Re: properties as nodes etc.

From: "Thomas B. Passin" <tpassin@comcast.net>

>  Seth was showing templates for n-ary relationships,
>  and for templates to work with RDF there has to be some
>  predicate that connects a template to its instance.

Well the instance can be connected to ~the template~ indirectly through the
class.  That way we connect the instance to it's class as per RDF spec with
rdf:type.  We then define a new predicate called 'nthDomain' to connect the
class to the sequenced bag  which is ~the template~ .   This is shown in my
diagram:

http://robustai.net/mentography/arity.gif

The 'nthDomain' predicate is identical to the 'domain' in suo kif ... that
used to be called 'domain' in sumo when I made that diagram ... the
corresponding predicate in CycL is 'argIsa'.    Note that there is a direct
translation from (rdfs:domain and rdfs:range) to ( (nthDomain 1) and
(nthDomain 2)).  Yes, there is a very slight leap from arity 2 to arity 3,
because the relationship rdf:type is between a reified triple in the former,
and a node in the latter. I think we can live with that, don't you?

>RDF cannot (of course)
> express the semantics of that connection.

What do you mean but that?

We need the ability to make templates now!  It seems that we could agree on
some predicate defined in a suitable namespace and have them to use
collectively right now ... huh ... huh?

Seth Russell

Received on Monday, 1 July 2002 09:43:31 UTC