Re: SUO: Re: Applications & ontology [was: Negotiation Instead ofLegislation]
On 2/26/02 10:26, "Douglas McDavid" <mcdavid@us.ibm.com> wrote:
> Bill --
>
> I wouldn't have asked the same question about the kinds of meta-ontological
> concepts that you list (and that are described in Chris and Nicola's recent
> article in Communications of the ACM:
> http://www.acm.org/cacm/0202/0202toc.html ).
Right - they've been at this a while now. However, these concepts are not
necessarily meta-ontological. They are ontological (and thus not "concepts"
but ontological categories) if one is a realist about universals (properties
and relations).
> What I am asking about are the concepts in SUMO (like the whole area around
> Organization and OrganizationalUnit), which are non-meta, and possibly purport
> to be creating structure for semantic interoperability among applications.
> What I am trying to push for is an understanding of a kind of semantic
> architecture that has roles for higher-level abstractions (upper
> ontolog(ies)), more domain-specific ontologies, and applications. What do we
> expect an application to do about a concept like WrittenLinguisticExpression
> or OrganizationalUnit? Or do we ever expect an application to see concepts
> from the upper level?
I think there are two kinds of upper level, really. First, you have what
you called the "meta-ontological level" above. Whether or not it talks
about anything real (i.e. you buy into the reality of universals) is a
matter for philosophers to argue over. However, as you point out, it is
useful for organizing what becomes the top-level ontology of particulars.
This "meta-ontological" level should be neither seen nor felt by normal
application writers and users. It should only be felt by ontology authors.
This is precisely how Guarino and Welty break their work up, I think, to
avoid a debate over the reality of universals. This top-level ontology of
particulars is going to be organized along different dimensions (perhaps
dependence, identity, unity, etc) and provides very high-level constraints
that are (or should be) never "felt" by users.
So that leaves the kinds of categories mentioned in SUMO. That I will leave
to Ian and others to address, since I don't know much about them.
> These are different questions, in my mind, from the meta-level issues that you
> point out below. Do you agree?
Yeah, I agree..
I admit to having not seen enough data points, so I wasn't sure what kinds
of things you were having trouble with and I took a guess. I thought your
complaint was a more general one than a concrete criticism of the SUMO
contents. Sorry about that.
BTW, this impinges somewhat on the "Legislation vs Negotiation" debate. I
believe it is perfectly fair to "legislate" the meta-ontological level and
the top-level ontology of particulars (however that might be delimited,
either by fiat or by some philosophical/logical principles or both). I
think not only that agreement can be reached on such issues, but that
agreement in this case is a good thing. Below those levels the situation is
less clear, and the benefits to a "legislated" ontology even murkier, as
your, John Sowa's, and others' comments point out.
One final point - With respect to the story I mentioned (which I will post
as soon as I formulate it clearly)... I do find the lack of concrete
examples where middle-level categories (e.g. those in SUMO) would be useful
for real applications to be a bit disturbing. In a conversation with Mike
Uschold a while back, he expressed the same lament to me. It would be
useful for the ontology community to compile a set of such real-life "war
stories" that can be readily summoned to validate what we're doing. If we
don't then it's hard to see why we should be taken seriously, especially in
the face of all the buzz generated by the "ontology-lite" interests that are
feverishly working to develop web "ontology" standards and the like.
Cheers,
.bill