Re: Unified Framework for Ontology
Rob,
We're not too far apart, but I'll add some qualifications
to your points. And at the end, I include a note that I
sent to the ONTAC WG on a related topic.
RF> The idea of abandoning a single central standard set
> of axioms sounds entirely sensible to me. But why stop
> at axioms?
I don't say we should stop there, but my proposal for UF
is intended as a useful intermediate step.
RF> Do you need WordNet? Words are perfectly useful categories,
> why not strike your balance between the general and the
> specific the same place language does, at words.
There are good reasons for developing formalisms, and even
though I often criticize them, I also recognize their value
for many applications. Something like WordNet is a useful
intermediate step.
RF> That is to say, if you are going to allow your users to
> make theory specific mappings, why not allow them to make
> theory specific mappings between words and categories too?
I would certainly want to do that.
RF> Surely if abandoning fixed theories gives you greater
> universality, abandoning fixed categories will give you
> more again.
I am not proposing that *any* category be fixed for all time.
On the contrary, my recommendation is to make the ontology
modular so that you can replace theories as units -- where
any theory could be as large or as small as you like.
Following is my previous note.
John
-------- Original Message --------
Subject: The possibility of a universal framework
Date: Sun, 04 Dec 2005 10:53:51 -0500
From: John F. Sowa <sowa@bestweb.net>
To: ONTAC-WG General Discussion <ontac-forum@colab.cim3.net>
I received an offline note related to some of this
discussion. Following is my response with the
features that identify the sender deleted.
John Sowa
________________________________________________________
I am firmly convinced that it is in principle impossible
to have a single, ideal, universal framework of any kind
that is also formal and precise. Such a system would
be far too rigid and inflexible to serve as a foundation
for knowledge representation for science, engineering,
business, and everyday life.
This does not mean that we cannot have formal ontologies,
but merely that we cannot have a single, monolithic ontology
that is fixed and precisely formalized from top to bottom.
We might have multiple ontologies or a framework that allows
modules to be replaced or modified, but the idea of a single
monolith is doomed.
On the other hand, I also believe that *every* natural
language is capable of being extended by means of metaphors
and related techniques to be a universal *informal* system.
Note the distinction:
1. No precise formal system can ever be universal.
2. But every natural language embodies sufficient
resources to serve as a universal informal system
that can be extended and modified by metaphorical
means to cover everything that is humanly conceivable.
I strongly disagree with the following:
> Although her doctoral supervisors ... applaud her
> exploration of "the nature of human language by using
> the experimental scientific method," they do not see
> her research as something useful for exact sciences
> and technology.
On the contrary, the fundamental methods of doing creative
research in science and technology *always* involve breaking
out of any rigid formalized system. Every creative advance
introduces new meanings that are totally foreign to the
framework that had been precisely defined beforehand. You
can have trivial innovations that reshuffle the old ideas,
but a creative advance must, of necessity, break the mold.
Just look at Einstein's papers of 1905 -- which destroyed
the foundations of classical physics. If Einstein had limited
his thinking to just those concepts that had been formalized
up to 1904, it would have been impossible for him to think
those new thoughts or to express them in a formal language
that had previously been defined.
Every creative advance in science *and* engineering destroys
some previous formalism and introduces a totally new framework.
That means that if science were restricted to only those
concepts that had been formalized at some point in time, *all*
of science and technology would be frozen at that instant,
and no further advance would be *conceivable*.
Note the word "conceivable". If you freeze the language,
you freeze thought. George Orwell understood that principle
very well -- read his book _1984_, especially his discussion
of Newspeak as the intended replacement for English.
These points don't imply that formalism is bad. It's necessary
for any kind of computer programming. But as soon as you have
formalized something for version 1.0 of your program, you can
be certain that at least part of it will be contradicted by
something in version 1.1. And by the time you get to v. 2.0,
you will have to rewrite the whole theory that you formalized
for v. 1.0.
John Sowa