Re: SUO: Re: Proposed SUO Content Outline
"Douglas McDavid" <mcdavid@us.ibm.com>
>
>You say:
>
> > Ah, OK. What you are talking about here is what one might call a
> > meta-ontology: a theory of ontologies, with information about their
> > strengths and weaknesses, what they are good for, how they relate to
> > one another, etc.. That is a noble ambition, but I don't think it is
> > what the SUO is trying to be. The SUO proposes to be a *single*
> > upper-level ontological framework which reconciles all these various
> > points of view into *one* classification heirarchy into which all
> > other ontologies can be fitted. Which is one reason why many of us
> > are very doubtful that it can be done.
>
>Let's see -- You dismiss your understanding of my suggestion
>as not what the SUO is about. At the same time, what you think it is
>about,
To be precise, what Adam and Ian think it is about, and indeed what
it has been about since it has been first mooted (by Bob Spillers)
about 5 years ago.
> you don't think is possible to do. Which begs the question of why
>you spend time on work that has no hope of success?
Well, to some extent that is my own business, but I will come clean
and respond.
There are several possible outcomes of the SUO effort.
1. A single, usable, well-designed, universal upper ontology is
created within which all other (or most other, or enough other to be
highly useful) ontologies can be fitted, thereby enabling
inter-translation, conformity, productivity and putting an end to
petty confusions.
2. A 'standard' upper ontology is created which is a shambles, being
a mishmash of ill-digested ideas, incompatible hidden assumptions and
so on, and is much more harm than utility; but it is given an IEEE
primateur, and becomes thereby widely adopted as a 'standard', and
acquires the practical force of law (like Microsoft Word, use of
which is now enforced by many federal government decrees.)
3. The whole enterprise is eventually abandoned, but everyone
involved has learned quite a lot about ontologies, one way or
another. (There are plenty of examples of this, eg the Japanese '5th
generation' initiative, and the US 'supercomputer center' plan, not
to mention the space race.)
Adam and Ian have 1. in mind, which I think is impossible. I believe
3. is the most likely outcome; but above all, I want to stop 2.
happening. My major fear is that many of the SUO enthusiasts don't
know the difference between 1. and 2.
>In response to my statement that:
>
> >I believe the world is ready for a huge new emphasis
> >on the *meaning* within software-based information systems. We're only
> >recently in this position, with the critical mass of computing power and
> >interoperation among enterprises of all types now in place.
>
>you say:
>
> > Sorry if Im one of the old cynical guys, but Ive been doing this
> > stuff for about 30 years now, as have many other people in AI, and it
> > doesnt seem that anything dramatic has changed. Moores law has been
> > going on for a while now, but that kind of advance doesnt make the
> > representational issues any easier. Ian is still talking about the
> > situation calculus (reference McCarthy 1961, or thereabouts), and
> > others are still getting hyped about modal logics (references from
> > mid-60s), so where are the new ideas coming from?
>
>This reveals an assumption that breakthrough new representation paradigms
>or
>mechanisms are needed. My perspective is that we have plenty of
>mechanisms,
Maybe you could tell us about some of them?
>but that what is needed is a lot of hard work to organize the universe of
>content,
>and more hard work to organize the work of doing that organization.
>
>I happen to be one of the old optimistic guys. I claim to have been doing
>knowledge management (a recent buzzword in the industry) for almost 35
>years.
>In one of your earlier postings you disparaged a comparison between
>libraries and ontology, but I have several times drawn just that comparison
>in this discussion. The national libraries, such as the Library of
>Congress,
>I maintain, have been doing ontology for centuries. Interlocking
>classification
>systems are used as locating devices and networks of semantic pointers to
>organize arbitrarily large and diverse corpora of information. This has
>been
>done with very old (paper) technology, and more recently with electronic
>technologies.
You are using 'information' in a different sense. If ontologies were
like libraries, then the information in them would be basically text:
it would be meant to be read by intelligent human beings who can
understand it. Formal ontologies are not meant to be read by human
beings. They are meant to be input to mechanical reasoners. A library
simply holds information in the way that a cup holds water; but an
ontology *is* the information. All a library classification has to do
is get the reader to the book; after that they are on their own. But
an ontology has to represent what is in all the books (and indeed
what isnt even in the books, but which the writers of the books
assume that all their readers know; the everyday knowledge that any
competent adult reader would bring to bear on the task of
understanding natural langauge). That is what we have to contend
with; and that has not been done before by anyone. The only field
that has seriously attempted it is AI. Libraries have *never* been
doing ontology in this sense; the nearest anyone has come to it
(before there were computers) were the compliers of lexica and
thesauri, who tried to make taxonomic classifications of entire
languages. But library classification systems are completely
irrelevant (or at any rate, they seem completely irrelevant, and if
you think otherwise then you need to make a very strong case.)
>The work I see to do is the organization of classification structures, not
>breakthrough algorithms or knowledge representation schemes. You ask
>where we should start? I think we have started, with Ian Niles's work. I
>think we
>could do well to survey all known existing ontologies to determine what it
>would take to support navigation among them. I would like us to take a
>look
>at library classification schemes in the same way. I posted a provocative
>list of classifications for the concept of "system". I look around and see
>a
>wealth of starting places.
>
>Then, again, if you are right, this is not what the SUO is about. Rather,
>it is about some Quixotic mission that is, in your opinion, impossible.
The only reason I am involved with the business is because of that
awful word 'standard'. Few things in the history of mankind have done
more harm than badly designed 'standards'. If the SUO can be made
into a reasonable standard, then fine. But if it can only be a
harmful standard, then my mission is to prevent it from happening.
Right now, Ian's merged ontology looks a hell of a lot more like 2.
than like 1., and I dont see any methodology being suggested which is
likely to change that.
The technique being used here, to use an analogy which might convey
my views to a librarian, is like trying to write a universal novel by
cutting and pasting paragraphs from other works of literature and
then doing copy-editing. We need a rousing opening speech: OK,
there's one in Henry V, so in goes "O for a muse of fire..." We need
a love interest, so we splice in something from Jane Austin mixed
with selections from "Hedda Gabler". We need sex, and someone points
out that parts of the Bhagvad Gita are pretty steamy, but someone
else says that it really doesnt matter since sex is pretty much the
same everywhere, so we toss in a chapter from "Fanny Hill". It gets
a bit rough at the edges, but thats OK since we have a really good
index (from Charles Darwin's 'Descent of Man') and everything is
being fitted into that, so it's bound to finish up OK, right?
Pat Hayes
---------------------------------------------------------------------
IHMC (850)434 8903 home
40 South Alcaniz St. (850)202 4416 office
Pensacola, FL 32501 (850)202 4440 fax
phayes@ai.uwf.edu
http://www.coginst.uwf.edu/~phayes