RE: Assertions vs. declarations

Thank you, Bijan.

In "The even more irresistible etc." it is said that "the 
roles knows, hasSibling, and properPartOf, should be 
*declared* as, respectively, reflexive, irreflexive, and 
antisymmetric." (I think I found other examples of this 
use of "declare," but I don't remember where.) In the rest 
of the paper, however, reflexivity axioms etc. are called 
"role assertions."

I agree that (Tbox/Rbox) "axiom" is the correct term, and 
that "(class/property) assertion" should be reserved to 
Abox axioms, but it would be useful to have a more 
specialized term that covers exactly those property axioms 
that are called "role assertions" in the above mentioned 
paper.

Any suggestion?

Marco

-----Original Message-----
From: Bijan Parsia [mailto:bparsia@cs.man.ac.uk]
Sent: giovedì 27 marzo 2008 20.25
To: Marco Colombetti
Cc: public-owl-dev@w3.org
Subject: Re: Assertions vs. declarations

On 26 Mar 2008, at 17:00, Marco Colombetti wrote:

> I’m kind of dissatisfied with the use of the terms “assertion” and 
> “declaration” in the OWL literature.

Some pointers to where this happens would be helpful. I've 
not noticed these.

> Axioms like DisjointClasses(C,D) are often called “class assertions”, 
> and axioms like FunctionalObjectProperty(P) are called
> “(object) property assertions”.

I wouldn't do that. I'd call them "class axioms" and 
"property axioms". (I wouldn't *want* to do that...perhaps 
I've slipped
somewhere.) It's relatively harmless since they are 
declarative sentences typically presented in an 
assertional voice. But they are TBox and RBox axioms, not 
ABox :)

> Unfortunately, also Abox assertions like ClassAssertion(a C) are 
> called “class assertion”s, and Abox assertions like 
> ObjectPropertyAssertion(P a b) are called (object) “property 
> assertions”.

Yeah, that's ok by me.

> The former type of assertions are sometimes called “declarations”,

Where? That's evil.

> but unfortunately OWL has a different notion of declaration, 
> likeDeclaration(OWLClass(A)).

And "declaration" for tbox axioms is just *wrong* :)

> Any suggestion on how to call axioms like DisjointClasses(C,D) and
> FunctionalObjectProperty(P) without creating confusion?

"Class axioms" and "Property axioms" and use 
ClassAssertion and PropertyAssertion for the abox 
statements. Or TBox and RBox in the right contexts.

Cheers,
Bijan.

Received on Saturday, 29 March 2008 17:13:57 UTC