Re: WOWG: first language proposal (fwd)

------- start of forwarded message -------
From: Ian Horrocks <horrocks@cs.man.ac.uk>
To: "Jonathan Borden" <jonathan@openhealth.org>
Date: Fri, 5 Apr 2002 21:39:27 +0100
Subject: Re: WOWG: first language proposal
Reply-To: Ian Horrocks <horrocks@cs.man.ac.uk>

On April 5, Jonathan Borden writes:
> Ian Horrocks wrote:
> 
> 
> > On April 5, Jonathan Borden writes:
> ...
> > >
> > > but in something RDF compatible:
> > >
> > > <Class rdf:ID="Trio">
> > >         <subClassOf>
> > >                 <Restriction>
> > >                         <onProperty rdf:resource="#hasMember"/>
> > >                         <cardinality>3</cardinality>
> > >                         <rdf:range rdf:resource="#Musician"/>
> > >                 </Restriction>
> > >         </subClassOf>
> > > </Class>
> > >
> > > which, gosh, isn't all that different, is it?
> >
> > But some problems that now arise are:
> >
> > 1. Various parts of the syntactic structure are given unwanted logical
> > significance by RDF.
> 
> That is a different issue, i.e. one of semantics not syntax i.e. Pat Hayes'
> "dark triples" proposal would solve this, one the other hand unless we can
> organize ourselves to _ask for_ dark triples, nothing will happen. My
> frustration is that there appear to be too many slightly differing proposals
> which get in the way of eachother and deadlock the WG.

Dark triples might solve the problem, but seem like a blatant hack. I
still come back to the question, why bother? What does RDFS give us
that justifies all this hassle.

> >
> > 2. Unlike in the XML form, it is difficult/impossible to restrict the
> > syntax to just this form, so we have to be able to cope with things
> > like:
> >
> > <Class rdf:ID="Trio">
> >         <subClassOf>
> >                 <Restriction>
> >                         <onProperty rdf:resource="#hasMember"/>
> >                         <onProperty rdf:resource="#playsMusicBy"/>
> >                         <cardinality>3</cardinality>
> >                         <cardinality>5</cardinality>
> >                         <rdf:range rdf:resource="#Musician"/>
> >                         <rdf:range rdf:resource="#Composer"/>
> >                 </Restriction>
> >         </subClassOf>
> > </Class>
> 
> Certainly an OWL ontology (which describes OWL) could, err, restrict the
> "cardinality" attribute to a cardinality of 1, etc.

Even if you are happy to use (F)OWL assertions to restrict (F)OWL
syntax, with all the semantic problems that implies (it will probably
turn out that we need dark axioms), you don't get what you want - you
get the inference that 3=5, Musician=Composer etc. All we want to do
is specify/restrict the syntactic form of the language - which is what
XML is good at.

> > It would be much more straightforward to have the parser reject this
> > as invalid (F)OWL - which we easily can do if we adopt Jeff's
> > solution.
> >
> 
> What parser? So far we have:
> 
> 1) no grammar
> 2) no parser
> 3) no software that deals with the output of the parser

Converting the abstract form to XML is trivial. You then write a DTD
and use an XML parser - that will take all of 10 minutes (or
so). Existing DAML+OIL software can easily be adapted to use the
output of the XML parser.

> Is this progress? (That said, I really am all for something that is an
> actual improvement).

No problems with semantics. A clearly and precisely specified
syntax. Increased readability (the above example avoids the dreaded
list issue). Yes, I think we can say that it is progress.

Ian


> 
> Jonathan
> 
> 
> 
------- end of forwarded message -------

Received on Friday, 5 April 2002 17:16:37 UTC