using the interlingua
pfps@allegra.tempo.nj.att.com (Peter F. Patel-Schneider)
Message-id: <9008161411.AA22778@vaxa.isi.edu>
Date: Thu, 16 Aug 90 10:13:58 EDT
From: pfps@allegra.tempo.nj.att.com (Peter F. Patel-Schneider)
To: bobrow@parc.xerox.com
In-reply-to: Danny Bobrow's message of Tue, 14 Aug 90 11:23:27 -0700 (PDT) <8am46TcB0V0wQ_R5ND@nero.parc.xerox.com>
Subject: using the interlingua
Cc: interlingua@vaxa.isi.edu
> Date: Tue, 14 Aug 90 11:23:27 -0700 (PDT)
> From: Danny Bobrow <bobrow@parc.xerox.com>
> References: <9008131834.AA03457@arisia.Xerox.COM>
>
> Excerpts from mail: 13-Aug-90 alternative interlingua P.
> Patel-Schneider@alleg (1122)
>
>
> > I am afraid that the major use of the interlingua will turn out to be
> > satisfying a, possibly implied, desire of various funding agencies. In
> > this climate, the main gain from using the interlingua will be to state
> > that it was used, and not putting it to good use. So, if it is easiest to
> > use the interlingua via a side-agreement between representation systems
> > that subverts the meaning of statements, then it will be so used.
>
> I believe that the people involved in the effort to build a common
> interlingua are people committed to sharing knowledge. So if the
> interlingua is any good (and we ought to test it before saying it is an
> interlingua -- or especially before saying it is a standard), then it
> should have significant use in sharing knowledge. I am sure that all
> the people have enough else to do that working out ways to subvert the
> system (rather than not using it at all) will be low on their priority
> list. I hate to see us commit to theDykstra religious fallacy "if a
> feature can be misused, then forbid it. " We are in fact a scientific
> community and can publicly judge how things are used.
>
I'm not worried about people who currently have a track record of doing
good KR research, and who are interested in designing and creating an
interlingua. These people will, I think, use the interlingua
appropriately. I am worried about people who are trying to get initial
funding. These people may have to demonstrate that they can share
knowledge in their KR system, and the easiest way to do this may be via
side-agreements.
Now if you were not particularly interested in the interlingua per se, and
had to demonstrate sharability in order to get funding, what would you do?
>
> > Forbidding this sort of side-agreement, and requiring that the mapping into
> > the interlingua be faithful to the meaning of the constructs of the source
> > language, and not allowing a quote mechanism, would, I think, make it much
> > harder for these political pressures to overwhelm the research benefits
> > inherent in a good interlingua.
>
> I think there ae significant reseaarch benefits for learning how to
> capture in better ways things which have been done through side
> agreements. But don't you think it is important to try to understand
> what we can get as common in current systems now?
>
> > My "interest" in an interlingua is that it is an attempt to build a
> > universal, or at least more powerful, representation logic. The "utility"
> > of an interlingua is simply using it as an intermediary to translate
> > between two representation systems. I painted a picture where this
> > "utility" did not come along with any real representational benefits.
>
> Do you believe that a universal (or even much more powerful
> representation logic) is just some design time away from coming into
> existence. I understood you and several others at the conference to be
> saying that there are interesting (read HARD) problems to be solved to
> make such a well-founded system. I support such research. Should the
> working group process that is going towards developing a standard now be
> sidetracked awaiting this hoped for train? I hope the result of the
> working group's deliberations are an initial language (call it SKL-1), a
> process for the evaluation of this language in use, and a process for
> the evolution of the language over the next several years (SKL-2, SKL-3
> ...). Any one of these could be a major jump (given that it could be a
> jump to something that subsumes the current state -- not necessarily
> syntactically-- and much more expressive). Is this too polyanna-ish to
> believe in?
I believe that we should try to create a representation logic that is more
powerful than first-order logic, and that can be used to represent quite a
wide variety of knowledge. There are some hard problems in this endeavour,
but I think that this approach will create a better interlingua than the
current, quotation-based interlingua. I don't know if I would go so far as
to say that the current interlingua is no better than straight first-order
logic with no extension mechanism, but neither am I prepared to say that
the current interlingua is significantly better than straight first-order
logic.
peter