LINGUIST List 2.94

Sunday, 24 Mar 1991

Disc: French, Parser, Cognitive

Editor for this issue: <>

Directory

1. , Query: objects of adjectives
2. Jonathan Mead, Re: Intro to French Linguistics
3. Mark Johnson, Parsers for the Macintosh?
4. Vicki Fromkin, Re: Responses: Cognitive Linguistics
5. , Modularity
6. , Teenage Mutant Ninja Cognitive Locality Conditions

F. Roger Higgins, in his dissertation, noted that certain adjectives
like 'proud' may not be used with their direct object when the subject
is an inanimate NP like 'Bill's manner':

1a. Bill was proud (of his son).
 b. Bill's manner was proud (*of his son).

2a. Mary was happy (with the results).
 b. Mary's words were happy (*with the results).

3a. Sue was careful (to turn on the lights).
 b. Sue's actions were careful (*to turn on the lights).

Are there any languages in which there is a difference in form between
'proud' predicated of 'Bill' and 'proud' predicated of 'Bill's manner',
and similarly for the other examples? Particularly interesting to me
would be a language in which the two uses of 'proud' were
morphologically distinct, but any information would be gratefully
received, including languages in which the (b) usage (or the (a) usage?)
is impossible.

Additionally, if anyone knows of a discussion of this phenomenon besides
Higgins' brief remarks, I would be interested in that too.

-David Pesetsky
 pesetskathena.mit.edu

Prof. Rooryck mentions a book by Prof. Y. Roberge about French
Linguistics. The book is titled "The Syntactic Recoverability of
Null Arguments" and is published by McGill-Queens University Press
3430 McTavish Street, Montreal, QC. H3A-1X9.

The book is not an introduction to French Linguistics. It focusses
on null arguments, clitic doubling and agreement in a range of
languages (mostly Romance).

Jonathan Mead

I've written a syntactic parser that runs on a MacIntosh. It implements
the Attribute-Value Logic I developed in my thesis (now a book
"Attribute-Value Logic and the Theory of Grammar", CSLI Lecture Notes Series).
It is really meant as a grammar development environment, since it has
fairly extensive graphics facilities (e.g. trees with mouse-sensitive nodes,
attribute-value matrices), and these graphic images can be copied as
standard MacIntosh PICT pictures into your favorite word-processor or
graphics program.

The grammars themselves are essentially what are commonly called
"unification-bases grammars"; Stuart Shieber's little book
"Introduction to Unification-based Approaches to Grammar" gives a
good survey of what such grammars look like. The parser itself comes
with a couple of sample grammars; one which is LFG like (using
attributes that name grammatical relations like SUBJ and OBJ), and
the other is Categorial-Grammar inspired and HPSG-like (with
SUBCAT lists, etc.).

It is written in CommonLisp, and in fact the parser sans graphics will
run on any CommonLisp. You need Apple's MACL if you want to use the
graphics, though. That's a superb implementation of CommonLisp, and
highly recommended for anyone wanting to do computational linguistics
in Lisp on the Mac. (In fact, unless you're after something that
runs on a Sun or similar workstation, I'd strongly recommend MACL on a Mac
over any other Lisp implementation I know of).

I am willing to send the MACL CommonLisp code to anyone interested.
MACL does have the capababilty of generating stand-alone applications,
so you don't have to own MACL to use the parser.
But they are so big that it would be impracticable to send it over
the net. If anyone without access to MACL is interested in getting
a copy of the parser, send me email.

By the way, I must also mention that there are a couple of fairly
good Prolog's for the Mac. Prolog is a 5th generation computer
language, and it has the advantage that relatively sophisticated
parsers can be built from scratch in relatively short time (i.e. a
few hourse when you're good at it). If you need a special-purpose
parser for some special job, it might make sense to roll your own
directly using Prolog. The main drawback is that you don't get the
graphic display capabilities that the parser I described above has:
at least not unless you implement your own graphics package!

The book by Pereira and Shieber "Natural Language Analysis using
Prolog" is a good place to start learning about parsing and Prolog;
I would recommend AAIS Prolog, Quintus Prolog, and ALS Prolog as
prologs for the Mac. These are not as good as the MACL implementation
of Lisp, but quite useable. (If anyone knows of a really good
Prolog for the Mac, please let me know!).

Mark Johnson.,
on leave from Linguistics and Cognitive Science, Brown University.
mjcs.brown.edu (forwarded automatically).

TO: Alexis_Manaster_Ramer
FROM: Vicki Fromkin

E-mail is really addictive and even though one swears under oath to oneself
that one will not respond (or read) stuff which pulls one away from one's
work (the impersonal 'one' is really imnpossible isn't it?) -- one
cannot seem to stay away. So more trivial:

1) I know no linguists who believe they cannot learn from non-linguists
about language or anything else. If it's interesting information you use
it; (2) But in my wanderings, apparently unlike yours, it is the non-linguists
who most often believe they know all about language and have nothing to learn
from linguists. This is beginning to change but it is amazing how people
believe that since they know how to speak they know about language
(3) Remember the many years and millions of dollars when non-linguists
predicted that in 5, then10, then 15 years automatic machine translation
of languages would be accomlished? The notion that all one needed to do
was put in the two dictionaries and feed in a sentence and 'poof' out comes
the sentence in the other language, which in reality Bar Hillel characterized
as the 'Language in,garbage out' MT results. My first linguistics 'workshop'
was the 1962 Venice NATO workshop on the machine translation of languages
and it was amazing how the engineers were sure that linguists who told them
language was more complicated than they thought were all stupid. I still tell
my Ling 1 classes about the famous or infamous machine translation of
"The spirit is willing but the flesh is strong" into Russian as "The vodka
is good but the meat is rotten". When John Pierce (*of Bell Labs, then Caltech)
headed the committee which finally said the predictions were wrong, the
scene shifted to those who were sure they could do automatic speech recognition
in a few years time. Oh well, such are the dreamers.
(4) finally, the study of languages is certainly a legitimate linguistic
pursuit but very different from the study of language; the only theory of
a language is a grammar of that language and unless you believe in the
Joos view that 'languages can differ in innumerable ways' you must believe
that an individual grammar must be based on a theory of grammar which
must account for all and only the grammars of all and only the possible
languages in the world.

I really think I use this 'bulletin board' as a way of postponing the things
I really have to do like income taxes and preparing for my next quarter
course. So forgive a procrastinator and I hope you are stronger than I
and can just delete a file without reading it. A lot of our (my)
discussion is pretty silly. But fun! Vicki Fromkin

Professor Fromkin's arguments for modularity strike me as uncompelling
for a very simple reason. In order to show that some aspects of
linguistic structure are modular, we would, I think, have to show
how this contrasts with what she dismisses as "general cognitive
ability". Or, if "general cognitive ability" is itself modularized,
then again there is nothing special about the allegedly special
linguistic faculties. I think, with all due respect, that the
advocates of modularity confuse very general (and entirely believable)
results that indicate that cognitive abilities are not a single
undifferentiated "cognitive faculty" with highly specific (and
to my knowledge non-existent) arguments for a certain specific
view of the organization of small parts of language (such as syntax)
held by a small minority of linguists and a number of cognitive
scientists. And, of course, the position that IS contradicted by
the data (viz., there IS a single undifferentiated cognitive faculty)
is not what the majority of linguists who dispute (or ignore) the
work on modularity hold. It is just another in a distinguished
line of straw men. And, again with all due respect, it seems to
me that the current attempts to link neurological and psychological
results with linguistic ideas about modularity, are bound for the
same ultimate destination as the once-celebrated arguments for the
reality of transformational grammar in real-time psychological
processes, or those for the reality of SPE-style underlying
representations in phonology, etc. If the neurological and
psychological data turn out not to support the linguist theories,
as I conjecture, it will not be the theories that will be given up.
Rather, we will suddenly discover that such "external" data are
not as relevant to linguistics as they are now claimed to be. And,
if, as is inevitable, the linguistic theories are themselves replaced
by other, more fashionable, ones (for purely linguistic reasons!),
then, the now-heralded neurological and psychological results will,
again, be forgotten. 

Alexis Manaster-Ramer writes in Linguist 2.90:

>It strikes [me] as bizarre that someone would insist that functionalists,
>or cognitive linguists, or whoever, should have to explain locality
>principles before they can be taken seriously. 

Fortunately, no one has made this bizarre assertion here (just as no one
made the assertions concerning non-linguists that you attributed to
Poser).

I do think that "functionalists or cognitive linguists, or whoever,
should have to explain locality principles before they can be taken
seriously" as explainers of locality principles. Ditto for any other
topic, from discourse to phonetics. Otherwise, we're in the kingdom of
vaporware.

-David Pesetsky