RE: SUO: Re: Ballot Comment
I broadly support what you are suggesting here.
In particular I think it is useful to be able to move
between theories at different levels for reasoning. So
you might start off with a question about gravity
expressed in every day terms, but move to a more rigorous
level to work out the answer. The rationale for this is
that what is complex for a human is sometimes easy for
a computer, and vice-versa, plus you are less prone to
error by moving to the more rigorous level.
There's a bit to do between here and there though.
Shell Information Technology International Limited
Shell Centre, London SE1 7NA, United Kingdom
Tel: +44 20 7934 4490 Other Tel: +44 7796 336538
> -----Original Message-----
> From: John F. Sowa [mailto:firstname.lastname@example.org]
> Sent: 17 August 2001 15:12
> To: West, Matthew R SITI-GREA-UK
> Cc: 'pat hayes'; Adam Pease; email@example.com
> Subject: Re: SUO: Re: Ballot Comment
> Matthew, Pat, and Adam,
> And following are more examples why a monolithic ontology won't work.
> In the framework I proposed in my two talks, you can have a hierarchy
> of concept types that include Space, Time, etc. Then,
> 1. Informal statements are mapped to the categories in the lattice,
> which have may different detailed definitions in
> different theories.
> 2. Some sentences are general enough that they can be answered in
> either theory (e.g., "What time is it here and now?").
> 3. More detailed problem specifications might only make sense in
> one theory or the other. The system tries both mappings and
> takes the one that has the best fit.
> 4. Other problems might not have a clean mapping to either theory.
> In that case, the system reports the areas that don't match and
> carries on a dialog with the user to refine the specification.
> I'm not claiming that the lattice of all theories is a magic solution
> to all problems. But I am claiming that it can handle a lot more of
> them than a single monolithic ontology.
> John Sowa
> "West, Matthew R SITI-GREA-UK" wrote:
> > Dear Pat and Adam,
> > Pat wrote:
> > >
> > > One thing that is for certain is, that if it is detailed
> enough to be
> > > useful, then some users will find it intrusive and
> awkward, perhaps
> > > to the point of being unusable. If it treats time
> > > the Barry Smiths of this world will refuse to have anything to do
> > > with it; if it fails to treat time four-dimensionally, the process
> > > control community will refuse to have anything to do with
> it. If it
> > > tries to both, then it will either become so confused that only
> > > highly trained specialists will be able to use it
> nontrivially (like
> > > Cyc), or else it will have to incorporate the kind of translations
> > > between ontological frameworks that would allow the rival views to
> > > coexist. And if it can do that, then it doesn't need to be a
> > > 'standard upper ontology', because with such translations
> > > we can translate between different upper ontologies.
> > >
> > Well first I agree with the thrust of Pat's argument. The idea of a
> > standard ontology in terms of "the one true way" to
> describe the world
> > is problematic.
> > On the other hand, a standard expression of some ontology(ies) is
> > potentially useful. Here we are not saying something about
> the ontology
> > itself, but of a standard representation of some ontology, so that
> > references to it can be made in a way that (dumb) computer
> systems can
> > recognise as being of the same thing. This basically comes
> down to agreeing
> > standard names for concepts, and defining axioms in a
> standard language.
> > You might then have a standard expression of a number of
> ontologies, without
> > being judgemental about the ontology itself. I think this
> has more chance of
> > being useful at this stage in the development of ontology
> and metaphysics.
> > Matthew West