[CL] Working on a definition of "ontology"
John F. Sowa
sowa at bestweb.net
Fri Jul 7 11:05:21 CDT 2006
Every quarter, I package up my email into convenient
batches for archiving. As I was going through 2Q06,
I came across the following point from April 10, which
I thought should be clarified:
RM> Further along in the book, Dr. Sowa indicates that
> "relations" are the same thing as "predicates", then says:
> In a conceptual graph, the boxes are called "concepts" and
> the circles are called "conceptual relations".
The terminology is somewhat variable among different authors,
but the most common position is that the word "predicate"
refers to the symbol and the word "relation" refers to the
mathematical structure that the symbol denotes -- i.e., a
predicate is a name of a relation.
In conceptual graphs, the terms "concept" and "conceptual
relation" refer to the nodes in a conceptual graph.
The type labels of conceptual relation nodes can be
mapped to and from the predicates of predicate calculus.
The type labels of concept nodes can be mapped to and
from monadic predicates of an untyped predicate calculus
or the types of a typed predicate calculus. For CL,
those mappings are defined in Annex B of the standard.
> So, if I read Dr. Sowa correctly, the ontology only defines
> the concepts; the relationships are defined by the logic.
The passage you quoted from my KR book might not have been
sufficiently clear, but in the sample ontology of Appendix C
of the KR book, I definitely include both concept types and
relation types in the ontology.
And in any case, a formal ontology is always defined in terms
of some version of logic. Therefore, all of those things
are defined by statements in logic, some of which may be
collected together in a package called an ontology.
More information about the CL