Re: Classes and predicates as first class objects

On Thu, 15 Aug 2002, Dan Brickley wrote:

[with regard to quantification over predicates (properties) in CycL]
> Interesting. Thanks for the example/pointers. Having implemented this, can
> you folks comment on whether such language features have proved
> problematic, useful etc...? Input from folk who've implemented large scale
> KR systems is one of the reasons we set up www-rdf-logic.

In Cyc, rules constitute only 5% of the KB content, ground facts are the
remainder.  And among the rules only a small fraction quantify over
predicates.  So I would summarize our experience with higher order logic
as useful for a few cases, yet worth including if your goal is
symbolically representing the full range of world phenomena.

And because we have special case inference engine support for the common
higher order query cases, we avoid much generic rule backchaining for
them.

Because Cyc has over 10,000 predicates (relationships between objects) it
is a challenge for the user to correctly specify the appropriate predicate
to express a desired relationship.  Although we now have RKF (Rapid
Knowledge Formation) tools to assist this search and clarification, the
primary use of quantification over predicates has been in queries.  You
can ask for example which predicates hold between concepts by making the
predicate position a variable in the query.

I can imagine a query language for OWL that allows variables in the
property position, even though quantification over properties is not part
of the asserted OWL language.

-Steve

-- 
===========================================================
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
         download OpenCyc at http://www.opencyc.org
===========================================================

Received on Thursday, 15 August 2002 14:28:56 UTC