Re: less-restrictive range and domain terms

From: David Menendez <zednenem@psualum.com>
Subject: Re: less-restrictive range and domain terms
Date: Tue, 04 May 2004 16:58:48 -0400

> Peter F. Patel-Schneider writes:
> 
> > From: "Phil Dawes" <pdawes@users.sourceforge.net>
> > > 
> > > Benja Fallenstein writes:
> > >  > 
> > >  > Peter F. Patel-Schneider wrote:
> 
>          Phil Dawes:
> > >  > |>I've recently found myself wanting a less-restrictive version
> > >  > |>of rdfs:range (or owl:allValuesFrom) and rdfs:domain. I want
> > >  > |>to say 'property *can* have range of class foo' rather than
> > >  > |>'property *must* have range of class foo'.
> 
> > >  > | Second, what do you mean by 'property *may* have range of
> > >  > | class foo'?
> > >  > 
> > >  > "There exist triples with property P and an object of class
> > >  > foo," rather than "All triples with property P have objects of
> > >  > class foo," is a useful interpretation, I presume.
> > > 
> > > Yep - that's what I meant. Sorry for not being clear.
> > 
> > But suppose that there just doesn't happen to be such a relationship
> > in the current world.  What happens then?
> > 
> > This is not such a problem for property ranges (but does cause
> > problems even here), but what about an OWL construct like 
> >      the class foo may have range bar for property p
> > Does this mean that foo has to be non-empty?
> 
> I would say, either foo is non-empty or the "may have range" statement
> is incorrect.

Well, this would be one way to go.  

> However, for the purposes of UI support, which (I think) was Phil's
> original goal, there's no need to make any implications at the RDF
> level.

Why not?  Oh, you mean that the UI doesn't need to make inferences?  Maybe,
but then the UI won't warn you if you are making contradictions (which may
be possible in this extension of RDF).

> For example,
> 
>     p:owner s:usualRange p:HumanAgent .
>     p:Gorilla owl:disjointWith p:HumanAgent .
> 
> If someone was entering RDF data and said:
> 
>     _:fido p:owner [ a p:Gorilla ].
> 
> this could raise a warning, rather than an error. "Fido's owner is a
> gorilla, but owners are usually humans or groups of humans. Are you
> sure?"

Well, if that is all that you want to use this construct for, then
annotations (i.e., uninterpreted things dangling off a class) seem to be
the way to go.

> It's not *wrong*, just *unusual*.
> -- 
> David Menendez <zednenem@psualum.com> <http://www.eyrie.org/~zednenem/>

Peter F. Patel-Schneider
Bell Labs Research

Received on Tuesday, 4 May 2004 19:20:15 UTC