nl-kr-request@CS.ROCHESTER.EDU (NL-KR Moderator Brad Miller) (01/05/88)
NL-KR Digest (1/04/88 23:56:34) Volume 4 Number 1
Today's Topics:
Mailing list on narrowing
Re: online dictionary needed
Re: semantics of "unless"
Is linguistics a science?
Re: verbal behavior as a natural phenomenon
Re: "Linguistic Science", Computer Languages, et al
Submissions: NL-KR@CS.ROCHESTER.EDU
Requests, policy: NL-KR-REQUEST@CS.ROCHESTER.EDU
----------------------------------------------------------------------
Date: Wed, 16 Dec 87 15:45 EST
From: Uday S. Reddy <reddy@b.cs.uiuc.edu>
Subject: Mailing list on narrowing
[Excerpted from the PROLOG digest]
A new mailing list dedicated to the discussion of narrowing in
functional and equational languages is being created. Narrowing is an
operational mechanism using which expressions with free variables can
be "executed" producing solutions for the free variables. This is one
of the approaches to combining functional programming and logic
programming paradigms into a unified framework.
The mailing list will be based at University of Illinois, at the
address narrow@a.cs.uiuc.edu. Please send requests for addition to
the mailing list to narrow-request@a.cs.uiuc.edu or
reddy@a.cs.uiuc.edu.
-- Uday Reddy
------------------------------
Date: Tue, 29 Dec 87 15:15 EST
From: Richard Shu <rshu@zodiac.ads.com>
Subject: Re: online dictionary needed
In article <17262@glacier.STANFORD.EDU> jbn@glacier.UUCP (John B. Nagle) writes:
>
> The entire American Heritage Dictionary ... is available, along with various other
>reference works, in CD-ROM format from Microsoft.
>
How about foreign language dictionaries? Anybody know if these are available
anywhere?
Richard F. Shu
------------------------------
Date: Tue, 22 Dec 87 11:31 EST
From: John_M._Lawler@ub.cc.umich.edu
Subject: Re: Semantics of 'unless'
I'm not going to add to the cascade of examples and counter-examples here.
I'll just point to a reference. Jim McCawley's _Everything_that_linguists_
_have_always_wanted_to_know_about_logic*_(*but_were_ashamed_to_ask)_, 1981,
University of Chicago Press considers and rejects *all* truth-functional
approaches to natural-language connectives like => and ~ (which would
pretty well rule out all the proposals made here so far. I don't have it
to hand, so I can't state with accuracy that he includes a discussion of
'unless', but I seem to recall one. It's a well-known problem for ling-
uistic semantics.
-John Lawler (jlawler@ub.cc.umich.edu) University of Michigan
"...Oh, how I wish I knew someone who has thrown all his words away,
So I could discuss ideas with him!"
-- Chuang Tzu
------------------------------
Date: Thu, 31 Dec 87 18:56 EST
From: The devil himself <daemon@csd1.milw.wisc.edu>
Subject: Re: semantics of "unless"
Summary: Unless = If Not
The summary says it all. I should point out that it is also a useful
keyword to have in a programming language.
------------------------------
Date: Thu, 31 Dec 87 19:19 EST
From: The devil himself <daemon@csd1.milw.wisc.edu>
Subject: Re: The semantics of "unless"
Summary: Unless = Whenever Not
Again, the summary says it all. On a more serious note, you'll find
that "unless" taking on a temporal aspect is a special case of a more general
phenomena. In natural languages, time, causality and logical inference are
closely tied together. Perhaps I should include Motion in here too. This
relation can be brought out in the following paradigm:
TO AT FROM
EFFECT MEANS CAUSE
GOAL PATH SOURCE
CONCLUSION INFERENCE HYPOTHESIS
FUTURE PRESENT PAST
From: markh@csd4.milw.wisc.edu (Mark William Hopkins)
Path: csd4.milw.wisc.edu!markh
Words used in one paradigm can be metaphorically extended to the other
paradigms. Therefore "unless" can take on the meanings "whenever not",
"if not" (for Logic), "wherever not" and "however not" (for Causality)..
------------------------------
Date: Thu, 24 Dec 87 07:48 EST
From: "Celso Alvarez,4226 Dwin.,2103,,MD17" <sp202-ad@garnet.berkeley.edu>
Subject: Is linguistics a science?
Since I assume this to be an interdisciplinary forum, please
allow me to interfere.
Is linguistics a science? Well, you tell me what *type* of
linguistics you are referring to and we will be in a better
situation to assess whether it deserves the adjective of
'scientific'. I assume most of you are referring to generative
linguistics. In that sense, we should make a distinction between
the method it applies (formalization, generation and testing of
hypotheses, etc.) and the results it has so far achieved. Nobody
can deny that building hypotheses and generalizations about the
ways in which the linguistic system works, as well as
establishing predictions on which structures are grammatical, are
scientific methods. Limited predictions have been established
that, for example, certain structures are impossible (e.g. "*I
ed-talk" for "I talked", or "*Would theater you to go like the
to?"). The problem arises when we consider sequences that,
although non-grammatical, are communicative. Here we are entering
the realm of social (communicative) behavior, and in this case
the question "Is a science of linguistic behavior possible?"
should be rephrased as "Is ANY type of science of human behavior
possible?" Indeed, even ethnographers of speech like Hymes fell
into the temptation of formalization by suggesting that "rules"
for describing acts of speech should be included in a more
developed model of verbal communication. Linguistic pragmaticians
and linguists of discourse continue to fall into this temptation
by elaborating cumbersome rules that in a way resemble logical
operations. But the question remains whether logic-based
linguistics is still describing with a minimum degree of accuracy
and operativeness the processes of generation (production) of
speech. Is, then, this formalized linguistic to be labeled as
"non-scientific" because until present it has proven unable to
give full account of the process of grammatical speech
production, or to elaborate "computer models" of specific
languages? Well, let's see those working computer models of the
structure of the universe!
I'm not comparing the degree of structural complexity of a natural
language with that of the universe! But one of the reasons why
*natural*, and not artificial languages interest linguists
primarily is because the former ones are considerably more
complex than the latter ones. Esperanto seems a dull topic of
research because it is based on what linguists precisely don't
like -- regularities.
The fact is, I'm not specially interested in formal
grammars. Even accepting the validity of the
competence/performance dichotomy, I am more inclined towards a
study of what we actually encounter in human communication --
'performance', that is, speech. Even generative models which
resort to speakers' judgements in order to assess the
grammaticality of a given structure, or to non-linguistic
constraints such as the level of formality of discourse --
thus introducing a probabilistic component in their --
description, fall short from incorporating
successfully into their analysis what seem to be
*unquantifiable* realities because of their very social nature.
Those who are familiar with Labov's work in variable rules
probably are aware of this. For so-called exact sciences, the
question of the relationships subject-object of research might
not be an issue: one studies external realities. But the question
in linguistics and in other social sciences is a metatheoretical
one -- whether human subjects can elaborate working formal models
of social behavior, that is, models whose *objects* of study are
phenomena in which the *subjects* themselves play a significant
part.
Again, these dilemmae and arguments are not new, but they seem to
be recurring as the international academic community shows a
growing interest in qualifying or disqualifying certain
disciplines of knowledge as "scientific" or "non-scientific", not
on the basis of their results in contributing to the
understanding of social and physical realities, but on the
discourse on which they operate. Hence the interest of some
structural linguists, sociologists and anthropologists in
"raising" their respective disciplines to the level of "science",
and the interest of mathematicians, physicists, computer
scientists, logicians, etc. in restricting the access to the
world of knowledge and possession of "scientific truth."
These ideas might connect with some of the questions raised by
Doug Landauer: "Why are linguists making no attempts to design
languages with specific features, to see just how useful the
features are?" Well, useful for what? If D. Landauer means
"useful for communication", we already have more than 5,000
natural languages used for communication in the world and we
don't know how to describe them yet! Is it possible to design a
"logical, infallible language"? Theoretically, yes. Practically,
in my view makes little sense. I am not familiar with the issue,
but the closest to this designing that I have seen, with
"practical" applications, is the task of language planners. They
are, in a sense, language designers, in that they deliberately
take steps to shape the form of languages in an acute situation
of internal fragmentation, diversity, etc., by trying to reduce
lexicon, offering preferable morphological solutions, and
designing, fixing and implementing theoretically adequate
spelling systems. Their role? They are the technocrats of human
language, as language engineers are the technocrats of computer
languages. But, as for planning an entirely new language... well,
look at the situation of Esperanto in the world. For most, it is
a mere curiosity. For some, it is a hope. The question is not
designing a language. The questions are, who creates it, for what
purpose, who will and will not gain access to it, how
will it be socially used?
------------------------------
Date: Mon, 28 Dec 87 16:49 EST
From: Mark Mehdi Towfigh <mmtowfig@phoenix.Princeton.EDU>
Subject: Re: verbal behavior as a natural phenomenon
In article <27@gollum.Columbia.NCR.COM> rolandi@gollum.UUCP (Walter Rolandi)
writes:
>
>The question here becomes, "Why are these 'postdictive', historical
>measures preferred to the experimental examination of 'live data'?"
>Linguistics has (at least potentially) a great deal more to offer.
>
>Why don't linguists study actual verbal behavior?
>
But they do! I suppose the branch of linguistics which depends the
most on experimentation is Psycholinguistics; experiments are
constantly being carried out with the aim of testing hypotheses
about how language is organized in the brain.
A typical experiment, for example, might present a subject with a
list of word pairs which are historically (i.e. etymologically)
related. The subject will not be told this, however, as the aim
will be to see if the subject feels the pairs are in any way related
in his or her common usage. Results of this experiment from a wide
range of subjects might be used to test the hypothesis that we store
many historically related words as individual units in our mind,
instead of keeping a central form, with ways to get to other forms.
For example, do we have the mental equivalent of a little box with
the word "real", from which we derive "really", "reality", "unreal",
etc.?An experiment like this might be able to judge this.
Another branch of linguistics which undertakes a large amount of
field work, research, and experimentation is the branch of Phonology.
Phonologists collect data about the pronunciation of languages, and
then form rule systems about pronunciation in those languages. But
these rule systems are hypotheses, and must be tested against the
language itself. One major hypothesis which phonologists have
grappled with is the Null Hypothesis, which states that natural
language units are remembered individually, and that we do not use
generative phonological rules in speech. The Hypothesis says about
the English plural, then, that you just have to memorize it; the fact
that there is an "s"-sound at the end of "cats" and "tulips" is not
related to the fact that both base words ("cat" and "tulip") end in
an unvoiced non-sibilant consonant, and the fact that "thing" ends
in a voiced non-sibilant consonant does not necessarily predict the
"z"-sound you tack onto it to make it plural.
Well, this is just a sketchy picture, as I myself don't know a whole
lot about experimental linguistics; I am sure it exists, however.
=======================================================================
Mark Towfigh If there's one thing I like better than a bologna
and whipped cream sandwich, it's honey and ketchup.
UUCP: mmtowfig@phoenix.princeton.edu BITNET: 6110480@PUCC
------------------------------
Date: Sun, 3 Jan 88 09:43 EST
From: Greg Lee <lee@uhccux.UUCP>
Subject: Re: verbal behavior as a natural phenomenon
In article <27@gollum.Columbia.NCR.COM> rolandi@gollum.UUCP () writes:
>...
>Why don't linguists study actual verbal behavior?
What kind of behavior are we talking about here? Current theories
certainly do make predictions about the acceptability of language
forms. When people judge a form to be acceptable or unacceptable,
is that not "actual verbal behavior"? Well, whether or not it is,
there is a very good reason for studying and making theories about
such judgments rather than pronunciations themselves. Feasibility.
There is no theory even faintly in prospect which could predict what
a person is going to say next. If one insists for some reason (I
don't know what it would be) that only pronunciations are worthy
of study, the result of such insistence will be that one is reduced
to diagnostic taxonomical "theories" which make no predictions.
We cannot just look at pronunciations to test theories which make
predictions about acceptability, for the minor reason that people
make mistakes sometimes in what they say, and for the major reason
that the unacceptability of certain forms is often crucial.
Greg, lee@uhccux.uhcc.hawaii.edu
------------------------------
Date: Wed, 30 Dec 87 13:44 EST
From: John_M._Lawler@ub.cc.umich.edu
Subject: Re: "Linguistic Science", Computer Languages, et alii
It's important to understand that Science isn't any particular me-
thod, or collection of methods, though we sometimes like to pretend it is.
It's a social institution, or more precisely, a collection of social in-
stitutions that differ widely in their goals, cultures, beliefs, prac-
tices, and myths. There's practically nothing one can say theoretically
about What Science Is that isn't contradicted by one or more particular
sciences. John Chambers <jc@minya.UUCP> has already made some excellent
remarks on this score.
Like all social institutions, Science (and sciences -- note the para-
llelism here with the standard Linguistic distinction between "Language"
and "languages") is self-sustaining and self-justifying. It continues to
exist because it already exists and we are part of it; the reasons why it
exists (like the reasons the U.S.A., or the Boy Scouts, or F.I.D.E., or
NASA, or Rotary International exist) are useful for various purposes, but
I doubt you could find any social institution that is completely defined
by its expressed goals. There's a lot more going on.
The thing that distinguishes Science (if anything does) is that sci-
entists are the participants in the institution. (Definition: Science is
what Scientists do). Study scientists, then, to see what their motiva-
tions are, and you know more about Science. This much is evident, given
the preceding. Now we come to the personal opinion: for my money the most
common motivating factor in Science is essentially *esthetic* in nature
(assuming that curiosity is a possible vehicle of pleasure in its satis-
faction, which I will assume).
And let's not get into arguments about the difference between Science
and Art. Those are political distinctions originating only within the
last century, and have rarely led to anything but acrimony. Now, back to
Linguistics: Is It A Science? As Greg Lee <lee@uhccux.UUCP> and Rick
Wojcik <rwojcik@bcsaic.UUCP> have pointed out, that depends on how you
define Science. The real question is: What *kind* of thing (Science or
Whatever) is it? To find out, the history of the discipline becomes
important.
Linguists, as Rick noted, get to wear lots of hats. Linguistics is a
recent amalgam of a number of diverse ancient, medieval, and modern
traditions of scholarly inquiry, all of which take (various facets of)
human language as their subject matter. To begin with, there is the med-
ieval European tradition of The Trivium (Linguistics is Trivial in the
best sense of the term :-): Grammar, Logic, and Rhetoric. These were the
'kindergarten' subjects one had to know cold before studying anything
else. This trinity is still a major part of the core of linguistics. In
fact, one could make a case (I won't) for renaming them as Syntax, Se-
mantics, and Pragmatics.
More recently (18th-19th C), the disciplines of history, literature,
and anthropology became important parts of linguistic study, and the dis-
covery of the ancient Indian discipline of phonetics added what was the
first part of linguistic study that was recognizably a physical science by
modern standards. Since then, we have happily borrowed from psychology,
sociology, biology, geography, mathematics, and just about any other dis-
cipline we could find. No bones about it, and no apologies. It's *hard*
studying language, and we need all the help we can get. Linguistics,
being thus oriented towards its object of study rather than its meth-
odologies, slops untidily across the modern bureaucratic boundaries of So-
cial Science, Natural Science, and Humanities. It's a perennial problem
for modern universities as to how to classify us. Too bad for them.
As to paying back the loans, that's another story. Our export
product is admittedly small, largely because of trade restrictions.
Computer science, for instance, finds little of use in most linguistics
because human language is in fact predicated on having a human being in
circuit. The fact that language is important and that we have terms and
(perhaps -- here's the Whorf meta-problem again) concepts that seem to de-
scribe it as a unified and isolable phenomenon does *not* mean that it ex-
ists thereby as something that can or should be formalized separately from
all other human experience in terms that fit the esthetic requirements of
computer science. The null hypothesis would seem to be that it does not;
that its successful explanation requires explanation of (at least part of)
the human mind. The experience of computer science and psychology in
trying to use linguistic concepts and theories would seem to support this
hypothesis. Though the question is far from resolved.
Now, on to another topic, which may shed some light on why Linguis-
tics is so difficult to export: its basic premise contradicts one of the
most fundamental beliefs of Western society. Linguistics is based on the
assumption that its object of study, natural language, is spoken, not
written. This seems harmless until you consider the ramifications.
Biologically, one of the species traits of human beings is that they talk.
We (the species) have been doing this for at least a quarter-million
years. We are biologically adapted to it. Writing, by comparison, is
modern technology. It is now the case and has always been so that
literacy is a trait of only a minority of the human race, and we are *not*
biologically adapted to it. This is a *huge* difference.
But Western society is based on literacy, which means that the writ-
ten word is held to be the archetype of language. This isn't the first
society to be founded on technology, of course; but this means that what
linguistics has to offer is in many ways irrelevant to the technological
needs of the society, which have to do with easier and better access to
and control of written language. A naive but almost inescapable belief
that language consists of strings of letters has led to the current state
of affairs in which psychologists use either written or oral stimuli in-
distinguishably; in which 'string manipulation' is almost synonymous with
natural language processing; in which typing is the model for 'word pro-
cessing'. If one were going to use Linguistics successfully, one would
have to start with sound, not letters, and one would have to anchor one's
work in actual human interaction, not in written texts. This is not to
say there's nothing Linguistics has to say about writing, just that the
presuppositions of its analyses are rarely carried over to applied work,
with predictable results.
With all this in mind (ready?), we come to the interesting question
of computer languages. To my mind, (as to John Chambers') this is a real
linguistic question; at least it is if the study of writing is. Rick
Wojcik pointed out, in reply to John Chambers' picquant remarks, that nat-
ural languages were not invented by people. Quite true; language evolved
with humans. But writing didn't, and as I've tried to show, writing, not
natural language, is most of what's involved in 'Natural Language Pro-
cessing'. Hence computer languages are at least as valid as topics of
interest for linguists (Definition: Linguistics is What Linguists Do) as
is writing.
To start with, it's very instructive to look at the similarities and
differences between natural and computer languages. I prepared the fol-
lowing (incomplete) list a few years ago. Additions and discussion are
welcome.
NATURAL COMPUTER
------- --------
1. Infinite number of units ==SAME== [recursive only]
(S's, texts, etc.) (programs)
2. Finite number of atoms ==SAME==
(# > 2000, usu. > 20,000) (# < 1000)
3. Large and complex system Small and simple system
of combination rules of combination rules
4. Many inconsistencies, Names totally arbitrary,
exceptions, arbitrary grammars totally consistent.
features.
5. Ambiguity as design feature. Unambiguousness as design
feature.
6. Used between two or more Written by human programmer
human participants reci- for human user via machine
procally. interface.
7. Vocal prior to written form. Written prior to vocal form
(vocal form yet to come).
8. Naturally evolved, in vocal Product of human technology
form. in any form.
9. Many languages, many dialects. ==SAME==
10. Learned interactionally ==SAME==
(with exceptions). (with same exceptions).
11. Integrated with many human Integrated with many
phenomena (music, poetry, machine phenomena (memory
politics, humor, love, etc.) size, speed, hardware
configuration, etc.)
12. Exists in language ==SAME==
communities. (communities of users)
13. Used for many purposes. Used to solve problems.
14. Many fixed phrases, standard ==SAME==
utterances. (tools, functions, algorithms)
15. Must be interpreted. May be compiled.
___________________________________________________________________
...and a final question to chew on: if there were a programming language
designed specifically to handle natural language (*not* 'text', *not*
'strings'), what would it be like?
-John Lawler <jlawler@ub.cc.umich.edu>
University of Michigan
______________________________________________________________________
"Language is the most massive and inclusive art we know,
a mountainous and anonymous artifact of unconscious generations."
-- Edward Sapir
------------------------------
End of NL-KR Digest
*******************