Re: Ontolingua vis a vis Cyc

thompson@zuben.ca.boeing.com (John Thompson)
Date: Fri, 19 May 95 10:41:01 PDT
From: thompson@zuben.ca.boeing.com (John Thompson)
Message-id: <9505191741.AA20989@zuben.ca.boeing.com>
To: ontolingua@hpp.stanford.edu, CCHERPAS@cccpp.com
Subject: Re: Ontolingua vis a vis Cyc


> Date: Tue, 16 May 95 11:35:37 pst
> From: "CCHERPAS" <CCHERPAS@cccpp.com>
> To: ontolingua@HPP.Stanford.EDU
> Subject: Ontolingua vis a vis Cyc
> 
> Y'all,
> 
>  Is anyone out there familiar with the Cyc project led
>  by Doug Lenat at MCC?  If so, how does it compare with
>  the work being done in ontolingua?
> 
> Best regards,
> cc

Since I haven't seen any other replies to this question yet, I`ll take
a stab at it.  I was working with Cyc in 1992 and looking at Ontolingua
in the first part of 1993, so I can make an outdated comparison (at the
risk of offending my friends in either project).

One major difference between the Cyc and Ontolingua efforts is that Cyc
enforces a global ontology, while Ontolingua's knowledge bases are
being developed more or less bottom-up.  In Cyc every new class that
you create must be placed under the correct superclass(es) and must
also be declared a member of the correct metaclass(es).  This means
becoming familiar with Cyc's exisiting ontology before you define any
classes of your own.  Ontolingua does not provide a global ontology and
does not require you to name the superclass and metaclass that your new
class belongs to.  But the expressive power of both notations is very
similar.

Another major difference is that Cyc knowledge bases are strongly
oriented toward common sense knowledge, while the Ontolingua knowledge
bases I've seen have been devoted to narrow topics in math, physics,
and engineering.  The Cyc KB's intentionally have a relatively naive,
informal view of the world, while the Ontolingua KB's are relatively
rigorous, precise, and mathematical.  Cyc is more willing to tackle
difficult subjects like human emotions, values, goals, and so forth,
while the Ontolingua KB's have stayed with subjects that have
mathematical definitions.  Cyc may be more useful for Natural Language
Understanding and for making common sense checks on data values in
databases, while the Ontolingua KB's may be more useful for exchanging
engineering knowledge.

Finally, Cyc is being developed by a single team that closely
coordinates its efforts and constantly looks for patterns of expression
that could be simplified with new predicates that everyone on the team
agrees to use.  Ontolingua is more of an open effort, where anyone who
wants to develop a new KB can do so, using his or her own style of
expression, and making use of whichever existing KB's seem useful.  And
obviously Cyc is a proprietary product (now a separate spin-off company
from MCC, called Cycorp), while the Ontolingua effort is an open
government project under the ARPA Knowledge Sharing Effort.

I hope to hear other comparisons expressed here.  (My opinions do not
represent any official position of my employer.)

John A. Thompson
The Boeing Company
thompson@zuben.ca.boeing.com