Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: SUO: RE: More KIF-ified Ontology Content



At 06-02-01 14:02 -0400, John F. Sowa wrote:

Pat,

I'm happy with those suggestions:

>>Pat maintains that "a Tarski-style model" is or should be...
>
>No; COULD be.  I hope that isnt a provocation to war. My claim is
>that the theory is entirely agnostic as to the nature of the things
>in the sets.

>With the understanding that 'concrete models' includes 'abstract
>models' as a special case, this would be fine. Apart from that
>quibble, I entirely concur.

To summarize the agreement,

 1. The term "model" won't make any commitment about the nature
    of the base entities:  they could be abstract things like
    surrogates or pointers, or they could be actual physical
    entities (whatever one might think the physical world
    happens to be made up of).

 2. The word "table" can be used as a synonym for "extension
    of a relation" as long as we agree that tables could be
    generalized to infinitely many (even uncountably infinite)
    tuples of entities (which may themselves be pointers,
    surrogates, names, or even "physical things", if need be).

 3. If a model is entirely made up of abstract, mathematical
    entities, it could be called an "abstract model" to make
    it clear that there are no physical things in it.

By the way, many people in database land make a distinction
between "lexical object types" (LOTs) and "nonlexical object
types" (NOLOTs).  LOTs are entities like character strings,
numbers, and pointers, and NOLOTs are things like cats, dogs,
trees, people, houses, etc.  This illustrates the typical IT
perspective that makes physical things secondary to the kinds
of things that are represented in computers.

### having been THERE when the terms "LOT" and "NOLOT" actually
were coined back in 1978, when I was but a lowly member of Sjir
Nijssen's staff at Control Data's Data Management Research Lab
in Brussels where we were thinking up the NIAM methodology for
database design, it is my sad responsibility :-) to correct John
on this grading of the respective importance of things and
references. Things (NOLOT instances) were definitely the more
"fundamental" items, at least in NIAM! I proposed the acronym
"NOLOT" only because it was (a) catchily pronounceable and (b)
an artifice, as such avoiding the confusion necessarily
surrounding any existing noun, and also forcibly turning the
attention of poor NIAM methodology disciples to the distinctness
from "mere" LOTs. So these are just signs and as we all know
signs lack intrinsic meaning --I'm ducking, Jon!  :-)

### This said, it may serve to draw our collective attention to
methodological issues within ontology engineering. From my
experience with database- and other system modeling activities
these issues can rarely be separated from representational ones,
even at the axiomatic level (when agreeing on good primitives).
After all, we probably would like the building of ontologies to
become a teachable discipline... in the case of fundamentally
semantic tools such as ontologies, this implies a "scalability
of the conceptual grasp", so to speak. Our "understanding" of
a modern CPU chip, for all practical purposes, is entirely based
on our willingness to trust its user-level specification, not
on a formal correctness proof from the quantum physics of its
transistors and some Zermelo-Fraenkel stuff. (Yes I know, Intel
bashers will at this point say,... blah etc. blah) --I hope the
point is somewhat clear.


I have no objection to making such distinctions, as long as
the basic assumptions are stated clearly up front.

John


Prof Dr Robert A Meersman       VUB (Vrije Universiteit Brussel)
Department of Computer Science  STARlab   --Building F-G/10
Pleinlaan 2                     B-1050 Brussels Belgium
phn (+32)(0)2 629 3308          fax (+32)(0)2 629 3525