Re: Urgent! Semantic question about rdfs:domain.

From: Ian Horrocks ([email protected])
Date: 04/25/01


On April 25, pat hayes writes:
> >Pat, I believe the current semantics are correct. The RDF Schema Spec
> >(sect. 3.1.4) says that rdfs:domain is "used to indicate the class(es)
> >on whose members a property can be used." To me, this indicates that an
> >instance of the class does not have to have a value for that property,
> >i.e., that there could be some Animals for which there is no hasParent
> >property, but any thing with a hasParent property must be an Animal.
> 
> Interesting. To me, that form of words suggests the oppposite 
> interpretation. (If a property has no value for some individual then 
> it cannot be "used" on that individual, surely? If it gets used, what 
> is the value of the property?)
> 
> The real moral is that vaguely worded specifications are worse than useless.

Agreed!

> OK, but I will take your advice. My problem now is that I don't feel 
> competent to rewrite the walkthru explanations, as I no longer feel 
> that I really follow the intended interpretation of properties and 
> domains. As far as I can see, with this interpretation, there is 
> never any point in declaring a domain or a range. Those statements 
> have no utility, since they do not allow a reasoner to draw any 
> conclusions.

I don't quite understand this. If Range(P,C) and P(x,y) then the
reasoner can infer C(y) - and similarly for Domain. As I said in my
other email on this subject, range and domain can simply be seen as
syntactic sugar for inclusion axioms. This kind of axiom can (as Jeff
also pointed out) be used to state either/both the universal and
existential forms of restriction. They offer just as much/little
potential for inference as any other kind of axiom, don't they?

Ian


This archive was generated by hypermail 2.1.4 : 04/02/02 EST