SUO: Re: CNN article about Cyc
John,
My major concern with Cyc is implied by your concern:
How do you maintain a net of a zillion axioms if .01 % change every year?
Furthermore, Cyc's matching is based on the assumption that terms mean
the the same thing, unless explicitly distinguished.
If need reliable structures - as in repetetive business transactions
(the intel folk need that less because they are still willing to vet
results manually) then you'd better make conservative assumptions,
namely that no term from a different context means the same thing
(i.e., refers to the same set of objects), unless explicitley stated
do so -- in an application context.
Gio
/Gio Wiederhold/
http://www-db.stanford.edu/people/gio.html