Re: WOWG: first language proposal

On April 5, Peter F. Patel-Schneider writes:
> From: Jeff Heflin <heflin@cse.lehigh.edu>
> Subject: Re: WOWG: first language proposal
> Date: Wed, 27 Mar 2002 17:43:04 -0500
> 
> > Frank et al.,
> > 
> [...]
> > Now on to my
> > comments...
> > 
> > First, I'd like to suggest that we attempt to maintain the frame-nature
> > of the language for properties as well (kind of like facets). For
> > example, we could have the productions:
> > 
> > <definition> ::= Property ( <propertyId> ,<facet>* )
> > <facet> ::= <domain> | <range> | <supersprops> | Transitive |
> > SingleValued | UniquelyIdentifying 
> > <domain> ::= domain( <classId> )
> > <range> ::= range( <classId> | <dataTypeRange> )
> > <superprops> := supers( <classId>*)
> > etc.
> 
> Great idea.  Ian Horrocks had a similar idea that was prepared just a
> bit too late to make it into the initial proposal.
> 
> > I'd even like to see this idea carried over into the non-frame portion
> > of the language. What do you think?
> 
> I'm not sure what it would mean to carry this over into the non-frame
> portion, except, perhaps, to allow <description> in the frame-like
> constructs. 

Maybe what Jeff means is that if such a property definition can
capture (almost) everything that can be said about properties in the
non-frame portion, then there seems little point in supporting (many
of) the non-frame style property axioms.

> > Second, I think we should include some kind of syntax to indicate that a
> > specific set of definition make up an ontology. Perhaps include:
> > 
> > <ontology> ::= Ontology ( <definition>*)
> > 
> > This can also serve as a place holder for us to later attach ontology
> > metadata and versioning information.
> 
> Again a great idea.  Ian and I were just discussing, on the phone, what it
> would take to do this.   Something like a KB is a sequence of ontologies,
> and each ontology is a sequence of definitions, facts, ontologies, and
> ontology references.  Dublin core information could then be associated with
> the ontologies.

Yes, I think that some such mechanism could address many issues. E.g,
as an "ontology" could be as small as a single axiom, it would provide
a mechanism for arbitrarily fine grained documentation. It also
facilitates the design of tools where it is useful to maintain
groupings or orderings.

> > Third, what are you[r] thoughts on using AND, OR, and NOT instead of
> > intersectionOf, unionOf, and complementOf (as is done in OIL)? I think
> > these might be more intuitive, and would certainly be easier for people
> > to type.
> 
> I could go either way here.
> 
> > Finally, an important issue will be finding a way to map your abstract
> > syntax into XML/RDF and still preserve its simplicity. I believe that in
> > order to get a good, intuitive syntax, we'll have to seriously consider
> > dropping the idea of using triples to represent the language, i.e., do
> > not layer on top of RDF Schema (but this is a point I've already raised
> > in another thread).
> 
> Again, a great idea.  The abstract syntax sort of alludes to this in that
> the fact portion is written the way it is so that it can easily be mapped
> into RDF/XML but the non-fact portion is different from RDF/XML.

I would also vote for this solution. Using RDFS causes many very
serious problems, e.g., it is an awful syntax specification language
(look at the trouble with lists, verbosity, unconstrainedness etc.)
and is difficult/complex to use as a semantic foundation for a more
powerful language (look at the problems pointed out by Peter). 

I asked some time ago what are the great advantages of RDFS that
justify us paying such a high price - I am still waiting for a
convincing answer (and considering the price, the answer needs to be
pretty convincing).

Ian

> 
> > Jeff
> 
> Peter F. Patel-Schneider
> Bell Labs Research

Received on Friday, 5 April 2002 13:05:41 UTC