Re: Types vs. monadic relations
Robert MacGregor <macgreg@ISI.EDU>
Message-id: <9202040507.AA13025@quark.isi.edu>
To: sowa@watson.ibm.com, interlingua@ISI.EDU
Reply-To: macgregor@ISI.EDU
Subject: Re: Types vs. monadic relations
Date: Mon, 03 Feb 92 21:07:01 PST
From: Robert MacGregor <macgreg@ISI.EDU>
> > What do you consider to be the semantic difference between
> > a "type" and a "monadic relation"? Why should we care?
>
> a) (Ex)(red(x) & ball(x)).
> b) (Ex:ball)red(x).
> c) (Ex:red)ball(x).
> d) (Ex:red)(Ey:ball)x=y.
> .... But ontologically, [the above statements] they make
> different assumptions
> about the nature of reality.
You state that there is a difference, but you haven't yet said *what*
it is. I'd like a rule of inference or something similar to
motivate the difference. For example, suppose an end-user comes up
to me and asks what difference it will make to his application
whether they define Male-Person as a type or as a unary predicate.
I'd like to have a good answer for them. If the answer is that
doing it one way makes the system run faster, then the distinction
shouldn't necessarily even be considered declarative knowledge!
Your previous message next listed three arguments in favor of a
sorted/typed logic. I agree with all three, and I think KIF ought
to be typed, but that is a different question. Suppose I decide to
make every unary predicate in my KB a type. Then there is no
distinction between the two, but I can still use a typed logic.
You haven't yet presented a concrete (i.e., non-philosophical)
reason for making the distinction.
- Bob