RE: SUO: Language Discusssion (formerly under "Maintenance - related issues")
John,
. Thank you for this. Obviously what I am envisaging would be
a development of ACE that would be less clumsy than you exemplify. Surely it
wouldn't be so hard to extend forward from this by using concepts underlying
algorithms used by developers of language translators.
. I can't help thinking English speakers don't generally find
formalising their own language as "sexy" (read engaging) as doing it in a
highly symbolic language. The fact that Fuchs is German, and based in
Germany, tantalises me with the possibility that the foreignness of English
is one reason why the work has occurred there.
. I'd appreciate your perspective on the resource for a
powerful first order language I see contained within English.
. I feel a very important issue arising out of the reference
you gave is to not confuse the logic with its representation in a highly
symbolic language such as KIF. I get the impression from other e-mails that
some participants tend to regard expressions in such languages as actually
being "the logic", which of course it isn't.
. From here, I am wondering about the relative ease of
developing a symbolic language compared to one structured like a controlled
English. By this I am referring to our need to resolve all sorts of logical
quandaries, such as have generated large numbers of e-mails already. Would
it be so different if we were extending ACE? I'd love to do the work myself,
if only I could get a job that funded such work. I'm intrigued at your
comment that catering for such things as plurals are "the hard part".
. Where can I actually get a copy of ACE, or the spec for it?
I presume I missed the link in the places I looked so far.
. I already responded to your comments on ForTran in my
previous e-mail to you, but note your observation about the most commonly
used subset.
Cheers Graham Horn
Australian Institute of Health and Welfare
================================================
Phone: 02.6244.1094
Fax: 02.6244.1199
Email: Graham.Horn@aihw.gov.au <mailto:graham.horn@aihw.gov.au>
-----Original Message-----
From: John F. Sowa [mailto:sowa@bestweb.net]
Sent: Friday, October 20, 2000 2:01 AM
To: Horn, Graham; 'sowa@bestweb.net'; 'standard-upper-ontology@ieee.org'
Subject: RE: SUO: Language Discusssion (formerly under "Maintenance -
related issues")
Graham,
Some answers to your questions:
>How far do you feel ACE goes towards meeting SUO
>requirements? - or forming the basis of something that would?
They have implemented all of first-order logic, and the subset they have
already implemented is sufficient to simulate a Turing machine. They have
not, as yet, implemented sets in a way that takes advantage of English
syntax for dealing with plurals. In the current version of ACE, for
example, it is possible to state the first paragraph, but not possible to
restate it as the single sentence:
1. There is a room X. In X, there is a set A, and there
is a set B. Every element of A is a boy, and every
element of B is a girl. The cardinality of set A is
greater than the cardinality of set B.
2. There are more boys than girls in room X.
>What opportunity do you think there may be for extending it
>to cover more English structures, and to accommodate more
>subject diversity?
In principle, simple FOL as currently implemented in ACE, is capable of
expressing every subject on earth. However, that is a lot like saying that
a Turing machine is as expressive as any programming language ever written.
You have to add a very big "YES, BUT" because many of the things you would
state in modern languages with a short statement would take many megabytes
with a Turing machine.
Given that caveat, I would say that languages such as ACE can be used now,
and they can be improved, just as we have come a long way from FORTRAN I in
1957. But note that more than 40 years have elapsed, during which many
programmers continued to use assembly language because they weren't
satisfied with any of the existing "higher-level" languages. However, for
many purposes, the original subset of FORTRAN I is still the most widely
used part of the language.
>Is it the best Controlled Language, or is there something
>else you would also suggest warrants serious consideration?
ACE currently implements the "low-hanging fruit" -- that subset of English
which can be translated very easily to and from FOL. A good programmer with
a fair linguistic background can repeat the ACE achievement very quickly.
The hard part is to extend that subset to cover things like plurals,
contexts, etc. Some of those areas are doable in the short term, but a
complete solution will involve a lot of long-term research.
For further discussion of controlled English and the mapping to and from
natural language, see Section 5 of my paper on "Ontology, Metadata, and
Semiotics":
http://www.bestweb.net/~sowa/peirce/ontometa.htm
John Sowa