Date: Tue, 6 Jul 93 00:05:55 EDT
From: lojbab@grebyn.com (Logical Language Group)
Message-Id: <9307060405.AA28423@grebyn.com>
To: doug@netcom.com
Subject: Re: reply.txt - Lojban and Loglan: A Linguist's Questions and an Amateur's Answers
Cc: nsn@mullian.ee.mu.oz.au
X-Mozilla-Status: 0011

Using Lojban as an intermediate language, you of course have to find a way
to disambiguate metaphors etc. that occurred in the source language.
I would envision that the subset of Lojban used as an intermediate language
would rarely if ever use JCB's 'metaphors' (our term 'tanru'), which are
indeed semantically ambiguous.  The language has the grammar to syntactically
expand source language metaphors to make it clear what the semantics is, and
that disambiguity is more important (in my mind at least) than the space
saving that tanru cause by effectively compressing the semantics.

Indeed, we have pretty convincing evidence that LOjban is a superset of
PROLOG, so as an intermediate language, and trying to save development
money, we would probably use a Lojban subset as close as possible to
PROLOG in scope, adding in the bells and whistles that LOjban has that
PROLOG does not (especially important to the translation process wouold be
metalinguistic notes, attitudinals (which Lojban has to an enormously greater
degree than older Loglans), etc.).

I've got the proposal to send to you, but it will come oin a separate message.
I can't recall whether we said more about this subject.  One observation we
made is that Lojban has a little problem in selling as an intermediate
language.  It is so different from natlangs, and also from computer languages
that the problems we would forsee in intermediate language representation
are likley to be quite different from the problems that other NLP processors
face.  Lojban makes some things easy that other languages find hard, and
possibly in some cases, the reverse.  The areas where research has been
concentrated, of course, have been those that are the stumbling blocks
for other language types - some of which will be non-problems for Lojban,
whereas we will find stumbling blocks that other researchers are not working
on, primarily because they haven't gotten tthat far to find the problem, or
because it is a non-problem at the level of abstraction found in other
languages.

Nick Nicholas just did a term project that converts Lojban into PROLOG, and
I think, makes some semantic deductions.  It is printed in our new journal
issue, but he can probably send you a summary and/or answer questions
specific to your view of the problem (I don't even play an NLP researcher on TV
%^).

lojbab