Copy Link
Add to Bookmark
Report
NL-KR Digest Volume 02 No. 48
NL-KR Digest (6/02/87 00:13:19) Volume 2 Number 48
Today's Topics:
Are syntacticians doing the right things?
S level grammar
Re: NL vs. AL
Re: Natural Languages: They're about the World
clarification on 'unicorn' semantics
The Generic Horse
Brentano on Intentionality
Text Critiquing and Eliza
----------------------------------------------------------------------
Date: Fri, 29 May 87 08:12 EDT
From: David Pesetsky <PESETSKY%cs.umass.edu@RELAY.CS.NET>
Subject: Are syntacticians doing the right things?
In NL-KR 2.44, Bruce Nevin <bnevin@cch.bbn.com> replies to my
contribution to the debate on current work in syntax. He writes:
>There is a deeper problem here, and that has to do with the
>relatively recent practice (only about 30 years old) of relying
>on anecdotal evidence in linguistics. THERE IS NO COMPLETE
>GENERATIVE GRAMMAR OF ANY LANGUAGE. One has only fragments of
>grammars. The occasional attempts at a grand synthesis (e.g.
>Stockwell, Schachter, and Partee's _The Major Syntactic
>Structures of English) have all failed.
> This is the reason I emphasize the need for a comprehensive
>approach, aiming for complete coverage from the outset.
What is meant by "relying on anecdotal evidence"? Judging from
the rest of the paragraph, this means using as subject matter
paradigms DRAWN FROM the syntactic data displayed by the
language, rather than some "complete" corpus of syntactic data.
If one were to accept this as a valid methodological criticism of
linguistics, then linguistic research would have to stop. (1) We
have no "complete" corpus, of course, for principled reasons.
(2) While we aim at the deepest explanations we can come with,
staring at 3000 sentences does not -- or has not -- revealed in a
flash to anyone the nature of the grammar as a whole. Instead,
as in any science, the ocean of accessible data is combed (mixed
metaphor) for sets of facts that seem to make sense together.
Painstakingly -- over many years -- we struggle so that our
little islands of understanding can grow larger and coalesce.
If we were to take the statement "THERE IS NO COMPLETE GENERATIVE
GRAMMAR OF ANY LANGUAGE" to be a sign of trouble, we might as
well take statements like "THERE IS NO COMPLETE PHYSIOLOGY OF THE
HUMAN BEING" or "THERE IS NO COMPLETE GEOLOGY OF THE EARTH" to be
indications of serious trouble in biology or geology.
In calling for a "complete generative grammar", Nevin is stating
the goals of the field -- where by "complete generative grammar"
we mean a full description of the contribution of the language
faculty (Universal Grammar) and of the parameters/rules supplied
by experience to the syntactic competence of the adult. In
correctly noting the absence of such a "complete generative
grammar", Nevin is simply noting that there is work to be done.
In noting the "failure" of Stockewell, Schachter and Partee,
Nevin is merely noting that there was also work to be done in the
early 70's, when that book was written. However, if Nevin
believes that this work will be done more quickly by taking a
"comprehensive" (i.e. non-"anecdotal") approach, just try it!
Sure, if you just want a description or catalog of sentence
types, a half-century or two of work might do it (cf. as a
foundation the better descriptive grammars, such as Quirk et al.,
or the wonderful work of Jespersen).
If, however, you are really interested in understanding WHY
languages show the organizational patterns they do, in
understanding HOW languages can differ, in understanding WHAT
facts about a particular language might be linked together
(aiding the child in the acquisition of that language), the
difficulties in dealing even with some small fragment of the
syntax are profound -- and the difficulties in dealing with some
larger fragment of the syntax are at present insuperable.
Linguistics is hard. In choosing some fragment to explain, the
linguist surely runs the risk of picking the wrong fragment (a
non-natural class of phenomena, if you will), but that is the
risk in any field of inquiry and is the reason why science so
often marches slow.
>Attempts to generalize from a seemingly successful fragment have
>not worked.
I do not know what is being referred to here; but if many or most
attempts at "generalizing" a theory of some set of data to a larger
set of data fail, I am not surprised or too disappointed. You have
to expect this: I repeat, linguistics is hard -- at least for me.
>Re the linguistic phenomena that Rizzi discusses, I recommend
>studying carefully section 3.1, `Linearizations and
>Transpositions', in Harris, _A Grammar of English on
>Mathematical Principles.
Summer reading.
------------------------------
Date: Mon, 18 May 87 00:02 EDT
From: GHERMON@UIUCVMD
Subject: S level grammar
I was very glad to see Pesetsky's note on the issue of studying S grammar. I
was also appalled (but too busy to answer) at the the level of discussion so
far. How can people argue about S grammar and discourse while disregarding
most of the work in linguistics in the last 30 years?? And for those who
doubt whether current research on S grammar has anything to do with the way
children acquire language- read any of the articles in B. Lust "The
Acquisition of Anaphora" or in Roeper and Williams "Parameter Setting" (both
volumes are published by Reidel, 1987). Or even better, read Nina Hyams' study
of how children deal with the range of facts connected to the pro-drop
parameter (also published by Reidel in 1987).
------------------------------
Date: Sun, 31 May 87 04:30 EDT
From: Steve Bloch <ma188saa@sdcc3.ucsd.EDU>
Subject: Re: NL vs. AL
Michael T. Gately <gately%resbld%ti-csl.csnet@RELAY.CS.NET> writes:
>I was thinking about the 'natural' vs. 'artificial' language discussion
>and beginning to believe that natural languages could be characterized
>by their use by people in communicating mental processes to each other.
>
>I recall being told a story about some LISP Machine programmers at MIT who
>used to communicate in LISP. For example, a programmer might send a
>message to another programmer reading "LUNCH-P."
>
>[One can argue that such use of Lisp, or whatever, is still NL use:
>A lot of inference of the intentions of the `speaker' by the
>`hearer' must go on for the query to make semantic/pragmatic
>sense, most of which would be picked up from context. Such
>inference does not go on in (normal) programming languages...BWM]
I've certainly used "lunchp" and variants in that way; similarly the "begin
flame ... end flame" often seen on the net. These are obviously not true uses
of computer language (henceforth CL) for human-human communication, but simply
NL showing off its ability to absorb new vocabulary. HOWEVER... Imagine two
scientists debating whether there's an effective method (in the technical
sense) to construct something that they dearly wish to know is constructible.
One day one of them sends the other email, or scrawls on his office door, an
algorithm (in Lisp, or Lisp-looking pseudocode, of course, they being
intelligent scientists :-) that she thinks will do it. Even if it's
pseudocode, the communication is a lot closer to CL than to NL. In
particular, it avoids the criticism (made about both of the previous examples)
that the CL is being applied to subjects outside its intended domain of
discourse. On the other hand, the CL communication is not just Exhibit A (and
thus an OBJECT of communication rather than a communication itself) because
there may be no NL wrapped around it, so although some context is required
(i.e. "I remember we were talking about this problem"), most of the meaning
is textually present in the message itself. The message's function is to
convey (a) the information "yes, there's an effective method," and (b) an
argument to hopefully convince the recipient of the veracity of (a). This
makes it human-human communication through CL.
------------------------------
Date: Sun, 31 May 87 03:29 EDT
From: Steve Bloch <ma188saa@sdcc3.ucsd.EDU>
Subject: Re: Natural Languages: They're about the World
In article <4568@milano.UUCP> wex@milano.UUCP writes:
>I agree that there are certain classes of thoughts I have (call them I)
>which are about what might be called "imaginary" things. These are thoughts
>about courage or God or unicorns, etc. These thoughts are distinct from
>others in that I recognize a priori that there is no physical manfiestation
>of the thing I am thinking of. Therefore, it is reasonable to postulate a
>mental object which is the target of I-class thoughts.
>
>However, there is another class of thoughts (call them R) which are about
>what might be called "real" things. These are thoughts about, for example,
>the chair I sit in, the keyboard I type at, or that `unicorn' that Ringling
>Brothers have in their circus. These R-class thoughts are distinct in that
>I recognize a priori that there is a physical manifestation of the thing I
>am thinking of. Even in cases where I am mistaken (the object is not what I
>thought it was or it fails to exist altogether), I still think that the
>intentionality of R-class thoughts is a real thing.
I object to equating "real" with "physical". Yes, that's an ancient
metaphysical problem, but here where we're only concerned with the
subjects of thoughts it's more tractable. Of course, to a confirmed
Platonist, thoughts about courage or God or unicorns ARE thoughts
about a real, existing thing which just doesn't happen to have a
physical manifestation. But you don't have to be a Platonist to
have trouble with this: when I'm searching for a bug in a program,
or trying to figure out the distinction between Artinian and
Noetherian rings, those thoughts seem very similar to my thoughts on
trying to fit a couch through the front door. As Laura said, "My
thoughts are my thoughts."
Would someone be so kind as to post the original definition of
"intentionality", by Brandon or whatever the name was?
------------------------------
Date: Fri, 29 May 87 00:41 EDT
From: STANKULI%cs.umass.edu@RELAY.CS.NET
Subject: clarification on 'unicorn' semantics
point of order:
NL-KR Digest Volume 2 No. 42
aweinste@Diamond.BBN.COM
> In typical usage, the word "unicorn" typically doesn't refer to
> anything. This is only a puzzle if you suppose that every word or
> thought-element must have a referent in order to be meaningful, but
> philosophers have understood since Russell that this is false.
it has been a while since i studied formal semantics so perhaps someone
will refresh me on the history of possible-worlds semantics. as i recall,
Wittgenstein first introduced the informal notion that concepts were the
meanings of words. Bertrand Russell formalized that concepts were sets,
and that 'unicorn' had the extension of the null set.
who was it that pointed out that 'john is searching for a unicorn' would
have the same extension as 'john is searching for a centaur' and hence the
same meaning? was it Carnap? this set the stage for formal possible-worlds
semantics, but i thought that Russell was committed to extensional semantics
in which knowledge acquisiton becomes so prohibitive. it seems that our
present artificial intelligence is struggling with these same issues of
acquiring knowledge in single domains because of the exponential explosion
of possible-world model domains. isn't that why constraints on model
specification are so important in research?
stan
------------------------------
Date: Fri, 29 May 87 13:24 EDT
From: Ken Laws <LAWS@IU.AI.SRI.COM>
Subject: The Generic Horse
From: wex@milano.UUCP
Subject: Re: Natural Languages: They're about the World
If you "think about a horse" you are visualizing one. The intentionality of
your thought is directed at that visualization, not at the class of all
horses. However, it is possible to have thoughts about the class of all
horses, in which case the class-of-all-horses is the intention of your
thought.
My own subjective experience seems at odds with this. I gather from
what I read of the mental image literature that most people visualize
an object when they think of it. (The mental image debate is over
whether this visualization is used as a substrate for computation or
is merely epiphenomenal -- a unused artifact of our perceptual
computations.) I am also told that the the subjects chosen for mental
image research are those who do well at forming mental images in
screening tests.
I am not able to form mental images, except in a very vague and
unreliable manner. Perhaps the images are formed and I do not have
conscious access to them, so believe what I say at your own risk.
Still, I seem to have the opposite of a photographic memory. I can
recognize visual objects, but I have extreme difficulty forming an
image of my wife, my kids, my house, etc. On the other hand, I have
no difficulty imaging a horse -- a specific, generic horse -- without
visualizing it. I do have some vague image of a body and four legs
(which seem to be moving), but it has no particular color, markings,
or even definite shape. (I admit that it is a standing horse rather
than a rearing or jumping horse, so the image may well be there.)
My own interpretation is one of spreading activation. When I see a
particular horse, many neural circuits are activated (or many
propositions are asserted). Different horses activate somewhat
different clusters of circuits. The generic horse is simply the
activation of some of these clusters. People who visualize strongly
apparently employ enough feedback or constraint propagation to
lock onto a cluster representing a specific, detailed horse -- even
if the details match no horse they have ever encountered. Others of
us probably accept either the broad pattern of activation as a
representation of "horseness" or apply so much inhibitory feedback
that only the central core of "horseness" is left activated (e.g.,
four-legs-and-rideable). I would expect different kinds of creativity
from individuals with these different internal representations, so there may
be advantages in maintaining such diversity in the human genetic stock.
I have heard that the oriental philosophers (of the Classical Chinese
period) did not have the "Western" notions of mentalese representation.
Whatever process caused a man viewing a horse to say "horse" was the
same process that caused a man viewing the [variable] character for horse
to say "horse", and the [written or unwritten] character filled the
role that we ascribe to mental objects. (This is integrated with
other aspects of Oriental thought, such as the propensity to use
language as a guide or "program" for correct behavior rather than
a description of the world or a "true" abstraction of the world.
It's beyond my current comprehension, but apparently a consistent
world view is possible without the need for reasoning about beliefs,
desires, and other Western preoccupations. The ideas are due to
Dr. Chad Hansen, Prof. of Philosophy, U. Vermont.) Perhaps the Chinese
philosophers were on the right track. Perhaps there are no mental
objects and no levels of recursive processing of such objects.
-- Ken Laws
------------------------------
Date: Mon, 1 Jun 87 12:03 EDT
From: Michael P. Smith <mps@duke.cs.duke.edu>
Subject: Brentano on Intentionality
In <399Asdcc3.ucsd.EDU> Steve Bloch asks:
>Would someone be so kind as to post the original definition of
>"intentionality", by Brandon or whatever the name was?
From Franz Brentano, PSYCHOLOGIE VOM EMPIRISCHEN STANDPUNKT (1874):
Every mental phenomenon is characterized by what the Scholastics of
the Middle Ages called the intentional (or mental) inexistence of an
object, and what we might call, though not wholly unambiguously,
reference to a content, direction toward an object (which is not to be
understood here as meaning a thing), or immanent objectivity. Every
mental phenomena includes something as object within itself, although
they do not all do so in the same way. ...
This intentional in-existence is characteristic exclusively of mental
phenomena. No physical phenomena exhibits anything like it. We can,
therefore, define mental phenomena by saying that they are those
phenomena which contain an object intentionally within themselves.
(pp.88 -- 89 of translation by McAlister, et. al., of
Kraus's 1924 edition, RKP, 1973)
There is much good material in Ausonio Marras, ed. INTENTIONALITY,
MIND AND LANGUAGE, U. of Ill. Press, 1972.
------------------------------
Date: Thu, 28 May 87 22:29 EDT
From: Ethan Scarl <ethan@BOEING.COM>
Subject: Text Critiquing and Eliza
[Excerpted from AIList]
The "grammar checker" discussions were stirring some old memories which I
finally pinpointed: a 1973 debate (centered on Joe Weizenbaum and Ken
Colby) over whether Eliza should be used for actual therapy.
The heart of the grammar checker issue is whether a computational package of
limited or doubtful competence should be given an authoritative role for some
vulnerable part of our population (young students, or confused adults). What
was most shocking in the Eliza situation (and may be true here as well) was
the quick and profound acceptance of a mechanical confidante by naive users.
Competent and experienced writers have no trouble discarding (or
extrapolating from) Rightwriter's sillier outputs; the problem is with
inexperienced or disadvanted users. Many of us were (are) enraged at this
automated abuse as absurd, irresponsible, and even inhuman," only to be
stopped short by a sobering argument: "if competent human help is scarce,
then isn't this better than nothing?"
The Rightwriter discussion summarizes/coheres rather well: Such systems are
suggestive aids for competent writers and may be useful in tutoring the
incompetent. Such systems will be unsuitable as replacement tutors for some
time to come, but may be worthwhile (in time and effort expended for results
achieved) as aids to be be used by a competent tutor or under the tutor's
supervision.
We are in deep trouble if there are no competent humans available to help
others who need it. But the secondary question: "Is sitting in front of a
CRT better than sitting in a closet?" can at least be tested empirically.
In the Rightwriter case, I would expect that most students will quickly
understand the the program's analytic limitations after they are pointed out
by a teacher. However, the human teacher's perspective is essential.
------------------------------
End of NL-KR Digest
*******************
