > My suggestion (which I might have mentioned in the past) is to drop > owl11:declaredAs keyword completely and use rdf:type instead. In > the RDF/XML mapping, the triple C rdf:type owl:Class would be > mapped to Declaration(OWLClass(C)) directly (similarly for > properties, individuals and datatypes). I completely agree. Matthew
Received on Thursday, 26 July 2007 13:36:14 UTC