W3C home > Mailing lists > Public > www-webont-wg@w3.org > July 2002

Re: LANG: closing issue 4.6 (was Re: ADMIN: Draf agenda for July 25 telecon)

From: Dan Connolly <connolly@w3.org>
Date: 24 Jul 2002 16:53:30 -0500
To: Jonathan Borden <jonathan@openhealth.org>
Cc: "Peter F. "Patel-Schneider <pfps@research.bell-labs.com>, www-webont-wg@w3.org
Message-Id: <1027547611.11038.286.camel@dirk>

On Wed, 2002-07-24 at 12:47, Jonathan Borden wrote:
> Dan Connolly wrote:
> 
> > > I vote for option 2.
> >
> > That's my preference for how sameClassAs works too.
> 
> I can't vote on this because I have no idea what the implications of these
> two options are. That is to ask: what is the real difference between
> "strong" vs. "weak" class as instances?

I don't understand that question...

> > A couple test cases: first, an obvious one:
> >
> >
> > :Car owl:sameClassAs :Automobile.
> > :car1 rdf:type :Car.
> > ==>
> > :car1 rdf:type :Automobile.
> >
> > now, one that shows the distinction:
> >
> > :Car owl:sameClassAs :Automobile.
> > :Car :averagePrice "20000".
> > =?=>
> > :Automobile :averagePrice "20000".
> 
> Help! My brain doesn't know how to react to this!
> 
> Is this the same thing as saying that an instance of the class :Car has a
> property :averagePrice which is constrained to have the value "20000"?

No; the class itself has that property.

> If
> not, why would a _class_ have an averagePrice


Consider some system that does natural language
understanding of news stories or something,
coming across...

	He left his key in the car.
	It's going to cost him $10 to get it replaced.

From knowledge of english grammar and
parts of speech, you can get as far as

	(left-in He ?k ?c)
	(type ?k key)
	(type ?c car)
	(replacement-cost ?it $10)

But the machine can't tell whether ?it
refers to his key or the car. A typical
approach is to ambiguity at the syntax
layer is to return all possible matches,
and disambiguate at the semantic level.

But if it knows cars cost around $20000, then it can demote
the possibility that 'it' refers to a car.


> -- are we going to sell them.
> If so I'd like to have as many different classes as possible :-))
> 
> If it does mean that instances of this class do have this property, then
> what would the meaning of sameClassAs be? Doesn't that mean that both
> classes have the same property constraints?
> ...
> 
> >
> > :Car owl:equivalentTo :Automobile.
> > :car1 rdf:type :Car.
> > :Car :averagePrice "20000".
> > :Automobile :averageWeight "2000".
> > =?=>
> > :car1 rdf:type _:someClass.
> > _:someClass :averagePrice "20000".
> > _:someClass :averageWeight "2000".
> >
> > I'm still thinking about whether I really, really need this in
> > owl or not.
> >
> 
> This one doesn't look terribly solipsistic if we are still concerned with
> such things :-) What are the pros and cons of each approach?

"each approach"? You mean the approach where the test passes
and the one where it fails?

Well, an approach where the above passes just involves
the RDF Core model theory plus a few horn-clause rules:

{ :x = :y } log:implies { :y = :x }. # symmetric
{ :x = [ = :z] } log:implies { :x = :z }. # transitive
{ :x :p :y. :x = :z. } log:implies { :z :p :y }.
{ :x :p :y. :p = :q. } log:implies { :x :q :y }.
{ :x :p :y. :y = :z. } log:implies { :x :p :z }.

(taken from http://www.w3.org/2000/10/swap/test/sameThing.n3,
lightly edited)


My experience with this approach is pretty positive.
Our implementation is kinda slow, but it's tolerable.


I don't have any experience with an approach where
that test doesn't work.


[hmmm... Jos, I see a rule for reflexivity of
equivalentTo in
  http://www.agfa.com/w3c/euler/owl-rules
but not one for transitivity nor symmetry,
let alone substitution-of-equals-for-equals.].


> If classes have
> weights, then solipsism is looking better.... my laptop might suddenly get
> very heavy :-))
> 
> Jonathan
-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
see you in Montreal in August at Extreme Markup 2002?
Received on Wednesday, 24 July 2002 17:53:21 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:57:51 GMT