SEM/TEST: restricted complement, unrestricted InverseFunctionalProperty

In yesterday's telcon, I mentioned that I brought
some of the layering (now embedding?) issues to
the attention of TimBL and some of the
Semantic Web Advanced Development folks, and
he was prepared to accept that owl:Class is
smaller than rdfs:Class; Peter said he was surprised
but happy. So perhaps this bears elaboration...

Basically, TimBL and I don't expect that
RDFS classes are closed under complement, but
we can see that it would be useful to define
some notion of class that _is_ closed under
complement.

[[ some potential test cases to make this concrete:

  :bob a :Person.
  -- does not ential --
  _:x owl:complementOf :Person.

but

  :Person a owl:Class.
  -- entials --
  _:x owl:complementOf :Person.

]]

  
Meanwhile, please don't standardize
a nifty notion like inverse functional without
letting us use it for RDFS properties
general.

[[test case: sameState.

http://www.w3.org/2002/03owlt/sameStateP.rdf
http://www.w3.org/2002/03owlt/sameStateC.rdf
]]


A bit more elaboration: I don't expect
us to specify that RDFS classes are closed
under union and intersection (but maybe
Pat has that worked out... I dunno...)
but please let's allow the notions
of union and intersection to work, somewhat,
for RDFS classes too.

That is:

	:bob a [ owl:intersectionOf (:Person :Student)].
	-- entials --
	:bob a :Person.

even though we don't know that :Person is an owl:Class.
But we do might need that extra premise to
conclude that the intersection exists:

	:Person a owl:Class.
	:Student a owl:Class.
	:bob a :Person.
	:bob a :Student.
	-- entials --
	:bob a [ owl:intersectionOf (:Person :Student) ].

So I hope we'll specify the semantics of
things like InverseFunctionalProperty, disjointFrom,
unionOf, intersectionOf over the whole RDF universe,
though somewhat weakly. Then we'll give stronger
conclusions over the restricted/stratified DL
style universe.

One problem that occured to me during the telcon yesterday
is that it might be a pretty hard problem to figure
out that an ontology meets all the DL constraints.
So it'll probably be useful to have an explicit
property on ontologies to say "this ontology
meets all the DL contraints." Maybe a subclass
of owl:Ontology called owl:FastOntology or whatever...
it promises that all the DL constraints are met
and enables certain inference techniques.

Hmm... I'm starting to see an actual design along
these lines... maybe I can update my model theory/axioms
along these lines... stay tuned...

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Friday, 6 September 2002 10:26:33 UTC