alternative email@example.com (Peter F. Patel-Schneider)
Date: Thu, 9 Aug 90 14:17:38 EDT
From: firstname.lastname@example.org (Peter F. Patel-Schneider)
Subject: alternative interlingua
A Proposal for a Different Sort of Interlingua
I think that part of the problem with the current interlingua is that it is
currently designed to be able to implement all sorts of information. Note
that I said "implement" and not "represent", as I think that using
quotation is not really representing information. There are, of course,
advantages to the quotation approach. In particular, "extending" the
interlingua is easy. However, I do not think that much information, if
any, can be reliably passed between KR systems through the quotation
mechanisms of the current interlingua. For example, I think that it will
be very hard to pass information between LOOM and CLASSIC using the current
I think that the ideal interlingua would be able to "represent" all sorts of
information. Unfortunately, this is a "holy grail" of knowledge
represntation, and a goal that is not likely to be even approached in the
near future. Therefore, I suggest that an interlingua that is able to
represent most of the common, declarative types of information.
The interlingua would not have any extension facilities, quotation
facilities, or procedural facilities, as these are precisely the facilities
that hinder transmission of information between KR systems. There would be
many things that the interlingua would not be able to represent; there
would be no requirement that translation from a KR system into the
interlingua following by translation back out to the same system would be
the identity mapping.
I think that the interlingua should have facilities for representing:
1/ first-order modal logic,
2/ terminological definitions and sorted quantification,
3/ simple defaults with preference,
4/ predicate completion, local closed world statements, unique name assumptions,
5/ knowledge and belief,
I think that this set of facilities should provide approximate coverage for
most declarative information.
Now there is a lot of work that has to go into the preparation of this
interlingua. The interlingua should be a real logic, with a complete
syntax and a model-theoretic semantics. The best semantics for several of
the above facilities have not yet been determined. The approach I would
like to take is to come up with a reasonable semantics, leaning on the fact
that exact translation is not required.
I think that this approach to an interlingua will result in better
translation from one KR system to another.
Peter F. Patel-Schneider
PS: Note that this proposal is actually quite similar to Bob Macgregor's
alternative interlingua proposal. Also note that this proposal is NOT a
proposal for a high-level KR language per se, although it has some of the
flavour of a high-level KR language.