From: cbmvax!uunet!PRC.Unisys.COM!dave
Message-Id: <9108081712.AA03626@gem.PRC.Unisys.COM>
Date: Thu, 8 Aug 91 13:11:51 EDT
To: cbmvax!snark.thyrsus.com!lojbab
Cc: lojban-list@snark.thyrsus.com
In-Reply-To: Bob LeChevalier's message of Thu, 8 Aug 91 00:08 EDT <m0k81fD-00014jC@snark.thyrsus.com>
Subject: Re: response to Dave Matuszek
Status: RO


Thanks for your thoughtful reply to my rather acerbic posting.

> 1. I was stating the argument from the linguist's point of view, which
> I have admittedly pretty much adopted.  But recognize that what I
> said, even if it is a turnoff, is reality.  Linguists will not be much
> interested unless we can meet THEIR standards.  A Sapir-Whorf test is
> a linguistics experiment, and linguists have to be involved and
> satisfied that their involvemnet is meaningful.

I entirely accept your point that it is important to you to have
linguists accept your work.

Many years ago I did graduate work in Mathematical Psychology, hoping
to learn something about human cognition.  I didn't.  I moved to
Computer Science (and Artificial Intelligence), because I came to
believe that computer scientists, by attempting to understand
intelligence well enough to build their own, were far more likely to
advance our knowledge of the nature of human cognition.  I still
believe this.

I've been involved in natural language processing by computer (mostly
teaching, not doing, unfortunately).  What I know of linguistics comes
from my attempts to find something worthwhile to use in my own
programming.  However, I differ from much of the AI community in that
I think linguists are not only barking up the wrong tree, but barking
in the wrong forest....  Be that as it may, I think your attempts to
construct a new language, based on logical principles rather than on
existing grammars, are far more likely to shed light on the nature of
language than anything the "real" linguists are doing.  I feel this
way because my strong "learn by doing, not by observing" attitude
applies to linguistics as readily as to artificial intelligence.

Since what you are doing is quite different from what linguists in
general are doing, your work is not going to be accepted as
"linguistics."  Linguistics is what linguists do.  You're not doing
that.

At the end of the '60s, psychologists were entirely concerned with
emotions, not at all with reasoning.  (Believe me, I was there.)  In
1969 (just as I gave up and moved into computer science) the new field
of "cognitive psychology" was created, in large part by psychologists
who had started playing with computers.

[Begin unsolicited advice]

The point I want to make is that, while Lojban will not be accepted as
linguistics, you (or someone) should attempt to found a new subarea of
linguistics, with some suitably descriptive but catchy title, that is
devoted to the things one can learn from, or should know in order to
build, conlangs such as Loglan, Lojban, Gua!spi (did I spell that
right?), etc.  [I'd like to see it include the problem of constructing
languages for communication between humans and computers, but that's
just my personal interest showing.]  "Constructive Linguistics"? --
surely we can do better than that!

Once you have the buzzword, immortalize it in the title of a book or a
scientific journal (one carefully NOT limited to Lojban).  This will
attract young linguists and probably a few AI types, and if there's
any substance to the field, it will catch a couple of imaginations
(among readers young enough to still have them), start to feed on
itself, and grow.

[End unsolicited advice]

I don't think Sapir-Whorf is particularly important.  It's my
understanding that most linguists accept the Sapir-Whorf Hypothesis as
somewhat true, but not really relevant to anything.  Even if S-W is
true in the strong form, Lojban won't get you more than one step
outside English.  (But I'm glad to see you take that step.)

I'm glad you're using a context-free grammar, because that removes one
needless complication in getting a computer to understand Lojban.
However, that also effectively removes the only thing modern linguists
are interested in:  adding more epicycles to their grammars.

> If the language is going to ever have practical application (i.e. if it is
> ever to be a real langauge), we have to sell the language to one of the
> three 1) an international language community 2) the artificial  intelligence
> mnatural language processing community 3) the linguistic communtiy.

As should be apparent by now, I am more interested in "theoretical"
than in "practical" applications.  (The quote marks are because this
distinction is not nearly as clear as it sounds; it is just a
convenient labeling.)  How does language work?  How do people use
language?  Is language involved in reasoning, and if so, how?  How can
we (computationally) represent and manipulate knowledge expressed in
language?

To the extent that you ignore theoretical issues, you will be ignored
by (2) the artificial intelligence natural language processing
community.  Those in the AI community enamored of syntactic language
processing will find nothing in Lojban to help them process English;
those enamored of semantics will find you are not addressing issues of
interest to them.  IMHO, of course.

I think what you are doing is potentially very important to natural
language processing.  I would like to see you get more involved in
hard theoretical issues of relevance to AI, because I think Lojban
might make a solid contribution here.  If you follow my suggestions as
given above, you might even form the foundation of some "real"
linguistics, in the correct forest....

> 4. I am not hostile to General Semantics, though I admit to knowing
> not that much about it.  It is merely one theory among many, and I am
> doing my best to make the language design independent of various
> metaphysical theories, and especially those that linguists are prone
> to dismiss.

I do not claim you should incorporate General Semantics into your
language.  (I think you may be confusing me with Eric Raymond, who did
suggest this.)  I will be blunt, in order to avoid any
misunderstanding--I think you, personally, should learn a bit about
General Semantics.  You occasionally say things to which my reaction
is "No, no, that's meaningless!"  As Linus Pauling (?) once said of
someone's theory, "That's not right.  It's not even wrong!"  Please
don't take this as any sort of a personal attack; it's merely more
unsolicited advice.  I admit that it is entirely possible I am simply
misunderstanding what you say in such a way as to provoke this
reaction in me, in which case the advice is inappropriate.

When I think about why I feel this way, I come up with ideas that I
absorbed decades ago from some books I once read about General
Semantics.  Doubtless there are other sources for these ideas; Martin
Gardner remarked, in his critique of General Semantics, that
everything that was right about it was not original to it.  I suspect
Bacon may have been a primary source, but my knowledge of the
philosophy of science is too weak to say for sure.  However, some of
the ideas are good ones, and deserve wider circulation.

I will attempt to scare up a couple of suitable current references,
and if I succeed I will post to this group.


-- Dave Matuszek (dave@prc.unisys.com)  I don't speak for my employer. --
-------------------------------------------------------------------------
| Flon's Axiom:                                                         |
|   "There does not now, nor will there ever, exist a programming       |
|    language in which it is the least bit hard to write bad programs." |
-------------------------------------------------------------------------