Re: [topicmaps-comment] OpenCyc and Topic Maps: Progress Report

["W. Eliot Kimber" <eliot@isogen.com>]

"W. Eliot Kimber" wrote:

> This is not the place to argue about the Cyc ontology, but I think the
> distinction made is appropriate: all dogs are domesticated animals but
> not all dogs are pets. However, if you created an instance of dog and
> asserted that it resides in a dwelling with humans Cyc will probably
> infer that that dog is also a pet--I'll try it out when I get a chance.

At least with the OpenCyc KB, there is no inference that relates dogs in
residence to pets. But in thinking about it, I think this makes
sense--there could be any number of reasons that a dog resides in a
house, so the assumption that it is the pet of the other residents is
not warranted. Cyc can probably determine that a dog could potentially
be a pet of one of the human residents of its house but can't assume
that it is the pet of any of them. But I don't yet know how to ask Cyc
whether a particular proposition is potentially true.

I think this just goes to point up how complex the whole task of
representing common knowledge is. It also points up the need for
contexts, which Cyc handles well. For example, Martin could easily
define a context for himself in which there is a default implication
relating dogs to fellow dwelling residents.

Cheers,

E.

-- 
W. Eliot Kimber, eliot@isogen.com
Consultant, ISOGEN International

1016 La Posada Dr., Suite 240
Austin, TX  78752 Phone: 512.656.4139