Re: SUO: Re: Sowa's piece and thoughts on tractability
Leo, Jean-Luc, et al.,
I would like to start with a comment by Jean-Luc with which I disagree
very strongly. In fact, I believe that the attitude embodied in the
following comment virtually guarantees that no progress in solving
the problems of making different ontologies interoperable can
even begin to be done:
JLD> And, once more, DON'T AIM AT NATURAL LANGUAGES, for now and
> probably quite a long while.
I would also like to cite the term "AI-complete problem", which has
the following definition in the dictionary of computer jargon:
> Used to describe problems or subproblems in AI, to indicate that the
> solution presupposes a solution to the `strong AI problem' (that is,
> the synthesis of a human-level intelligence). A problem that is
> AI-complete is, in other words, just too hard.
>
> Examples of AI-complete problems are `The Vision Problem' (building
> a system that can see as well as a human) and `The Natural Language
> Problem' (building a system that can understand and speak a natural
> language as well as a human). These may appear to be modular, but all
> attempts so far (1999) to solve them have foundered on the amount of
> context information and `intelligence' they seem to require.
Jean-Luc's comment expresses the feeling, embodied in the term "AI
complete", that natural language understanding is too difficult to
consider. I agree that it is difficult, but I also believe that it
is no harder than the problem of reconciling or relating multiple
diverse ontologies to one another. I also believe that it is much
easier than the problem of developing a good monolithic ontology
that is suitable for a wide range of practical problems.
(I realize that is the goal of SUMO, but note that I added the words
"good" and "suitable". I don't believe that any ontology developed
with the methodology of SUMO could ever be good or suitable for a wide
range of practical problems. That doesn't mean that SUMO wouldn't be
useful as a resource that could be taken apart and be reused in a more
suitable form, but that is a different story.)
Furthermore, I would maintain that both ontology development and
NL understanding are AI-complete problems. Unlike some people who
immediately dismiss such problems as unworthy of serious consideration,
I believe that we can make a great deal of progress on both ontology
development and NL understanding if and only if we approach both of
them together. Approaching either one by itself is a total waste of
time. But both of them together are much easier to address and solve.
I would also like to cite the following point by Leo Obrst, with
which I agree very strongly:
LO> My point in all this: you need both (1) and (2).
>
> 1) the knowledge expressivity problem, i.e., encoding declarative
> representations which are close to the human conceptual level
>
> 2) the efficient run-time execution of such expressive knowledge
>
> Now, will the rich expressivity you have (in a very expressive FOL/HOL
> language) in (1) get lost or vitiated when you transform it based on
> (2)? Yes, but again there are smart things you can do to minimize
> this. But we need (1), i.e., we need expressivity and if you force
> knowledge workers to encode their human conceptual level
> representations to some "tractable" level of expressivity, ultimately
> you doom the field of KR to a fatal irrelevance.
I believe that there are automated and semiautomated tools for dealing
with the rich expressivity of NL, and those same tools are essential
for solving the ontology problems we have been discussing on this list
for the past two years.
I also believe that the brute-force hand coding that has been done
for SUMO (and for Cyc) is a dead end. I wouldn't want to throw away
that work because it is a valuable resource. But I believe that it
is essential to develop the appropriate tools and structures to assist
in the mapping from large NL resources (such as the entire Internet)
into our ontologies.
There are very good tools that have been developed that address these
problems, and I'll discuss them in future postings.
John Sowa