Re: Guide/Lang: UML as a presentation syntax for OWL

I agree with most of what Chris has written w.r.t. UML (although I
don't pretend to be an expert). I disagree, however, with the use of
the term "primitive" to describe RDF (or DAML+OIL/OWL) classes.

It is not true even of RDF that it "does not provide a way to specify
sufficient conditions for class membership" (as stated in the
compliance level 1 document). E.g., if class C is the domain of
property P, then from P(x,y) I can infer C(x). Of course it is "even
less true" of DAML+OIL/OWL: having a "primitive definition" (i.e., a
subClass axiom) in one place doesn't preclude the possibility of
specifying sufficient conditions elsewhere in the ontology. For this
reason, I believe that terms like "primitive" and "defined" should be
avoided.

Ian


On May 15, Christopher Welty writes:
> In reference to 
> http://www.swi.psy.uva.nl/usr/Schreiber/docs/owl-uml/owl-uml.html
> 
> (are there reasonable tools for suggesting changes in HTML like there is 
> in MS Word??
> 
> 
> My comments:
> 
> As I mentioned in my previous note 
> (http://lists.w3.org/Archives/Public/www-webont-wg/2002May/0100.html), I 
> disagree with the note [LANG note: proposal to drop "primitive classes" 
> and use "class" instead. This implies that, by default, classes are 
> primitive.]   Since UML will only be able to specify primitive OWL 
> classes, I believe we should be specific about this in the document.  Thus 
> the first paragraph of the section:UML notations for OWL Lite, 
> subsection:Classes and subclasses should read: 
> "A primitive OWL class is shown in a similar fashion as a UML class. Note 
> that ontologies typically do not specify class behavior, so the operation 
> compartment will stay empty. Classes may have primitive subclasses."
> 
> The second sentence is not strictly true, since in OO languages most of 
> what we can specify in OWL, such as inverses, transitivity, functional 
> roles, etc. would need to be implemented.  "Inference" per se is not 
> something one can specify directly in UML, and we may want to have some 
> conventions of our own for specifying standard inferences - from a UML 
> perspective I think these things would belong in the "class behavior" 
> section.
> 
> Let's try to use decent examples, because they become canonical.  Male and 
> Female are not subclasses of Animal.  Let's change the subclasses to 
> "Mammal" and "Reptile"
> 
> Some details of the level 1 language should probably be worked out before 
> we decide on many of the finer points.
> 
> We should show the precise OWL for each UML diagram.
> 
> The Animal has-parent Animal diagram is confusing, first of all my 
> understanding of UML was that associations to the class require a single 
> box, with a line going from that box out and then back.  Second, this is a 
> cycle - does OWL allow cycles?
> 
> That's as far as I got.
> 
> -Chris
> 
> Dr. Christopher A. Welty, Knowledge Structures Group
> IBM Watson Research Center, 19 Skyline Dr.
> Hawthorne, NY  10532     USA 
> Voice: +1 914.784.7055,  IBM T/L: 863.7055
> Fax: +1 914.784.6078, Email: welty@us.ibm.com

Received on Thursday, 16 May 2002 16:03:49 UTC