SUO: Re: Re: IFF Model Theory Ontology 1.0, Request For Comment
Hello Richard,
(preliminary responses below)
----- Original Message -----
From: "Richard Cooper" <rich@valutech.com>
To: "'Robert E. Kent'" <rekent@ontologos.org>; "Jon Awbrey"
<jawbrey@oakland.edu>; "West, Matthew R SITI-ITPSIE"
<Matthew.R.West@is.shell.com>
Cc: "SUO" <standard-upper-ontology@ieee.org>
Sent: Tuesday, May 14, 2002 1:17 PM
Subject: RE: Re: IFF Model Theory Ontology 1.0, Request For Comment
> In Open Knowledge Base Connectivity, at
> http://www.ai.sri.com/~okbc/spec/okbc2/okbc2.html
> An Entity (i.e., an instance of the :THING class) can be either
> a Class or an Individual, but these two are mutually exclusive.
> So an instance of a class is necessarily an Entity, and may or
> may not be an Individual. If not an Individual, it must be a
> class.
>
> So the method of identification for Individuals is different
> from Entities. It seems to me that identification ought to be
> designed specifically for Entities, since literally everything is
> descended from an Entity. So a pair of shoes is one Entity,
> and the left shoe and the right shoe of the same pair are
> also Entities. If I know that something is an Entity, but know
> nothing else about it, I can efficiently represent it relationally
> as I discover more and more information about the Entity. I can
> immediately assign it a unique ID for keying relations, although
> I may create other Entities that are semantically related to the
> Entity, themselves also Entities.
>
> The point is that the identification process would seem to have
> advantages in being Entity based rather than Individual based.
Yes, I believe that the approach used in the IFF-MT is compatible with what
you have said. The universe of an IFF-MT model would seem to correspond to
the OKBC notion of Entity. Perhaps we on the IFF team should also define an
IFF OKBC portal. Let me just hazard a guess (I have forgotten my OKBC)
based upon your comments above.
Let A be an OKBC knowledge base.
Then we make the following two metalevel identifications:
universe(A) = instance(entity(A)) = entity(instance(A)) = Entity
type(entity(A)) = entity(type(A)) = Class
We also introduce the following constraint (implicitly defining a set of
"individuals" in this context to be things that are not types):
Entity = Class + Individual (disjoint union)
Then signatures of relation types will be indexed collections of classes,
and the tuples (relational instances) will be indexed collections of
entities.
Here entity(A) is an IFF set, instance(A) is an IFF hypergraph, and type(A)
is an IFF type language (a kind of referenced hypergraph).
Robert E. Kent
rekent@ontologos.org