Re: SUO: RE: Topic Maps
I would like to emphasize and expand upon some points
by Pat Hayes, which I heartily endorse:
>>Topic Maps is slated as the infrastructure for the way that semantics on the
>>web are going to be managed.
>
>Really? By whom? W3C seems to have committed to RDF
>http://www.w3.org/2000/07/hs78/rdf-ms as the basis for its
>'semantic web' effort (http://www.w3.org/2000/01/sw/), and Darpa is
>funding DAML+OIL http://dtsn.darpa.mil/iso/programtemp.asp?mode=347
>, http://www.cs.man.ac.uk/~horrocks/DAML-OIL/ which is now close to
>an international standard, being compatible both with the
>EEC-mandated OIL http://www.ontoknowledge.org/oil/ and RDF. Several
>of us are involved with the DAML+OIL initiative, so I think that on
>the whole we are fairly well on, if not ahead, of this curve.
I would like to point out that all of these notations, Topic
Maps, RDF, DAML, OIL, SQL, Express, Epistle, and even English
and other natural languages, are built upon exactly the same
primitives as good old fashioned first-order logic. Whatever
strength and generality any of these notations may have, it is
derived almost entirely from their ability to express some or
all of the primitives of first-order logic. Whatever weaknesses
or limitations these notations may have is almost always the
result of not being able to combine all the primitives of FOL
in the simplest and most flexible ways.
That is why we have adopted FOL as the foundation for KIF and
conceptual graphs. There are many more issues to be discussed
about logic and language, but support for FOL is the minimum
prerequisite for beginning to address the issues. For starters,
I recommend my ontometa paper, which presents the five basic
primitives and their expression in English, RDF, predicate
calculus (or KIF), existential graphs, and conceptual graphs:
http://www.bestweb.net/~sowa/peirce/ontometa.htm
See Sections 2 and 3 of that paper for a summary of what kinds
of features *MUST* be present in any adequate notation. (There
may be other desirable features as well, but any notation that
cannot express these is, as they say, a "nonstarter".)
>>If we wait before getting involved with Topic Maps, then some other group
>>will grab the opportunity, and we risk being consigned to the dustbin of
>>irrelevance (the Betamax of ontologies).
>>
>>Let me put it the other way. If we don't want to be an ontology for the web,
>>what do we want to be an ontology for?
>'an ontology for the web' is a tricky phrase. We certainly want to be
>an ontology ON the web, and hopefully a valuable web resource. But
>the 'ontology FOR the web' is often taken to mean an ontology for use
>by web agent software, and I think that most of the SUO isnt in fact
>directed chiefly to that goal.
The basic point is that if you can support FOL, you've got
a foundation for everything. If you can't support FOL,
you've got a limited notation with built-in limitations.
KIF, CGs, and other logic-based systems can represent each
other (in the sense that anything represented in either of them
can be mapped into the other without any semantic loss), and
they can also represent Topic Maps, RDF, and whatever else may
be thrown at them.
If anyone has any other pet notation to propose, their first
step to prove minimal adequacy is to demonstrate that they can
support KIF and CGs.
John Sowa