[Prev][Index]

CSLI Calendar, 2 June 1994, vol.9:31



        
        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
______________________________________________________________________________

2 June 1994                      Stanford                       Vol. 9, No. 31
______________________________________________________________________________

      A weekly publication of the Center for the Study of Language and
Information (CSLI), Stanford University, Ventura Hall, Stanford, CA 94305-4115
                               ____________
                                    
               CSLI ACTIVITIES DURING 2 -- 10 JUNE 1994

  THURSDAY, 2 JUNE
        10:00 - STASS Seminar
                Cordura Hall, Room 100
                Discourse Structure and Some Problems of Repair
                Livia Polanyi, Rice University
                Abstract below

         2:30 - CSLI Seminar
                Cordura Hall, Room 100
                Philosophy and the Slingshot
                Stephen Neale, UC Berkeley Philosophy
                Abstract below

         4:15 - SSP Forum
                Building 60, Room 61-F
                Interacting with the World of Networked Media
                Terry Winograd, Stanford Computer Science
                Abstract below

         5:30 - Special Linguistics Colloquium
                Ventura Hall, Room 17
                Semantic Atoms of Anaphora: Selfish Languages
                and Selfless Ones
                Ken Safir, Rutgers Linguistics
                Abstract below

  FRIDAY, 3 JUNE
       8:30am - CSLI Workshop
                Cordura Hall, Room 100
                Third CSLI Workshop on Logic, Language, and Computation
                Schedule below

         3:30 - Linguistics Colloquium
                Ventura Hall, Room 17
                Cutting the Ergativity Pie
                Chris Manning, Stanford Linguistics
                Abstract below

  TUESDAY, 7 JUNE
         3:00 - Seminar on Computational Learning and Adaptation
                Cordura Hall, Room 100
                Concept Learning Rates and Transfer Performance of Several
                Multivariate Neural Network Models
                Patrick Suppes and Lin Yiang, CSLI
                Abstract below
                
                               ____________

The CSLI Calendar appears on Thursday of each week (more or less) throughout
the academic year.  Announcements, abstracts, and other information to appear
in the Calendar on a given Thursday should be submitted by e-mail to
incalendar@csli.stanford.edu by 5:00 p.m. on the previous Tuesday.

Past issues of the CSLI Calendar, a quarterly schedule of upcoming CSLI
events, and other information about CSLI are available on-line by way of a
CSLI gopher server on kanpai.stanford.edu.  The Calendar, with all available
abstracts, is also posted each week to the csli.bboard newsgroup.

                               ____________

This will be the last issue of the CSLI Calendar for the 1993-94 academic
year.  Weekly publication of the Calendar will begin again in the last week of
September 1994.

                               ____________

                              STASS SEMINAR
                           on Thursday, 2 June
                    10:00 a.m., Cordura Hall, Room 100
             Discourse Structure and Some Problems of Repair
                              Livia Polanyi
                             Rice University
                         livia@csli.stanford.edu

While most previous presentations about the Linguistic Discourse Model (LDM)
have concentrated on discourse syntax, in this talk I will present an overview
of the model as a framework for investigating the complex semantic issues
raised by the analysis of naturally occuring interactive discourse.

Under the LDM the unit of discourse formation is the elementary _discourse
constituent unit_ or _dcu_, a semantically motivated semiotic structure which
expresses a single event or state of affairs in some semantic "world of
discourse."  Elementary dcus are always indexed for situation of utterance,
including (real or modeled) Interaction and Speech Event, as well as
linguistic contexts such as genre unit (if any) and modal contexts including
modality, polarity, specificity, point of view, and so forth.  (Contexts may
be underspecified as is the case for polarity in YES/NO questions, for
example.)  Complex discourse is formed from elementary dcus by recursive
sequencing and embedding of dcus to dcus.

The structural description of any discourse is an Open Right Discourse Parse
Tree (DPT) assigned incrementally as each incoming elementary dcu is attached
to the developing DPT as the right daughter of an appropriate existing or
newly created node along the Right Edge of the Tree.  The attachment of the
elementary dcu to the DPT's Right Edge is taken as an instruction to discourse
semantics to create a new discourse representational structure to accommodate
the propositional information asserted by the dcu or to increment or otherwise
modify existing representations with the new information.

In this talk, I will motivate the LDM framework by presenting a simplified
example of a complex interactive discourse and, after describing the LDM
syntactic and semantic machinery, I will present analyses of example
discourses involving repair sequences which illustrate how that machinery
works and which show why, under the LDM, discourse SYNTAX is taken to be
monotonic while discourse SEMANTICS must be non-monotonic.

                               ____________
                                     
                               CSLI SEMINAR
                           on Thursday, 2 June
                    2:30 p.m., Cordura Hall, Room 100
                       Philosophy and the Slingshot
                              Stephen Neale
                          UC Berkeley Philosophy
                        neale@garnet.berkeley.edu

The philosophical literature contains many interesting examinations of the the
"collapsing" or "slingshot" arguments presented by Church, Godel, Quine, and
Davidson.  I maintain that none does justice to the subtlety of the
linguistic, logical, and metaphysical issues particular slingshots raise.  In
this talk I shall examine Godel's and Davidson's slingshots in connection with
an ontology of facts and with correspondence theories of truth.  I will argue
that (1) Godel was correct to point out that Russell's account of facts is
saved by his Theory of Descriptions; (2) there is a common conception of
`fact' that is completely undermined by Davidson's slingshot; (3) there are
other conceptions of `fact' that survive the slingshot; (4) it is unclear
whether the survivors have any philosophical utility or significance; (5)
there can be no generalized slingshot argument; (6) only one (of four) popular
non-Russellian treatments of descriptions provides the means of turning
Davidson's slingshot into a serious cause for concern; (7) there is much
handwaving in the literature in connection with non-Russellian treatments of
descriptions and a model-theoretic account of logical equivalence; (8)
Davidson's cases against conceptual relativism and scepticism depend upon the
success of a particular version of his slingshot.

                               ____________

                          SYMBOLIC SYSTEMS FORUM
                           on Thursday, 2 June
                    4:15 p.m., Building 60, Room 61-F
              Interacting with the World of Networked Media
                              Terry Winograd
                        Stanford Computer Science
                         winograd@cs.stanford.edu

The explosive growth of the Internet and information spaces like the World
Wide Web has fueled a rapid wide proliferation of information servers and
browsers.  The resulting richness and chaos have been exhilarating and have
also created many problems, both social and technical.  I will talk about of
each of these dimensions and some of the projects underway to "tame the
information wilderness."

                               ____________

                          LINGUISTICS COLLOQUIUM
                            on Thursday, 2 June
                     5:30 p.m., Ventura Hall, Room 17
     Semantic Atoms of Anaphora: Selfish Languages and Selfless Ones
                                Ken Safir
                           Rutgers Linguistics
                         safir@zodiac.rutgers.edu

Do anaphors have semantic content?  It is often tacitly assumed that elements
used syntactically as anaphors, that is, elements that must be bound by an
antecedent in a given syntactically determined domain, are "bleached" of
semantic content.  If anaphors differ in their distribution, the relational
meaning of morphemes internal to the anaphor are usually treated as accidental
or only historically relevant to these differences.  For example, it is
generally assumed that the difference of the independent meanings of /self/ in
English or /meme/ ("same") in French is not a relevant factor in
distinguishing differences in the distribution of the anaphors /himself/ and
/lui-meme/ ("him-same") in English and French, respectively.

The central thesis of this talk is that the semantic content of anaphors can
indeed be an active factor in determining the anaphoric pattern.  I will argue
further that anaphors which respect familiar treatments of Principle A are
contentful predicates and that the semantic relational content of these
predicates will restrict their distribution, at least in part, in a very
general way.  Not only will predicates of different semantic content behave
differently from one another, but those anaphors that are not contentful will
behave differently as a class from all of the contentful ones.  Although the
existence and distribution of relational anaphors will be consistent with
Principle A, I will also argue that not all elements that have been called
anaphors fall under it.  My proposal will be developed in the context of a
more general approach to anaphora that I will call "universalist" (Safir
1993).  The central premise of the universalist position is that all anaphora
specific statements are derived from independently motivated properties of
lexical items that do not modify or select amongst principles of anaphora.

                               ____________
                                     
                              CSLI WORKSHOP
                      Friday 3 June -- Sunday 5 June
               8:30 a.m.--5:00 p.m., Cordura Hall, Room 100
         Third CSLI Workshop on Language, Logic, and Computation
                     {atocha,nino}@csli.stanford.edu

The Center for the Study of Language and Information (CSLI), Stanford
University, will host the Third Workshop on Logic, Language, and Computation
on 3--5 June.  The workshop will take place in the large conference room in
Cordura Hall (Room 100).  Registration is free and open.

This meeting is a follow-up to similar workshops held in the previous two
years.  This annual event brings together philosophers, linguists, and
computer scientists with an interest in logic, with the overall aim of
facilitating interdisciplinary interaction.

The organizers of the workshop are Johan van Benthem, Stanley Peters, Atocha
Aliseda (atocha@csli.stanford.edu), and Maria-Eugenia Ni~no
(nino@csli.stanford.edu).

                             WORKSHOP PROGRAM

Each talk will consist of 30 minutes of presentation followed by a
10 minute question and discussion period.

--------------------------------------------------------------------------
Friday, 3 June

Morning      I. PROCESS LOGICS AND PROOF THEORY
                Chair: Jens Erik Fenstad

8:45-9:00       Opening Remarks by Johan van Benthem

9:00-9:40       Grigori Mints (Stanford) 
                Hilbert's Program and the Substitution Method

9:40-10:20      Dick de Jongh (Amsterdam)
                Exact Provability in Intuitionistic Logic and 
                Provability Logic

10:20-10:35     Coffee break

10:35-11:15     Vaughan Pratt (Stanford)
                Complementarity and Uncertainty in Rational Mechanics

11:15-11:55     Patrick Lincoln (SRI)
                Automating Linear Logic

11:55-12:35     Marianne Kalsbeek (University of Amsterdam)
                A Logic for Propositional Prolog

12:35-1:25      Lunch break

Afternoon   II. MULTIMODAL AND VISUAL REASONING
                Chair: Grigori Mints

1:25-2:05       Jens Erik Fenstad (Xerox PARC)
                Formal Semantics Needs Geometry

2:05-2:45       Mun-Kew Leong (Stanford)
                Visual Reasoning with Maps

2:45-3:00       Coffee break

3:00-3:40       Yoav Shoham (Stanford)
                Knowledge, Certainty, Belief, and Conditionalization

3:40-4:20       Joe Halpern (IBM Almaden)
                Conditional Logics of Belief Change

--------------------------------------------------------------------------
Saturday, 4 June

Morning    III. PROCESSING NATURAL LANGUAGE
                Chair: David Israel

9:00-9:40       Ronald Kaplan (Xerox PARC)
                Restriction as a Grammatical Descriptor

9:40-10:20      Mark E. Stickel (SRI)
                Upside-Down Meta-Interpretation of the Model Elimination
                Theorem-Proving Procedure for Deduction and Abduction

10:20-10:35     Coffee break

10:35-11:15     Bob Moore (SRI)
                A Natural Language Understanding System based on UG

11:15-11:55     Narciso Marti-Oliet and Jose Meseguer (SRI)
                Action and Change in Rewriting Logic

11:55-12:35     Patrick Blackburn (Utrecht) and Maarten de Rijke
                (CWI Amsterdam)
                Logics of Communicating Structures

12:35-1:25      Lunch break

Afternoon   IV. CONTEXT CHANGE AND DYNAMIC INTERPRETATION
                Chair: Greg O'Hair

1:25-2:05       Tom Burke (Stanford)
                Dynamic Logic and Situation Theory

2:05-2:45       Sasa Buvac (Stanford)
                Logic of Contexts

2:45-3:00       Coffee break

3:00-3:40       Giovanna Cepparello (Scuola Normale Superiore, Pisa)
                Tarskian Variations

3:40-4:20       Paul Dekker (ILLC, Amsterdam)
                Predicate Logic with Anaphora

4:20-5:00       Jeroen Groenendijk (Amsterdam)
                Coreference and Modality

Evening         Party

--------------------------------------------------------------------------
Sunday, 5 June

Morning      V. QUANTIFIERS, ANAPHORA, AND DISCOURSE
                Chair: Alex Lascarides

9:00-9:40       Friederike Moltmann (UCLA)

9:40-10:20      Filippo Beghelli, Tim Stowell, and Anna Szabolcsi (UCLA)
                "Predicate" and "Distribute" in Scope Assignment

10:20-10:35     Coffee break

10:35-11:15     Seungho Nam (UCLA)
                Argument Orientation of Locative PPs in English

11:15-11:55     Jaap van der Does and Henk Verkuyl (Utrecht)
                Quantification and Predication

11:55-12:35     Lauri Hella (Helsinki), Jouko V"a"an"anen (Helsinki),
                and Dag Westerstahl (Stockholm)
                Definability of Polyadic Lifts of Generalized 
                Quantifiers

12:35-1:25      Lunch break

Afternoon       Chair: Stanley Peters

1:25-2:05       Ed Stabler (UCLA)
                Finite Incremental Parsers for Human Languages

2:05-2:45       Makoto Kanazawa (Stanford)
                What Makes Categorial Grammars Learnable?

2:45-3:00       Coffee break

3:00-3:40       Alex Lascarides (Stanford)
                Intentions and Information in Discourse

3:40-4:20       Chris Pi~n'on (Stanford)
                Comparative Aspectual Composition

4:20-5:00       Eric Jackson (Stanford)
                Negative Polarity and Strong Statements

5:00            End

                               ____________

                          LINGUISTICS COLLOQUIUM
                            on Friday, 3 June
                     3:30 p.m., Ventura Hall, Room 17
                        Cutting the Ergativity Pie
                              Chris Manning
                           Stanford Linguistics
                        manning@csli.stanford.edu

Anderson (1976) suggested that the majority of ergative languages appear to
have accusative syntax if tested via the major cyclic transformations.  This
thesis led to the current orthodoxy (Dixon 1979) of a world consisting of
accusative languages and ergative languages with the latter mainly
morphologically ergative languages but with the odd one syntactically ergative
(Dyirbal seems the only uncontroversially accepted exemplar).

One problem has been that the Philippine languages have always sat uneasily
outside this classification system.  This is especially disturbing since a
minority faction (Johnson 1980, Payne 1982, Blake 1988) has repeatedly
observed the similarity between Philippine and various ergative languages.
This talk begins with Kroeger's (1993) analysis of Tagalog (which pace
Schachter (1976, 1977) analyzes the Philippinists' topic as being the subject)
and extends it to cover ergative languages, building on what I take to be the
essence of Bittner (1994) and Sadock's (ms) analyses of Inuit and Campana
(1992) and Van Valin's (1982) analyses of Mayan languages.  The result is a
large supply of syntactically ergative languages (but under a different
definition of subjecthood!) and "semantic" (argument structure based) theories
of binding and control.

                               ____________

             SEMINAR ON COMPUTATIONAL LEARNING AND ADAPTATION
                            on Tuesday, 7 June
                    3:00 p.m., Cordura Hall, Room 100
            Concept Learning Rates and Transfer Performance of
                Several Multivariate Neural Network Models
                       Patrick Suppes and Lin Yiang
                                   CSLI
                     {suppes,liang}@csli.stanford.edu
        
In this talk we present results for several related learning models.  We
compare the models in two major ways: their relative rates of concept learning
and their relative transfer performance when they learn one concept followed
by another. We compare the models on six sets of data: (i) the well-known
Edgar Anderson data on three species of iris, which provides four measurements
for each sample; (ii) measurement data on genuine and forged Swiss bank notes
(Flury & Riedwyl 1988); (iii) measurement data on three species of beetles
(Lubischew 1962); (iv) randomly generated data on two related artificial but
interesting problems of classification, one in which positive instances are
bounded by two concentric circles and the other in which they are bounded by
two simply connected regions of irregular shape; (v) the classical XOR
problem; and (vi) transfer data from experiments reported by Suppes (1965) on
young children learning identity and equipollence of sets. We begin by
describing a multivariate normal learning model, and then introduce a
"meta-model" for comparing the learning behavior of several alternative
models. We present and analyze learning results for these models on data sets
(i) through (v), then discuss problems of transfer, focusing on data set (vi)
in particular. Finally, we compares the results to those for a variety of
other models discussed in the literature.

The goal of this seminar is to increase communication among local researchers
with interests in computational approaches to learning and adaptation.  If you
would like to be added to (or removed from) the mailing list, or if you are
interested in giving a talk in the seminar, please send email to
<langley@cs.stanford.edu>.

                               ____________