[Prev][Next][Index]
CSLI Calendar, 25 January, vol. 5:14
-
Subject: CSLI Calendar, 25 January, vol. 5:14
-
From: csli@csli.stanford.edu
-
Date: Wed 24 Jan 1990 15:49:06
C S L I C A L E N D A R O F P U B L I C E V E N T S
-----------------------------------------------------------------------------
25 January 1990 Stanford Vol. 5, No. 14
_____________________________________________________________________________
A weekly publication of the Center for the Study of Language and
Information (CSLI), Ventura Hall, Stanford University, Stanford, CA 94305-4115
____________
CSLI ACTIVITIES FOR THIS THURSDAY, 25 JANUARY 1990
12:00 noon TINLunch
Cordura 100 The Role of Central Conceptual Structures in the
Development of Scientific and Mathematical Thought
Robbie Case
School of Education
Stanford University
(ka.rob@forsythe.stanford.edu)
Abstract in last week's Calendar
2:15 p.m.
Cordura 100 CSLI Seminar
HPSG from Afar
Paul John King
CSLI Postdoctoral Fellow
(pjking@csli.stanford.edu)
Abstract in last week's Calendar
____________
CSLI ACTIVITIES FOR NEXT THURSDAY, 1 FEBRUARY 1990
12:00 noon TINLunch
Cordura 100 Reading: A Logical Model of Machine Learning:
A Study of Vague Predicates
by Wlodek Zadrozny and Mieczyslaw Kokar
Discussion led by Jerry Hobbs
(hobbs@ai.sri.com)
Abstract below
2:15 p.m. CSLI Seminar
Cordura 100 Some Aspects of Word-Retrieval Errors in the
Speech of Aphasic Adults
Audrey Holland
Professor of Otolaryngology and Communication
and Associate Professor of Psychiatry
Director of the Division of Speech, Language,
and Voice Pathology, School of Medicine
University of Pittsburgh
(alh@med.pitt.edu)
Abstract below
____________
NEXT WEEK'S TINLUNCH
A Logical Model of Machine Learning:
A Study of Vague Predicates
by Wlodek Zadrozny and Mieczyslaw Kokar
Discussion led by Jerry Hobbs
In this paper, we apply a logical framework to the problem of
recognizing vague predicates. We formulate a rule of abduction and
apply it in identifying objects; we formally account for the context
sensitivity of recognition observed by W. Labov. We show how multiple
theories of "cup" can be combined in a new theory of the concept.
Finally, we conjecture that operationality in explanation-based
learning is related to the cost of abduction.
The logical theory we use assumes that reasoning and learning take
place in an interaction of theories on three levels: methodological
level, object level, and referential level. An object level theory
describes the current situation; the referential level encodes
background knowledge; while the methodological level is responsible
for choosing methods of reasoning and ways of constructing models.
____________
NEXT WEEK'S CSLI SEMINAR
Some Aspects of Word-Retrieval Errors in the
Speech of Aphasic Adults
Audrey Holland
The talk will include a brief sample of a videotaped interaction with
an aphasic man who has a moderate conduction aphasia, frequent
phonemic paraphasic errors, and word-retrieval deficits. How the
patient's speech production deficits are modified as a result of the
feedback with which he is provided, how such deficits and their
resolutions are classified and coded, and some difficulties in
description of word-retrieval deficits will be the focus of the talk.
I will try to relate some of the realities of speech production
deficits to problems of building computer models of aphasia, at least
briefly.
____________
SEMINAR ON ISSUES IN LOGICAL THEORY
Philosophy 396
Approaches to the Liar Paradox, Part II
John Etchemendy
(etch@csli.stanford.edu)
Thursday, 25 January, 3:45-5:30
Cordura 100
In the logic seminar this week, John Etchemendy will talk about the
treatment of the liar paradox in Barwise and his book {\it The Liar}.
Next week, Bernie Linsky will present material from "General
Intensional Logic" by C. Anthony Anderson (chapter II.7 of the
_Handbook of Philosophical Logic_). Copies will be made available at
this week's meeting.
____________
SYMBOLIC SYSTEMS FORUM
Conversations with Apes: Reflections on the
Scientific Study of Language
John Dupre
Department of Philosophy
(dupre@csli.stanford.edu)
Thursday, 25 January, 4:15
Building 60, Room 62G
(Please note room change!)
In this talk, John Dupre will look at some of the different attempts
that have been made to teach symbolic systems, including American sign
language, to various apes. He'll also discuss the main criticisms
that have been directed against these attempts. He will argue that
these criticisms reveal serious conflicts between assumptions about
correct scientific methodology and the very possibility of the kind of
research project envisaged by the ape language researchers. Finally,
he will offer a few suggestions about why this research and its
evaluation has seemed so important to some people.
At the next Forum (1 February), Annie Zaenen will talk. Title: Verb
Varieties: Syntax or Semantics?
____________
LINGUISTICS DEPARTMENT COLLOQUIUM
Dependency Relations and Syntactic Functions:
Heads and Bases
Arnold M. Zwicky
Ohio State University and Stanford University
Friday, 26 January, 3:30
Cordura 100
A particularly simple approach to syntactic functions/relations would
treat them as located within a hierarchy (or perhaps within several
overlapping hierarchies) arising from the split between the dependency
relations HEAD-OF and DEPENDENT-OF, from the split of DEPENDENT-OF
into ARGUMENT-OF and MODIFIER-OF (and the corresponding split of
HEAD-OF into the converse relations OPERATOR-ON and MODIFIED-BY,
respectively), and from the split of ARGUMENT-OF into syntactic
functions like Subject and Direct-Object and of MODIFIER-OF into
syntactic functions like Adjectival and Adverbial.
There are well-known problems with one part of this view, involving
arguments that clearly have the syntactic functions (and sometimes the
syntactic categories as well) normally associated with modifiers: for
example, the locational dependent of the verb PUT in I PUT THE BOX ON
THE TABLE and the manner dependent of the verb WORD in WE WORDED OUR
RESPONSE CAUTIOUSLY. Phenomena like these argue for a dissociation of
the dependency relations from syntactic functions; an Adverbial is
standardly a MODIFIER-OF a nonnominal head, and the standard
non-Subject ARGUMENT-OF a verb head is a Direct-Object or an
Indirect-Object NP, but these are merely the default alignments, not
invariable associations. On the one hand, then, we have dependency
relations like ARGUMENT-OF and MODIFIER-OF, while on the other hand we
have an inventory (perhaps very large, and perhaps very complexly
organized) of syntactic functions like Subject, Adverbial, Predicator,
and so on, each with its own default associations with syntactic
categories (NP for Subject, AdvP and PP for Adverbial, V for
Predicator, and so on).
Granting dissociations that will allow (for instance) an Adverbial PP
to serve as an ARGUMENT-OF a Predicator V, we should also expect
nonstandard pairings of OPERATOR-ON and MODIFIER-OF with syntactic
functions. I propose that one of these dissociations is in fact
abundantly exemplified, when an Adverbial or Adjectival constituent C1
serves as OPERATOR-ON another constituent C2. In such a construct C0,
we would expect C1 to exhibit some of the properties of the "head"
within C0; as an OPERATOR-ON C2, C1 should be expected (following
Keenan-style generalizations) to act as agreement target with respect
to the agreement trigger C2, and as government trigger with respect to
the government target C2, and it should be expected (following the
default inheritance generalizations in most current approaches to
phrase structure) to be the locus of morphosyntactic marking for
properties belonging to C0 as a whole (unless the rule for this
construction specifically stipulates that these are marked elsewhere,
for instance on an edge of C0). But we should also expect C2 to
exhibit some of the properties of the "head" of C0; as an Adverbial or
Adjectival, C1 should be expected to be optional, and C2 should be
expected to be the "syntactic determinant," the constituent that
predicts the external distribution type of C0.
This is the split of properties that often occurs for combinations of
an auxiliary verb with its complement; for combinations of a
complementizer with a clause; for combinations of a "grammatically
used" adposition with its object NP; and for combinations of a
determiner with a nominal constituent. The syntactic literature on
these matters -- as in the _Journal of Linguistics_ exchange between
Zwicky and Hudson, or in Abney's dissertation -- is largely taken up
with attempting to argue that one or the other of the two constituents
involved *really* is the head, though there are those (Warner in
Linguistics, Fenchel in WCCFL) who propose to side-step the problem by
declaring *both* constituents to be heads. I maintain that there is
no problem, since there are two conceptually distinct entities here,
for one of which I reserve the label "head," for the other of which I
suggest the label "base."
This is not yet a theoretical proposal. Rather, I am elucidating a
distinction that I believe will have a reflex in any adequate theory
of syntax, without at the moment taking a stand on how the distinction
should be realized in a particular theory.
____________
COMMONSENSE AND NONMONOTONIC REASONING SEMINAR
Formalizing Various Intuitions About Inheritance
in Logic Programs
Fangzhen Lin
Stanford University
Monday, 29 January, 2:30
Margaret Jacks Hall 301
(Please note room change for this meeting only!)
Reasoning about inheritance is one of the earliest applications of
nonmonotonic logics. It is also one of the motivations for developing
such logics. Unfortunately, so far attempts at formalizing
inheritance hierarchies using general purpose nonmonotonic logics,
like default logic and circumscription, seem not as successful as the
ones using ad hoc methods, like the ones used by Touretzky and the
Horty trio. This raises an important question: Are these nonmonotonic
logics appropriate for the job? In this paper, we'll show that for
default and autoepistemic logics, the answer is positive.
Specifically, we'll propose a methodology for formalizing various
intuitions about inheritance in logic programs with negation as
failure (a subclass of default and autoepistemic theories). We'll
prove that one of our formalizations includes Horty's skeptical theory
as a special case. Among other things, the methodology is remarkably
simple and very similar to the ones used by McCarthy and others.
____________
SEMINAR ON COMPUTERS, DESIGN, AND WORK
Social Aspects of Expert Systems
Bryan Pfaffenberger
Wednesday, 31 January, 12:15
Ventura 17
No abstract available.
____________
NEW CSLI POSTDOCTORAL FELLOW
Paul John King
(pjking@csli.stanford.edu)
Manchester University, England
Dates of visit: January 1990-January 1991
Graduating from London University with a mathematics degree, Paul
studied for a mathematical logic Ph.D. under Peter Aczel at Manchester
University. Drawn to him for his work on nonwellfounded sets, Aczel
soon introduced him to Barwise and Etchemendy's _The Liar_, helping
him to write a dissertation that simplified some of the maths behind
their work. Fueled by this and other dealings with situation
semantics, Aczel encouraged Paul to learn a little about linguistics.
This culminated in his Ph.D. dissertation, which dealt with the
mathematical foundations of Pollard and Sag's _Information-based
Syntax and Semantics_. Having, seemingly, developed a habit for
putting mathematical "meat" on CSLI "bones," Paul hopes to deepen this
habit while here by working alongside the situationists and/or
unification-based grammarians (whoever wants him most!) as a sort of
mathematical "janitor"!
____________
NEW CSLI VISITOR
Hiroshi Nakagawa
(nakagawa@csli.stanford.edu)
Professor of Electrical and Computer Engineering
Yokohama National University, Japan
Dates of visit: January 1990-January 1991
Hiroshi is interested in commonsense-knowledge representation and
situation theory. While at CSLI, he would like to represent
commonsense knowledge using situation theory. In trying to do this,
he also hopes to find out which parts of situation theory are useful
and/or what kinds of logical devices we should add to situation
theory. He thinks there are at least two ways in situation theory
that will allow him to do this. One is the "situation first" way,
i.e., defining operations on situations or situation types. The other
is the "infon first" way, i.e., defining operations on infons or
so-called infon algebra. Hiroshi feels he is now wandering between
these two ways.
____________