SUO: RE: KIF syntax and semantics and a basic ontology
Hello, Chris, thanks for your reflections.
> I suspect, listening to this exchange, that the problem is that you
> and Matthew (and I) come from different domains - with different
> interests.
>
> So, for example, you wrote:
> > > Suppose one wants to develop an ontology that embodies a conception
> > > of time on which the future is not determinate, and hence on which
> > > there are past and present objects, but no determinate future
> > > objects. (Such a conception might be useful, say, in a
> > > manufacturing system that has to calculate on the fly the number of
> > > kind of objects it is going to produce during any given run
> > > depending on demand.) On such a conception, the class of humans
> > > grows moment by moment with each birth (or conception, or whatever
> > > your criterion is for being human).
>
> Within your community ...
Just to be clear, although my home base is an ivory tower, and although
I do a good bit of work of a more theoretical nature, I have worked
regularly on the development of "real world" systems for solving "real
world" problems for "real world" organizations for 15 years.
> ...it may seem sensible to produce an axiomatisation that is a
> 'language' that can and will be used to characterize several
> conceptualizations - often of the same domain.
Well, only in the sense that the axiomatization might not decide a
certain question.
> Use your example, in manufacturing, you could end up with two or more
> conceptualizations, of 'human' each with a different intended
> interpretation function for the 'objects' you have introduced here. In
> other words, you are axiomitising a range of possible interpretations.
Not in the sense, which you seem to be implying, that a single
axiomatization will contain different characterizations of a single
class like HUMAN.
> At the risk of re-opening an old debate - I presume that this means
> this version of KIF is intended to support multiple ontologies - not
> just a monolithic standard.
I believe KIF is irrelevant to your point. KIF itself is just a
linguistic framework, and *any* linguistic framework is capable of
supporting multiple ontologies. You don't get an ontology proper until
you choose a "namespace" (i.e., names for your distinguished objects,
classes, relations, etc) and write some axioms for it.
> However in the business community, where we deal with the 'real'
> commercial problems :)
See above for credentials! ;-)
> it seems to me that the problem we face when trying to get
> inter-operability is to try and find a 'standard' ontology for a
> domain - one where we can agree consistently on a single
> interpretation. So deliberately allowing for different interpretations
> goes against the grain - it seems like deliberately choosing to be
> equivocal.
I understand this motivation for a standard, and agree that it is one
scenario of use for an "upper" ontology. However, another is to
facilitate the integration of *existing* ontologies, in which case
an upper ontology that is compatible with ontologies that differ in
detail seems in order. But I wasn't necessarily thinking of this when I
left out extensionality for classes (which seems to be the actual issue
between Matthew and me that is motivating your discussion here). It
just seemed to me that there was no unanimity even within the SUO
discussion group itself about the status of extensionality, so rather
than legislate a decision I just left the issue open.
> Maybe it would be useful to distinguish between 'theories' that have a
> single intended interpretation and those that, deliberately, have a
> range of possible intended interpretations. It may also be useful to
> explore the possible ranges of intended interpretation.
Maybe, though the notion of an intended interpretation is extremely
problematic. But I agree that it might be useful to flag those points
in an ontology where a question about the nature of a given class is
left open.
> It seems to me that one important question is whether the cost of
> including this kind of 'flexibility' in KIF is that is implicitly
> prefers some of the range of possible full blown ontologies that can
> be described.
Again, just to emphasize the point, KIF is not an ontology, it is a
framework for constructing ontologies.
> However, as others have said, this is not to disparage in any way the
> effort you and Pat have been making to date.
Of course, thanks.
Regards,
-chris
--
Christopher Menzel # web: philebus.tamu.edu/~cmenzel
Philosophy, Texas A&M University # net: chris.menzel@tamu.edu
College Station, TX 77843-4237 # vox: (979) 845-8764