[Prev][Next][Index]
CSLI Calendar, 26 April 1990, vol. 5:25
-
Subject: CSLI Calendar, 26 April 1990, vol. 5:25
-
From: csli@csli.stanford.edu
-
Date: Wed 25 Apr 1990 15:25:20
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
-----------------------------------------------------------------------------
26 April 1990 Stanford Vol. 5, No. 25
_____________________________________________________________________________
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, 26 APRIL 1990
12:00 noon TINLunch
Cordura 100 Truth Makers for Modal Propositions
Bernard Linsky
Visiting Scholar from the University of Alberta
(linsky@csli.stanford.edu)
Abstract in last week's Calendar
2:15 p.m. CSLI Seminar
Cordura 100 Controversies in Natural-Language Research 2
led by Stanley Peters
(peters@csli.stanford.edu)
Title: Developments in Phonological and
Morphological Theory
Speakers: Paul Kiparsky and Bill Poser
(kiparsky@csli.stanford.edu, poser@csli.stanford.edu)
Abstract in last week's Calendar
____________
CSLI ACTIVITIES FOR NEXT THURSDAY, 3 MAY 1990
12:00 noon TINLunch
Cordura 100 Fuzzy Logic and its Applications
Lotfi A. Zadeh
Computer Science Division
University of California, Berkeley
(zadeh@ucbarpa.berkeley.edu)
Abstract below
2:15 p.m. CSLI Seminar
Cordura 100 Controversies in Natural-Language Research 3
led by Stanley Peters
(peters@csli.stanford.edu)
Abstract below
4:15 p.m. CSLI Colloquium
Cordura 100 Metaphor Comprehension by Neural Networks
Kouichi Doi
Department of Electrical Engineering
University of Tokyo
(doi@mtl.t.u-tokyo.ac.jp)
Abstract below
____________
NEXT WEEK'S TINLUNCH
Fuzzy Logic and its Applications
Lotfi A. Zadeh
Fuzzy logic has been, and continues to be, controversial. Regardless
of the controversy, it is finding many applications in fields ranging
from medical expert systems and image analysis to the control of
cement kilns and subway trains. In Japan, in particular, it is widely
used in industry and is the object of a large-scale
government-supported program to expand its applications.
A concept that plays a key role in many of the applications of fuzzy
logic is that of a linguistic variable, that is, a variable whose
values are words or sentences in a natural or synthetic language. To
illustrate, _age_ is a linguistic variable of its values labeled
_young_, _very young_, _quite young_, _not very old_, etc. In
general, a linguistic value may be viewed as a label of a possibility
distribution. What is important about such values is that the
meaning, i.e., the possibility distribution, of any linguistic value
may be computed from the meaning of the so-called primary terms, e.g.,
_young_ and its antonym, _old_.
The concept of a linguistic variable provides a computational
framework for describing the relation between two or more variables in
qualitative, linguistic terms. It is this framework that plays a
major role in many of the applications of fuzzy logic. In the final
analysis, what fuzzy logic offers is a much better model for human
reasoning -- most of which is fuzzy rather than precise -- than the
traditional logical systems, which are intolerant of imprecision and
partial truth.
____________
NEXT WEEK'S CSLI SEMINAR
Controversies in Natural-Language Research 3
led by Stanley Peters
Next week's seminar will consist of a panel discussion/debate among
Ron Kaplan, Lauri Karttunen, Martin Kay, Paul Kiparsky, and Bill Poser
about the computational tractability of morphological theory as
presented two weeks ago and the theoretical underpinnings of the
computational approach to morphology presented this week.
____________
NEXT WEEK'S CSLI COLLOQUIUM
Metaphor Comprehension by Neural Networks
Kouichi Doi
In this paper, we propose a method of metaphor comprehension by neural
networks. One metaphor has many meanings, and an associative network
is useful to find candidates as the real meaning of the metaphor.
In our method, input sentences made of characters are decomposed into
independent words, and are given a neural network. Then the neural
network is constructed in the following way: situation and context are
treated as the activation values of neurons. Synonyms are treated as
the assignment of each neuron. When meanings of words alter, we treat
such cases by changing connective coefficients between neurons.
Candidates of interpretation are output from the neural network.
Among the candidates, a valid meaning is chosen based on the mixture
theory. Our network has a simpler structure, and the network
converges faster than ever. The result of the experiment by network
is valid.
____________
SYMBOLIC SYSTEMS FORUM
Applications of Nonwellfounded Sets in Symbolic Systems
Jon Barwise
(barwise@csli.stanford.edu)
Thursday, 26 April, 4:15 p.m.
Building 60, Room 61G
Nonwellfounded sets have been in ill repute for seventy-odd years.
The reason is that they were mistakenly implicated in the paradoxes of
set theory. In recent years, though, they have been rehabilitated.
This rehabilitation has been brought about in part by theoretical work
in set theory, where Aczel and others have shown how to make sense of
such sets, and in part because of the applications that are being
found for them in the symbolic sciences. In this elementary talk, I
will motivate them by means of example, and discuss the theory of
nonwellfounded sets a bit.
Next week: The Neurophysiology of Conscious Experience, Ben Libet,
Physiology, University of California, San Francisco.
Change in calendar: The announced dates of the talks by Professors
Girard and Dupuy in May have been switched. Professor Dupuy will be
speaking on 17 May (On the Self-Deconstruction of the Symbolic Order),
and Professor Girard will give his talk on 24 May (The Genesis of
Symbolic Forms in Human Culture).
____________
LINGUISTICS DEPARTMENT COLLOQUIUM
Ditransitives and Linking Theory
Steve Wechsler
(wechsler@csli.stanford.edu)
Friday, 27 April, 3:30 p.m.
Cordura 100
The goal of linking theory is to explain the "projection from the
lexical entries of predicators to the morphosyntax of the clauses or
phrases that those predicators head. A reasonable desideratum for
such a theory is that each semantic feature used in the grammar must
either (i) be definable in terms of universal semantic primitives
(which perhaps should be expected to meet a condition of
epistemological priority (Chomsky 1981)); or (ii) be associated with a
specific linguistic form. I will propose a variant of Direct Linking
Theory (Kiparsky 1989) that meets this criterion. Thematic role types
are fine-grained relations associated with linguistic forms --
typically Ps and semantic cases, sometimes special agreement morphemes
and configurational linkers -- so they are of type (ii).
Crosslinguistic principles of term (subject and object) linking are
stated in semantic features that meet (i). Since type (i) and (ii)
features are theoretically distinct, hierarchies like "agent -->
recipient --> theme" are convenient descriptive devices but not part
of grammar.
This theory will be applied to the analysis of ditransitives. First I
will give an account of the English dative/benefactive construction,
which captures its productivity in terms of a configurational "dative"
linker, without positing a derivation or operation at any level of
grammar. Then I will turn to Swedish, which has not only the dative
linker but incorporated Ps as well:
(1a) Man tog chefskapet fraon honom.
one took the-headship from him
(1b) Man fraon-tog honom chefskapet.
one from-took him the-headship
"They took the headship from him."
On the proposed theory the incorporated P, like the unincorporated
one, is a semantic linker (Ostler 1979) or equivalently a copredicator
(Gawron 1986), which is thus available to link a postverbal NP bearing
the appropriate role. This correctly predicts that (1b) has two
passives, "The headship was from-taken him" and "He was from-taken the
headship." Finally, I will look at the implications of this theory
for the analysis of applicatives and "relational preverbs" (Craig and
Hale 1988).
____________
COMMONSENSE AND NONMONOTONIC REASONING SEMINAR
The Frame Problem:
From Toy Worlds to a General Theory
Vladimir Lifschitz
Department of Computer Science
Stanford University
(val@sail.stanford.edu)
Monday, 30 April, 2:30 p.m.
Margaret Jacks Hall 252
Ideas related to formalizing action and change are usually tested on
"toy domains," such as the blocks world or the "Yale shooting"
example. We will discuss what appears to be the most promising
available formalization of the blocks world and investigate how
general it is. We will see that that solution is applicable to a
whole class of domains involving situations and actions, of which the
blocks world is only one example. The class is described in terms of
purely formal, mostly syntactic, properties of axioms.
____________
POETICS WORKSHOP
A Coherence Analysis of a Sonnet by Milton
Jerry Hobbs
SRI International
(hobbs@ai.sri.com)
Tuesday, 1 May, 4:00 p.m.
Ventura 17
The poetics workshop will resume meeting this quarter, after its
winter quarter hibernation. The workshop involves people from various
departments with a common interest in such formal properties of poetry
as meter, rhyme, alliteration, and parallelism, and their relation to
syntactic and phonological structures of language. It meets every
other week for informal presentation and discussion of current
research.
____________
PHILOSOPHY DEPARTMENT COLLOQUIUM
Conceptual Issues in the Divided Mutagenesis
of Bacteria Controversy
Sahotra Sarkar
Boston University
Tuesday, 1 May, 4:15 p.m
Building 380, Room 380F
(Note unusual day, place, and time)
No abstract available.
____________
SYNTAX WORKSHOP
What's in a Word? On the Syntax of Bantu Noun Class Prefixes
Joan Bresnan and Sam A. Mchombo
(bresnan@csli.stanford.edu, mchombo@csli.stanford.edu)
Tuesday, 1 May, 7:30 p.m.
Cordura 100
In virtually the entire tradition of Bantu linguistics, the gender or
class markers of Bantu nouns are analyzed as prefixes morphologically
bound to noun stems. Myers (1987) proposes an alternative syntactic
analysis, according to which the class markers are syntactically
independent words (like articles or particles) that are associated
with their noun stems only phonologically and not morphologically.
The syntactic analysis of the noun class markers raises the question
of what criteria to use in identifying words. The Principle of
Lexical Integrity, which states that the internal structure of words
is opaque to syntax, provides a rich set of criteria, which Myers
(1987) has not exploited. According to these criteria, neither the
morphological analysis nor the syntactic analysis of class markers is
completely correct by itself. Instead, there is compelling evidence
from Chichewa for the syntactic analysis of the locative class
markers, and for the morphological analysis of all other noun class
markers.
We compare two alternative explanations for this state of
affairs -- that locatives belong to a distinct category in syntax from
nouns (along the lines of work by Baker and Stowell), or that
locatives are simply incompletely morphologized noun class markers
(following a hypothesis of Greenberg). The evidence clearly supports
the latter hypothesis.
The next workshop will be Tuesday, 15 May, when Robert Van Valin will
talk.
____________
PSYCHOLOGY DEPARTMENT COLLOQUIUM
Context, Activity, and Participation
Chuck and Candy Goodwin
Xerox PARC and University of South Carolina
Wednesday, 2 May, 3:45 p.m.
Building 420, Room 050
No abstract available.
____________