Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: SUO: Where are the axioms that establish the condition of a class ?




Seth,

	See my comment below.

-Ian

> -----Original Message-----
> From: Seth Russell [mailto:seth@robustai.net]
> Sent: Tuesday, January 15, 2002 2:03 PM
> To: Ian Niles; standard-upper-ontology@ieee.org
> Subject: Re: SUO: Where are the axioms that establish the 
> condition of a
> class ?
> 
> 
> From: "Ian Niles" <iniles@teknowledge.com>
> 
> > I don't think you're proposed "criteria" definition has any 
> advantages
> over
> > this representation.
> 
> The only tangible advantage is that the KappaFn is to be 
> depreciated and has
> no axiomatic definition .. so if we don't define the KappaFn 
> and depreciate
> it, then we will loose this ability.   

It's true that 'KappaFn' has no axioms and is deprecated.  It is deprecated,
because in most cases it is possible to create a complex class by means of
the class-forming operators that don't give rise to Russell's Paradox, e.g.
'UnionFn', 'IntersectionFn', 'ComplementFn', etc.  Although 'KappaFn' has no
defining axiom, your proposed predicate "criteria" fares little better in
comparison.  

(<=>
    (criteria ?CLASS ?FORMULA)
    ( (exists ?FORMULA ?X)
           (<=>
                (equal ?FORMULA True)
                (instance ?X ?CLASS)

First of all, the axiom isn't true, because not every formula specifying the
intension of a class is true.  For example, the intension of the class of
unicorns is going to be something like 'an omnivorous quadruped with a horn
sticking out of its forehead', and the formula specifying this intension
will either be false or lack a truth value.  Thus, the axiom should say just
that the first argument of "criteria" is a 'Class' and the second argument
is a 'Formula'.  However, these are just argument-type restrictions, and, in
fact, 'KappaFn' has similiar such restrictions.  Thus, in sum, I don't see
that "criteria" has any advantages over 'KappaFn' (and the other
class-forming operators defined in the SUMO).

>I have some other 
> personal reasons
> why a binary relation that doesn't involve the use of the 
> term 'equal' would
> be more advantageous;  but I doubt those would mean much to 
> the rest of you
> all.
> 
> Seth Russell
> 
>