[topicmapmail] Getty Art and Architecture Thesaurus On Line

[Thomas B. Passin]

[Murray Altheim]
> Thomas B. Passin wrote:
> > [Murray Altheim]
> >
> > The terminology used by rdfs and OWL is sometimes very useful.  I'm
thinking
> > here of domain/range.  Say the relationship between "species" and "Homo
> > sapiens" is class-subclass.  Then you could say
> >
> > subclassOf domain Class
> > subclassOf range Class
> > "Homo sapiens" subclassOf species
> >
> > No ambiguity here. you could also invent a relationship instanceClass,
say,
> > to indicate that you want to consider a class to be an instance of
another
> > class, and you would give that relationship the same domain and range as
> > subclassOf.  In fact, it is not clear to me why the two would be
different
> > relationships.
>
> Except that to my understanding of RDF/OWL, what you've written, is that
> the statements are incorrect. "domain" is not a subclass of "Class", nor
> is "range" is not a subclass of "Class".

In OWL, "subclassOf" is an (RDF-style)  property.  It may be applied to
resources of type "Class" (its domain) and it may have values of type
"Class" (that is its range).

> Neither is "Homo sapiens" a
> subclass of a species, rather an instance of a species. Perhaps I'm not
> understanding what you're saying correctly, as this seems fairly straight-
> forwardly wrong. Could you clarify?
>

I assumed that when you said "species", you meant a categorizing term, and
so I am calling "species" a Class.  If this is correct, then "Home sapiens"
is precisely a subclassOf "species", isn't it?  If not, we are getting into
terminology, and I really meant to show how it is possible to distinguish
between different uses of what you are calling "instance", by means of
specifying the allowed domains and ranges.  IOW, I think you are using the
word "instance" to mean different things in the two cases, and I think that
some other term should be used to distinguish the two.

What is this distinction that I am claiming?  In the one case you expect the
target of "instance" to be a class (or species, or whatever collective term
you want to use), in the other you expect it to be an individual thing, like
an antelope or rose - that is, an individual as opposed to a collective
concept.

If topic maps had a defined vocabulary that included something like domain
and range, you could precisely make clear what you have in mind here.  And
if you truly want to include both senses of the term "instance", with OWL
you could do it by specifying (possibly anonymous) union classes to get both
ranges combined.

> >
> > My point here is that I think of these two usages of "instance" as
different
> > relationships, since they have different ranges.  [...]
>
> Actually, better definitions can be found in either Cyc, SUO, or the
> recently-announced ISO 15926 definitions. See
>
> Now, there are many ways to skin a cat. I'm not particularly fond of this
> scheme, but it apparently works for the audience of ISO 15926.
>

I see why you say that ... I just looked at the entry for "Class" and I
think it is pretty weak.
>
> The problem is not that there isn't an available definition of a class-
> instance relationship, it's that there are a number of competing ones,
> such as classes built upon their membership (extensive) or by their
> characteristics (intensive).
>

As I said, I'm not looking to get into terminology (and definitions of
"class" and "instance") here - well, I guess it can't be entirely avoided,
can it?  My claim, I think, is that if we could define the domain and range
(or some equivalents), it would be much easier to say whether in fact your
"instances" do or do not mean the same thing.

> > Maybe we should define what a class is, for the purposes of being more
> > precise about themeaning of superclass-subclass and instanceOf.
>
> I think a more appropriate thing than *us* doing it, is to simply use
> PSIs pointing to the various existing definitions as developed within
> each of the various competing schools of thought on the subject.

Yes, I would think so.

> ... As I said, I tend to agree with Peirce and think that
> knowledge representation only has validity as a form of communication
> of non-Platonic concepts within a specific, given interpretation. You
> need thirdness, context.
>

I really only know about Pierce's writings from reading Sowa so far, and I
am not that clear about thirdness.  I don't automatically equate thirdness
with context, but I can see that context might act in that capacity.  I need
to think about it some more.  Any suggestions for non-Sowa writings about
Pierce that help to penetrate the terminology (with an eventual aim of being
able to grasp the man's writings themselves)?

> I don't believe you can create static, universal class definitions. John
> Sowa has stated [1][2] that he thinks that through careful analysis the
> human community could eventually "discover" these Platonic ideals.

When I read Sowa's post, I did not take it that way.  He talks about
arriving at some satisfactory upper ontology by empirical means, and I think
of that as being quite distinct from discovering Platonic ideals.

Cheers,

Tom P