[topicmapmail] SUMO
Murray Altheim
m.altheim@open.ac.uk
Mon, 09 Jun 2003 01:48:06 +0100
Thomas B. Passin wrote:
> [Lars Marius Garshol]
>
>>* Murray Altheim
>>|
>>| [...] I think of Topic Maps as a *tool* for describing ontologies,
>>| not as the semantics of an ontology.
>>
>>That's what I think, too, and that's what I meant. Doesn't that sound
>>awfully like a "meta-ontology language"? I
>
> No, because Topic Maps have practically little semantic content out of the
> box. Topic Maps is (or can be seen as) a pattern for structuring
> information, one of many possible patterns. But even the class-subclass psi
> does not say much about what a "Class" is or how it behaves, which is needed
> to really build ontologies, I think.
Yes, I agree. Several months ago we had a minor debate on the choice of
"class" vs. "type" on the new set of PSIs being worked on by OASIS. That
devolved into a discussion about extensionality vs. intensionality in the
definition of classes. And that brought us to a question of what questions
we are actually trying to answer, what specific domain are we in? I hate
to say this publicly, but I'm quite happy to have "superclass-subclass"
and "class-instance" in XTM 1.0 regardless of definition.
> Owl is a meta-ontology language layered on top of RDF and RDFS. It __does__
> contain semantic definitions needed to apply it to defining ontologies.
>
> I read Murray's posts as asking in essence, what will play the role of OWL
> for Topic Maps, or at least for his thesis purposes? (Murray, I hope I am
> not drastically misrepresenting what you were getting at).
No, not at all. But what OWL is and what I'm looking for are a bit different.
I need both more and less. When you put 53 people in a room and expect to
create something pure and/or minimal, your head needs examining. So we can
at the outset expect to see that OWL is a compromise. There are certainly
overlaps in many of the features of OWL Lite and OWL DL that I would use in
my own ontologies, but these are features of pretty much any existing system
I know of. I don't have any particular fancy for description logics; to me
it's just one of many possible logics. For my application, I won't be able
to use DL for a variety of reasons, e.g., monotonicity. But I don't think my
application is necessarily typical, as I'm playing with issues of formality.
In the end, I'm not sure how much machine reasoning will be possible using
my ontologies, but I'm not sure how useful and valid OWL ontologies will be
in dealing with real world problems either. My real world problem only
requires one person to understand and use an ontology, so definitions of
terms like "class", "subclass", "extension" and "intension" are almost
irrelevant. My users are most likely to be like 99.995% of the planet, i.e.,
they have no idea what ontological engineering is, much less having any
expertise in it. Which is one of the reasons I expect OWL to die the same
grisly, painful and horrific death as every previous attempt at this kind
of thing: it will be used by the community that created it.
> Given some such semantic basis, then an upper ontology could be devised or
> adapted to be expressed in the new TM-OWL or TM-Cyc or TM-SUMO or whatever
> it ends up being.
Well, as you know, I actually do have some enthusiasm for upper ontologies.
I think they're kinda cool constructions, like extremely advanced Legos.
But I'm not going to fall prey to the same guileless enthusiasm some have
in believing that all those cruise missiles hit their intended targets,
that everyone arrested post 9/11 was guilty (none of the 762 were actually
charged, though some were tortured and many are still being held), or that
the world operates according to some nice, neat plan. Remember that one of
the primary financiers and clients of OWL is DARPA. Probably not far behind
is the Office of Information Awareness [1], which coupled with the new
Patriot II act [2] (which will allow US police to without cause swab your
cheek for DNA to put into that fancy new database), we all must remember
that the technologies WE are working on may likely be used against us. I
hope we don't all feel like Robert Oppenheimer in a decade or so.
So that I don't seem entirely off-the-wall with the above statements, I'm
going to repeat something I said earlier: while I can understand the desire
for the Universal Theory of the Universe, I firmly believe the whole idea
of implementing *universal* ontologies to be fundamentally flawed. So what
if 53 people can get together and agree on a definition of "class"? That
definition is not a consensus of the ontological engineering community (if
that were even possible), much less the common understanding of people
(whatever that might mean).
With permission from both parties, I'm including a message [3] sent within
the Conceptual Graph list recently. I agree with Stephen Downes that any
universal ontology would be authoritarian [4]. I don't share John's optimism
about point #1, expressed as the possibility that an upper ontology could
be "discovered" via empirical evidence, simply because I don't believe in
a universal empiricism, a universal experience.
Also, the world is enormously more complex and varied than is expressible
or discoverable. At best, we approximate. As a simple example, a typical
ontological category might be something like "animate" vs. "inanimate", or
"living" vs. "non-living". But as anyone following the current US supreme
court cases will attest, what exactly constitutes "life" is a very difficult
question to answer. And death? Well, we don't die suddenly. Our bodies die
over a rather long period of time. There are various legal and medical
definitions of death (e.g., brain death, cardiac arrest, etc.), but none
is universally "correct". Christians and Buddhists would say you don't even
die, though they'd disagree about what happens after. Another example I
think I've stated previously is that DNA evidence shows that the boundaries
between species are illusory, that the concept of differentiated species
just as flawed as that of "race". And even "cat on mat" is difficult: at a
molecular level, the boundary between the cat and the mat is permeable.
The cat and the mat are one! [okay, okay...]
Now, the point of a tool is to solve a problem. I'm happy in my little
world of building tools for specific purposes, and have no desire to try
to create universal tools. On some thread not discussing universals I'll
perhaps try to describe my own approach to computer-based ontologies, but
I've already typed more than enough for one email message.
Here's the two messages:
John F. Sowa wrote:
: Jean-Luc,
:
: That note you cited is very relevant to the current debates.
: I enclose a copy below for SUO readers who don't click on URLs.
:
: I like Downes's phrase "a regime of ontological authoritarianism",
: which is a very good characterization of what some people on SUO
: list are advocating. The question of whether a single SUO is
: possible is an empirical issue that can only be answered by
: analyzing a very wide range of data and determining whether
: a single SUO can emerge.
:
: For the record, following is my opinion on the subject, which
: has been influenced by the kinds of sources that Downes cites
: as well as many years of working on the problem and observing
: the results of other people working on similar goals:
:
: 1. I believe it is eventually possible to have an upper
: ontology that everyone can agree to.
:
: 2. For starters, the only universal point we can all agree to
: is a one-category ontology, which contains nothing but the
: empty node at the top. That is a start, but not a very
: useful one.
:
: 3. The next step is to determine a set of distinctions that
: can be used to classify entities. That is a time-honored
: principle since Socrates and Plato. I describe that
: process in Ch 2 of my KR book and summarize it on my
: web site: http://www.jfsowa.com/ontology/
:
: 4. The combination of all distinctions that are truly universal
: -- i.e. those that apply to any entities whatever -- form
: a lattice, as I summarize in the opening page of my web site
: and as the Formal Concept Analysis people have supported
: with a great deal of theory, practice, and software.
:
: 5. But many distinctions are conditional -- they do not apply
: to *all* entities, but only to those that presuppose some
: other combination of distinctions. Those can also be
: organized in the same lattice by using the FCA software.
:
: 6. However -- and this is the point of contention -- many
: distinctions (how many is an empirical issue) have optional
: and incompatible implications: they subdivide the ontology
: into subtheories or modules, which are individually consistent
: with the upper levels, but inconistent with one another.
:
: 7. One example of an incompatible distinction is the 3D vs. 4D
: representation of space-time, which everybody has been citing.
: The question of how many such incompatible distinctions there
: are is an empirical issue that can only be determined by
: analyzing the data.
:
: Summary: I believe it is possible to *discover*, not *legislate*,
: a single universal upper level. But finding it is an empirical
: issue, and any premature legislation that tries to impose one SUO
: on everyone would be a disaster. Furthermore, that single upper
: level, by itself, may be so small that it would be nearly useless
: without some significant number of incompatible distinctions that
: introduce mutually incompatible subtheories or modules. We already
: know some such incompatibilities, and we have not yet begun to
: catalog the remaining ones, which may be even more problematical
: than the 3D vs. 4D distinction.
:
: Bottom line: Finding an upper ontology is an empirical issue.
: Any authoritarian attempt to impose one would be as foolish
: as legislating the value pi=3 (which the Indiana legislature
: once tried in order to simplify life for school children).
:
: John Sowa
: ______________________________________________________________
:
: From: downess <Stephen.Downes at downes.ca>
: Date: Fri Jun 6, 2003 1:29 pm
: Subject: Re: SUO Ballot with 2 Questions
:
: Hiya,
:
: The question of whether there can be one ontology for any given field
: is one that I have already encountered in the field of learning
: object metadata.
:
: It is, indeed, a question that has been encountered especially in 20th
: century philosophy of language, so there are good grounds from which
: to arrive at an informed opinion.
:
: I deal with the question in this presentation:
: One Standard for All: Why We Don't Want It, Why We Don't Need It
: (It is a PowerPoint (sorry) but it reads like a paper)
:
: http://www.downes.ca/files/one_standard.ppt
:
: I think that the essential point that emerges in such discussions is
: that it is not merely a question of language. It is not possible to
: simply identify a set of terms and agree on them.
:
: Disagreements in language can be traced to disagreements in a wide
: variety of underlying hypothesis, among them including ontology
: (Quine), causation (Hanson), explanation (van Fraassen),
: categorization (Lakoff) and meaning (Wittgenstein). These five
: factors (among others) create a context of discourse, and it is the
: context of discourse (the 'pragmatics' (Morris)) that changes the
: meanings of the terms in question, changes the very language of
: discourse.
:
: Because there is no way around this (aside from a regime of
: ontological authoritarianism) any system of representation must adapt
: to this, learning, deploying, and teaching new vocabularies, new
: semantics, as the need arises.
:
: -- Stephen
:
: ----------------------------------------------------------------
: Stephen Downes ~ Researcher ~ National Research Council Canada
: Moncton ~ Canada ~ stephen@d... ~ http://www.downes.ca
: ----------------------------------------------------------------
Murray
[1] http://www.washingtonpost.com/ac2/wp-dyn/A40942-2002Nov11
http://www.thememoryhole.org/policestate/iao-logo.htm
[2] http://www.alternet.org/story.html?StoryID=15541
[3] http://mars.virtual-earth.de/pipermail/cg/2003q2/004880.html
[4] Stephen has asked that I include a link to another paper of
his at <http://www.downes.ca/files/one_standard.htm> though I might
note that this is an IE-tuned web page and didn't work for me on
Netscape/linux.
...........................................................................
Murray Altheim http://kmi.open.ac.uk/people/murray/
Knowledge Media Institute
The Open University, Milton Keynes, Bucks, MK7 6AA, UK .
"We can now hypothesise with some confidence that those apparently
happy, calm Buddhist souls one regularly comes across in places
such as Dharamsala, India, really are happy," said Professor Owen
Flanagan, of Duke University in North Carolina. -- BBC News
http://news.bbc.co.uk/1/hi/health/3047291.stm