UOS-conveners:
I think that the comments of
Mills Davis (below, sent to Brand Niemann) provide a good example of the work
that the upper ontology community has to do in order to explain to the wider
world ? including most people in the IT field ? why upper ontologies are
important for achieving broad semantic interoperability among diverse
applications.
The need for formalized definitions in
ontologies has been preached by Doug Lenat and others since before 1990. The basic ideas go back 50 years to the
birth of "AI". One might have
supposed that by now everyone in IT would understand why computers need precise
definitions to do logical inferencing on
knowledge. But this
understanding is apparently still confined mostly to a small group who work with
axiomatized ontologies. Most of us
in this group understand the potential benefits of automated reasoning over
large knowledge bases, but some who are (properly) concerned with what can be
done immediately remain to be convinced, either of the ultimate potential, the
time required to get there, or the economic or social feasibility of the goal. It appears to me that those who are
convinced of the benefits of upper ontologies need more effort at outreach to a
broader audience. The UOS was
intended to help in that goal. The
result may depend on how much we can achieve on Tuesday.
Mills has been familiar with
applications of knowledge representations for years, and yet he still asks:
"why have one upper-level
ontology when you can have them all? "
The answer, of course, is that
having more than one upper-level ontology makes accurate and automated semantic
interoperability impossible (or in the more cautious words of the joint communiqué
draft):
" . . . use of some formally defined common
upper ontology by an organization or community is the most cost-effective method
for achieving semantic interoperability that can scale from a few applications
to diverse knowledge-based reasoning systems ".
. . . but Mills is apparently not
convinced. And I think that he
represents a very large fraction of the IT community in this respect.
I think that we cannot assume
that just because someone has been exposed to arguments about the importance of
upper ontologies, s/he will have understood the point. Even worse, we cannot assume that
someone who has access to, or been pointed to, discussions of the need for upper
ontologies has even read such discussions, let alone understood them. The issues are apparently complex enough
that they need to be phrased as simply and
succinctly as possible, and be accompanied by pointers to examples
illustrating how UOs can benefit users.
The former might be helped by the Joint Communique, which can be
redistributed wherever it may do some good. The examples are an issue we need to
discuss on Tuesday and Wednesday morning.
Another point that has come up in
the course of these discussions is that most of those who are building domain
ontologies or other knowledge classifications need something simple to work
with, if they are going to align such knowledge bases with a more general
concept classification. This also
impinges on the question of how much agreement can be achieved among the upper
ontologists, and whether a "common subset" can be found. These issues were raised by
comments of (among others) Barry Smith, David McComb, Mills Davis, John Sowa,
and Roy Roebuck.
But to serve the purposes of
semantic interoperability, I think that most of us would agree that even a
"simplified" upper ontology must have
sufficient formalization to avoid the common ambiguities associated with
linguistic terms. This recalls
Einstein's dictum: "A theory
should be made as simple as possible, but no simpler. "
My own appreciation of these
concerns leads me to feel that it would be very helpful to have an axiomatized UO fragment that is simple
enough to be rapidly and widely adopted so that it is highly visible, but
axiomatized to sufficient precision to avoid the major ambiguities in
interpreting the meanings of the terms, and can demonstrate to a wide audience
the advantages of a common, formally defined higher-level ontology to enable
information transfer among semantic modules. Perhaps a common subset will serve such
a purpose.
The note of Mills Davis is
attached directly below.
The current situation reminds me
of the "muckraker" allegory from "Pilgrims Progress" and I have attached that
reference below as well, with a short comment.
Pat
Patrick
Cassidy
MITRE Corporation
260 Industrial Way
Eatontown, NJ
07724
Mail Stop: MNJE
Phone: 732-578-6340
Cell: 908-565-4053
Fax:
732-578-6012
Email: pcassidy@xxxxxxxxx
=======================================================
Note of Mills Davis to Brand
Niemann
===========================================================
I talked with Steve, and yes,
I'll do 20 minutes with Doug Lenat to
set a business context for
ontology. Why? Because, at a minimum, it
is good that the
various ULO custodians are agreeing to work together
to cross link and
explore samenesses and differences between their
approaches. This is
the essence of the public statement they will
come up with at the end
of the second day. I support their desire to
work together because I
believe that the semantic wave is a big one
and that all contributions
are welcome and encouraged.
If you were to ask me where I think that
this ULO activity will come
out, then I have the following
thoughts:
(1) If the ULO quest is to arrive at one upper level ontology,
then I
think that the result will not be this. Rather, that there
will
emerge some few, a small set of primitives, that link to
low level
concepts or instances of interest via a direct binary
relationship;
then, between these layers of abstraction, any taxonomy
or upper
level ontology can be interposed, allowing us to choose which
is the
best for our purposes. Bottom line, why have one upper-level
ontology
when you can have them all? I suspect that the only
reason to be
looking for one upper level ontology would be if you
believed that
relational databases would be for ever and that
every item of data
must have one and only one tag.
(2)
Secondly, my interest is in the development of a "scientific
method"
for testing and evaluating the efficacy and validity of
metaphysical
patterns (ontology). In an era of knowledge computing,
its is of
critical importance to be able to validate and ground the
ontology
specifications in experimental data. Here, I think that the
focus
should shift from the nouns (concepts) to relationships and the
theories that underly these relationships. Nouns require conventions
and agreements. Relationships and the theories that underly them are
much more limited and scientific.
Bottom line: I applaud and
encourage the efforts of this UO
community. By coming together, I hope
they deliver value. At the same
time, I'm reserving judgement (or,
remaining agnostic regarding the
value, pending evidence), and am
harboring a supposition that in the
next year or so, technology may be
emerging that will obviate UO
arguments by subsuming all of these
disparate approaches and
subjecting them to tests of efficacy.
====================================
Pat Cassidy: fable for the
day.
===================================
There will always be a tension
between those who are concerned with what is concrete and doable immediately,
and those who are concerned about the things that can be done in the near future that represent a qualitative
improvement over the immediate. The
fable of the "muckraker" from "Pilgrim's Progress" puts that in an allegorical
(though religious) form:
From Paul Bunyan, Pilgrim's
Progress, stage 2
------------------------
This done, and after those things had been
somewhat digested by Christiana and her company, the Interpreter takes them
apart again, and has them first into a room where was a man that could look no
way but downwards, with a muck-rake in his hand. There stood also one over his
head with a celestial crown in his hand, and proffered him that crown for his
muck-rake; but the man did neither look up nor regard, but raked to himself the
straws, the small sticks, and dust of the floor.
Then said Christiana, I persuade myself that
I know somewhat the meaning of this; for this is a figure of a man of this
world: is it not, good sir?
INTER. Thou hast said right, said he; and
his muck-rake doth show his carnal mind. And whereas thou seest him rather give
heed to rake up straws and sticks, and the dust of the floor, than to do what He
says that calls to him from above with the celestial crown in his hand; it is to
show, that heaven is but as a fable to some, and that things here are counted
the only things substantial. Now, whereas it was also showed thee that the man
could look no way but downwards, it is to let thee know that earthly things,
when they are with power upon men's minds, quite carry their hearts away from
God.
-----------------------------
[PC] Apparently, a common upper
ontology that can enable semantic interoperability "is but as a fable to
some". I hope that the UOS
meeting can help to provide evidence that the benefits of upper ontologies are
more "substantial" than many in the wider world currently believe.