[Top] [All Lists]

Re: [ontac-forum] Follow up question on Ontology, knowledge, languagecon

To: ONTAC-WG General Discussion <ontac-forum@xxxxxxxxxxxxxx>
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Sat, 08 Oct 2005 20:39:07 -0700
Message-id: <434890DB.5030500@xxxxxxxxxxx>
Pat and Leo,    (01)

Thanks for the comments.  I don't think that we're too
far apart, but there are some important points to make.    (02)

First, I'd like to emphasize my commitment to having good
ontologies, terminologies, and lexicons.  We will definitely
need them to make progress on both practical and research
systems.  In my 1984 book, Conceptual Structures, I was one
of the first people working in AI to use the word "ontology"
and to promote its centrality for knowledge representation,
and I added more emphasis in my 2000 book on KR.    (03)

But in both of those two books, I also emphasized the problems
and limitations.  The last chapter of the CS book was titled
"Conceptual Relativity:  The Limits of Conceptualization", and
Chapter 6 of my KR book was titled "Knowledge Soup".    (04)

As for upper ontologies, I believe that they're useful, but I
don't believe that a single, fixed upper ontology is possible,
necessary, or even desirable.  Doug Lenat has said that he
believes the middle levels are more important than the upper
level -- that's another way of saying something similar, but
not quite.    (05)

In my KR book, I proposed the lattice of all possible theories
in which the only thing common to *every* ontology is the single
top theory that contains only the tautologies.  Every concept and
axiom below the top is optional in one possible ontology or another.    (06)

PJC> We need at least one large community using the same COSMO in
 > order to achieve the benefits of multiple researchers testing
 > their own reasoning methods within the same paradigm of meaning
 > representation.    (07)

I am very happy with the phrase "same paradigm".  That's a great
idea as a project for developing the lattice and determining the
range of choices.  But the important point is that the only thing
that is common is the lattice and the stock of options, no single
one of which is necessarily present in any particular  ontology.    (08)

PJC> ... unless the same COSMO is used, the results will be difficult
 > to interpret and methods may not be reusable.    (09)

If we interpret "same COSMO" to mean the same lattice of theories,
then I would completely agree.  It is important to evaluate which
concepts and axioms in the lattice are the most fruitful for use
and reuse in multiple applications.  Those that are useful can be
registered in the Metadata Registry and become available for
further mix and match combinations with other theories.    (010)

LJO> Eleven years ago, Jackendoff's Lexical-Conceptual Structure
 > framework (focused on lexical semantics) was shown to be expressible
 > in ordinary model-theoretic terms...    (011)

Of course, Jackendoff has a few good insights into some aspects
of semantics, but his understanding of logic is woefully inadequate,
and his comments about logic are hopelessly misguided.    (012)

And every theory in the lattice of theories that I have been
discussing is satisfiable by one or more models (except for the
inconsistent theory at the very bottom of the lattice).    (013)

LJO> Ad hominem bricks against Montague are not appreciated, even
 > if by Henry Kautz.    (014)

That poem said nothing negative about Montague, but it does say
a lot about people who think they've done something important
when they've translated some trivial ideas into logic.  The point
I wanted to emphasize is one that Lord Kelvin made a long time ago:    (015)

    Better an approximate answer to the right question
    than an exact answer to the wrong question.    (016)

If somebody has a half-vast theory, expressing it in logic does
nothing to improve it.  On the contrary, that may make it harder
for people to see the limitations.    (017)

LJO> Model-theory is not just about consistency. It's about entailment
 > relations, and a host of other notions...  I'm also interested in
 > category theory...  Information Flow Framework...    (018)

You're preaching to the choir.  Of course we need logic, model theory,
lattices...  And I love category theory, but I don't talk about it
very much because it tends to frighten people.    (019)

LJO> I'm afraid you sometimes tend to obliterate necessary and
 > under-appreciated distinctions in your rhetorical quest to make
 > particular points.    (020)

Please note my quotations from Whitehead in both the opening quote
of my knowledge soup paper and several other quotations in the paper
on signs, processes, and language games.  Whitehead was certainly a
strong promoter of logic, but he was very critical Russell's logical
atomism and Carnap's logical positivism, both of which were influenced
by the early Wittgenstein, who spent the second half of his life
preaching against the "grave mistakes" of his own and others
(including Frege, Russell, and Carnap).    (021)

I am firmly convinced that the major weaknesses of Cyc and many other
ontology projects is that they are based on Wittgenstein's early 
philosophy instead of his later corrections.    (022)

John    (023)

Message Archives: http://colab.cim3.net/forum/ontac-forum/
To Post: mailto:ontac-forum@xxxxxxxxxxxxxx
Shared Files: http://colab.cim3.net/file/work/SICoP/ontac/
Community Wiki: 
http://colab.cim3.net/cgi-bin/wiki.pl?SICoP/OntologyTaxonomyCoordinatingWG    (024)
<Prev in Thread] Current Thread [Next in Thread>