# Re: Parallel standards projects for KIF and CGs

sowa <sowa@turing.pacss.binghamton.edu>
Date: Mon, 2 May 94 05:24:38 EDT
From: sowa <sowa@turing.pacss.binghamton.edu>
Message-id: <9405020924.AA25996@turing.pacss.binghamton.edu>
To: interlingua@ISI.EDU, macgregor@ISI.EDU
Subject: Re: Parallel standards projects for KIF and CGs
Cc: cg@cs.umn.edu

Bob,
Some comments:
> Without adding types, a logic that distinguishes between types and
> unary relations cannot check out (apparently this is now fixed, or
> will be).
KIF has not added a new semantic construct for types, but Mike
has agreed to the syntactic extension for distinguishing them.
With that feature, I can map a CG into KIF and get a CG out that
looks pretty much like the one I sent in -- i.e. the types remain
types, and the relations remain relations.
> Presumably we will hit the same block on the notion of Proposition
> (but apparently we have 5 years in which to fight that battle).
> Expressing proposition using quotation does not solve the problem.
My solution with conceptual graphs is to define a type called Proposition,
which is linked via the statement relation (STMT) to a quoted statment
in some formal languages, such as KIF or CGs. Then the axioms for
STMT would specify the conditions for considering two different
statements to be expressions of the same proposition.
To avoid the complexity of writing STMT and quotes, I use type
coercion to expand the form
[Proposition: [Cat]->(ON)->[Mat] ]
into the lower level form
[Proposition]->(STMT)->[CG: [Cat]->(ON)->[Mat] ].
This graph says that there exists a proposition, which has
a statement in the CG form which follows. The type label CG
effectively acts like a quote.
Can you use something like this for KRSS?
John