Content-Type: text/plain; charset="us-ascii"
Date: Mon, 14 Feb 1994 12:16:45 -0800
John Sowa writes:
>That goal of an implementation-independent way of talking about
>propositions and concepts was my motivation for defining a
>proposition as an "equivalence class of intertranslatable sentences
>in a family of languages." If your family has only one language,
>and your translation rule is identity, then the terms sentence
>and proposition are in exact one-for-one correspondence. But if
>we are concerned about knowledge sharing in multiple languages and
>implementations, the notion of propositions independent of any
>particular representation becomes essential.
I concur with this statement. However, note that the notion of
"proposition" is absent in KIF. Thus, while nothing prevents one
>From using KIF to axiomatize the notion of equivalence classes
among sentences, such an axiomatization is subject to the "you can
check in but you can't check out" problem that still prevents KIF
>From being a true Interlingua.
It possible that a solution is as simple as defining a couple of
relations, e.g., "proposition" (a unary relation that identifies
the class of propositions) and "sentence-proposition", a binary
relation that maps a sentence to an equivalence class (i.e., to
a proposition). Possibly just the second relation is sufficient.
My purpose here is not to submit a proposal for representing
propositional equivalence classes, but rather, to point out that
this functionality is missing in KIF.
The necessary primitive(s) to define propositions could be placed
in an ontology rather than in KIF, just as the notion of "Type"
probably belongs in an Ontology (and not in KIF). However,
types and propositions have not been established as notions
within KIF, and no proposal for a basic ontology has been entertained,
so for now KIF is simply inadequate in this regard.
Robert M. MacGregor firstname.lastname@example.org
USC/ISI, 4676 Admiralty Way, Marina del Rey, CA 90292 (310) 822-1511