W3C home > Mailing lists > Public > www-rdf-logic@w3.org > May 2001

Re: What do the ontologists want

From: Stephen Reed <reed@cyc.com>
Date: Fri, 18 May 2001 15:39:15 -0500 (CDT)
To: Seth Russell <seth@robustai.net>
cc: www-rdf-logic@w3.org
Message-ID: <Pine.LNX.4.21.0105181445300.24478-100000@balrog.cyc.com>

Regarding CycL, the KR language for the Cyc KB, the vast majority of
our assertions are binary relationships, which can be easily expressed as
triples. We use higher arity relations for situations in which the
relationship between two concepts is always with regard to one or more
additional concepts, such as the ternary relationship taxRate having three
arguments Being, TaxType, Rate.  When we change the meaning of a predicate
we must modify all the current instances and derived facts which are
impacted.  Commonly we use one-time forward rules to make this task
easy. I agree that binary predicates are more flexible in that if one of
the arguments is not known, then simply do not make the assertion, as
compared with what to do about some N-ary relationships in which less than
N concepts are known.

I understand and agree with the reservations expressed about RDF
reification, and hope that RDF is revised to enable the expression of a
logical formula without resorting to reification.  Otherwise we will
handle negation and other logical operators in a fashion most compatible
with other DAML+OIL early adopters.


On Fri, 18 May 2001, Seth Russell wrote:
> ...
> Let's suppose I compose a CycL kind of predicate like this {f,  Afrom,  Bto,
> Csomething} and define the meaning of each position in the tuple, and
> accumulate a bunch of instances of the tuples in my graph, and  then
> subsequently find that I was missing a ~Dwiggy~ part of the concept.  With
> the CycL technique I must redraw all of my predicates according to the new
> knowledge ... not so with a node based slot system as proposed in [1].  With
> such a slot based system I can very easily write into the inference engine
> the idea that we may or may not have complete information about the concept.
> The class\model nodes of instances tells us what we should anticipate and\or
> seek out.  I call that more flexiable and more attuned to the way real world
> models evolve.

Stephen L. Reed                  phone:  512.342.4036
Cycorp, Suite 100                  fax:  512.342.4040
3721 Executive Center Drive      email:  reed@cyc.com
Austin, TX 78731                   web:  http://www.cyc.com
Received on Friday, 18 May 2001 16:39:36 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:45:37 UTC