Re: rdfms-assertion and webont

From: Jim Hendler <hendler@cs.umd.edu>
Subject: Re: rdfms-assertion and webont
Date: Fri, 24 Jan 2003 09:39:34 -0500

> jeremy -
>   the following is currently only my opinion, but I could see it being 
> a chair's ruling at some point:
> 
> I see the issue of social meaning to be extremely important and 
> powerful with respect to the success of the Semantic Web, the meaning 
> of RDF, etc. - but I also think it is NOT appropriate AS AN ISSUE for 
> WOWG -- basically, we as a WG do not have to have a consensus opinion 
> on this -- here's why:
> 
> 1 - rdfs:comment is in Owl Full, whatever happens.  Our Model Theory 
> does not CHANGE its meaning - so if in the Recommendation process, 
> the social meaning issue is left the same, then we accept that; if it 
> is changed, then we accept that.  (As individuals we should all try 
> to get this issue to end up where we want)

Well, yes, at least I guess so.  That is, if the RDF recommendation keeps
the social meaning stuff, then OWL will also get it, unless something is
done to block the application of RDF social meaning to OWL documents.  I
don't view this as an impossibility, as RDF social meaning is more recent
than the WebOnt charter, and thus I don't view it as a necessary part of
any OWL.

> 2 - we do have an issue that is a legitimate WOWG issue which is 
> whether the annotations are in OWL DL (and/or Lite).  This could be 
> influenced by the social meaning issue

Agreed.

> 3- we never in any way FORCE anyone to use refs:comment, label etc. 
> As such, a user of OWL, who doesn't believe in the social meaning 
> stuff, is never forced to use it -- they can avoid importing 
> documents with comments, they can avoid using them, etc.

I don't see this as a viable option.

> So, searching our documents, I don't see anywhere where we would 
> change a document or a test case BASED ON THE SOCIAL MEANING (we well 
> might change one based on whether this is in Lite or Full, but that's 
> a different issue as I emphasize)

I'm not so sure about this.  It may be that no test case changes, but it is
certainly the case that RDF social meaning dramatically changes the meaning
of RDF (and thus potentially OWL) documents.

> so - the example you give below would be an Owl issue if you are 
> willing to put in a test case which says:
> 
> >If A defines an innocent vocabulary,
> >If B describes a person "John Doe" in the vocab of A
> >If C defines a class using vocab of A and making defammatory assertions
> >about its members,
> >and
> >D imports  A, B and C, and "John Doe" fits the description in C.
> 
> DL-entails
> 
> owl:insulted "John Doe"
> 
> which of course is not true -- the RDF meaning insists that John can 
> consider himself insulted (i.e. he would be socially-entailed as you 
> call it below)
> but that does not come from OWL

If the OWL specification doesn't prevent the spillover of RDF social
meaning to OWL documents, then this RDF social meaning becomes *part* of
the OWL meaning of OWL document.

> write me an example where the presence of a comment in a DL document 
> changes something in the formal OWL entailment and I would consider 
> this an OWL issue.

It is probably the case that no formal OWL entailment should change because
of RDF social meaning.  This does not mean, however, that RDF social
meaning does not impact OWL meaning.

> So - In the very hypothetical case that  Guus and I did chair's 
> ruling that said "the issue of whether annotations are in DL and Lite 
> is a WOWG issue, and should be resolved; but the issue of the social 
> meaning of these annotations is out of our charter and in the charter 
> of RDF Core."
> 
> Why would this be something that would be appealed/objected to?

I would object most strenuously, as I strongly feel that RDF social meaning
would thus infect OWL and make OWL unusable.

> (note for real - I'm using this to pose the question in an extreme, 
> I'm really trying to probe the edges - I think if we enter this 
> realm, and especially in the very nasty case where we went to a vote, 
> and over some objections (including yours and mine) the OWL group 
> ended up with a result that went counter to RDF Core, we would create 
> a mess that we really don't need to from a technical specification 
> perspective??

> To put it in Dan's terms, show me a test case that exposes the issue 
> of social meaning -- NOT the issue of whether rdfs:comment is in Owl 
> Lite.

We currently don't have a place for social meaning in our test cases.
However, here is a potential ``social'' test case

Document A:

	...
	Class(X complete Person annotation(rdfs:comment "All X's are liars")
		restriction(nationality value(Cretan)))

	...

Document B:
	 
	...
	Imports(A)
	Individual(John type(Person) value(nationality Cretan)
		value(name "John H. Smith-Jones, III"))
	...


Does document B OWL-insult John H. Smith-Jones, III?

Note that it might not RDFS-insult John H. Smith-Jones, III, although there
is a reading of RDF social meaning where this insult is part of the RDF
meaning of document B.


peter

Received on Wednesday, 29 January 2003 16:24:44 UTC