Re: Quoting and unquoting variables in KIF
sowa <sowa@turing.pacss.binghamton.edu>
Date: Tue, 13 Apr 93 12:23:09 EDT
From: sowa <sowa@turing.pacss.binghamton.edu>
Message-id: <9304131623.AA07840@turing.pacss.binghamton.edu>
To: schubert@cs.rochester.edu, sowa@turing.pacss.binghamton.edu
Subject: Re: Quoting and unquoting variables in KIF
Cc: cg@cs.umn.edu, interlingua@ISI.EDU, sowa@turing.pacss.binghamton.edu
Len,
Yes, I believe that there is an intuitive notion of what it means
for two different statements in two different languages to "mean the
same thing." And a definition stated in terms of the mappings between
different formal languages seems to be a good way to capture that
notion without getting into nonobservable mental entities. Any
proof-theoretic constraints on such mappings would have corresponding
effects on the associated models. What I am calling a proposition
is a language-related notion that is closely related to the
situation-semantics notion of "infon". Ideally, the syntactic
constraints on the mappings and the semantic constraints on the
models should lead to an isomorphism between propositions and
infons.
In any case, this discussion has diverged quite a bit from the
original question of the syntax of quotes and commas in KIF.
In conceptual graphs, I am following Peirce's conventions for
lines of identity that cross context boundaries. The contexts have
the effect of quoting the contained graphs, but they don't block
the coreference links that have already been drawn. What they do,
however, is determine the conditions for extending coreference links
>From outer contexts into inner contexts. Modal and intensional
operators, for example, may create "opaque" contexts that prevent
a coreference link from being extended into the context. But that
is not a result of the quoting effect of contexts; instead, it
results from the nature of the operators that are attached to the
contexts themselves; i.e., you can copy something from a knowledge
context to a belief context, but not vice-versa.
Those rules for copying (or "lifting" as McCarthy calls them) are
not associated with the context mechanism itself, but with the
operators that are applied to a context. With conceptual graphs,
I follow Peirce's "lightweight contexts" that have no semantics
associated with them other than the their ability to enclose
other graphs. The semantics derives from the axioms that one can
state about permissible copying and extending operations that may
be performed in the presence of various context-modifying operators
(negation, possibility, belief, knowing, hoping, wanting, fearing, etc).
This approach makes it fairly easy for me to translate conceptual
graphs to the KIF quotes. It also makes it easy to use the KIF
metalanguage facilities to state the axioms for operating on the
quoted contexts in the presence of various modal operators.
But I want to make it clear that I am not proposing any special
theories of knowledge or belief or temporal logic, etc. What I
am trying to do is to define a framework in which people can state
any such theory that they find convenient. Peirce was developing
that approach very nicely in the 1896 - 1914 period, and it is a
pity that logic got sidetracked into a backwater by Bertrand Russell.
John