Re: SUO: RE: CYC event vs. SUMO Process -- really different?
Richard Cooper prefers not to have a necessary
relation expressing a spatial location for Events:
Richard Cooper wrote:
> In a nutshell, this is what I've been trying to say:
>
> If no 4-space specification is required in either events or processes,
> yet a 4-space specification can optionally be attached to both events
> and processes, then I am satisfied with the representations. But
> I don't want to have to deal with defaults
> unless I somehow represent that 4-space information is important in the
> situation(s) I want to represent.
>
This situation will likely occur for many choices -- whether a
relation is necessary or optional. Some will prefer that it be
necessary, others optional. The same will be true of certain
classes of entities: those building specialized ontologies
may not "want to have to deal with" certain concepts irrelevant
to their specialized concerns. For example, someone building an
ontology for IC chip manufacturing will have no use for
biology or psychology or (almost all of) chemistry. But if
they want to preserve interoperability with other ontologies,
they will have to have some way of relating their ontology
to a standard that does have these things in it. If we ever
do get to build what could be considered a "standard" ontology
I think it will be very desirable to have some utilities to
help users extract out only those parts that they need,
such as some subset of all the classes and relations available in
the full SUO. Richard's comment also suggests that it may be
necessary to alter some axioms defining the relations, at least
with respect to whether they are necessary or optional.
One possibility is to make all relations optional. This has
one serious problem: without necessary relations we lose the
defining characteristics of classes. In addition, it would
shift the burden of control of inferencing completely to the
procedural program using the ontology, increasing the chances that
different programs would diverge in the inferences they generate
from the same data. Since the primary point of trying to
build a "standard" ontology is to enhance interoperability
of programs, it seems to follow that where relations can
plausibly be said to be necessary in reality (even though
not necessary for particular applications), the default
preference should be to make them necessary in the SUO.
With a mechanism to alter the default, individual users
can switch any given relation to "optional". If they
wish to interoperate with an ontology where that relation
is necessary, they would have to come to an agreement on
that issue with the owner of the other ontology.
Since it can plausibly be said that all processes located
in Time are also located in Space, I would prefer that the
spatial location relation on Events be a necessary one.
The default location could be as innocuous as "in the
Physical Universe". Users of the SUO who would rather
not even deal with such a default can make the relation
optional.
I think that if there is going to be any chance at all
of finding agreement on an SUO that will serve to enable
interoperabilty among ontology-driven applications, some users
will probably do well to tolerate a certain degree of "clutter"
of classes, relations, and inferences that they do not absolutely
need. At worst it may slow down the application slightly, but
whether that cost is burdensome will be something to be determined
by each user. If it's a problem, the defaults can be changed
and interoperability may be preserved within any community of
users that agree to use the same defaults.
Pat
=============================================
Patrick Cassidy
MICRA, Inc. || (908) 561-3416
735 Belvidere Ave. || (908) 668-5252 (if no answer)
Plainfield, NJ 07062-2054 || (908) 668-5904 (fax)
internet: cassidy@micra.com
=============================================