[Prev][Next][Index]
CSLI Calendar, 13 April 1995, vol.10:23
-
To: friends
-
Subject: CSLI Calendar, 13 April 1995, vol.10:23
-
From: Tom Burke <burke>
-
Date: Wed, 12 Apr 1995 13:43:32 -0700
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
______________________________________________________________________________
13 April 1995 Stanford Vol. 10, No. 23
______________________________________________________________________________
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 12 -- 21 APRIL 1995
WEDNESDAY, 12 APRIL
4:00 - Seminar on Computational Learning and Adaptation
Cordura Hall, Room 100
Flexible Discriminant and Mixture Models
Trevor Hastie, Stanford Statistics
Abstract below
THURSDAY, 13 APRIL
12:15 - CSLI Seminar
Series on Intelligent Agents
Cordura Hall, Room 100
Meaning and Mechanism: Methodological Issues in Agent Design
Stan Rosenschein, Teleos and Stanford
Abstract below
2:15 - CSLI Seminar
Series on Visual Programming
Cordura Hall, Room 100
Rendering By Example
Ravi Krishnamurthy, Hewlett-Packard
Abstract below
4:15 - SSP Forum
Building 60, Room 61-F
The Verbmobil Project: Machine Translation
in the Late 20th Century
Dan Flickinger, CSLI
Abstract below
FRIDAY, 14 APRIL
12:30 - HCI Seminar
Skilling Auditorium
A Social Science Theory is a Practical Thing:
Social Responses to Communication Technology
and Microsoft's Bob
Cliff Nass and Byron Reeves, Stanford Communication
Abstract below
3:15 - Philosophy Colloquium
Building 90, Room 91-A
Laws of Nature and Natural History in the
Description of the World
James McAllister, Leiden Philosophy (Netherlands)
3:30 - Linguistics Colloquium (rescheduled from 7 April)
Cordura Hall, Room 100
So What Good are Linguists Anyway If They Can't Make
Computers Translate?
Martin Kay, Stanford Linguistics and Xerox PARC
Abstract below
TUESDAY, 18 APRIL
2:15 - CSLI Seminar
Ventura Hall, Room 17
Geometric Data Analysis
Henri Rouanet, University of Paris
8:00 - 1995 Immanuel Kant Lecture I
Building 420, Room 041
What is Empiricism, and What Could It Be?
Bas van Fraassen, Princeton Philosophy
WEDNESDAY, 19 APRIL
4:00 - Seminar on Computational Learning and Adaptation
Cordura Hall, Room 100
Feature Construction in Decision-Tree Induction
Cesar Montes Gracia, Technical University of Madrid
THURSDAY, 20 APRIL
10:00 - STASS Seminar
Cordura Hall, Room 100
Situation Theory and the Liar Paradox
John Etchemendy, Stanford Philosophy and Symbolic Systems
Abstract below
12:15 - CSLI Seminar
Series on Intelligent Agents
Cordura Hall, Room 100
Reacting, Planning, and Learning in an Autonomous Agent
Nils Nilsson, Stanford Computer Science
Abstract below
7:30 - Phonology Workshop
Building 460, Room 146
Rule Ordering and Constraint Interaction in OT
Young-Yee Yu Cho, Stanford Asian Languages
Abstract below
8:00 - 1995 Immanuel Kant Lecture II
Building 420, Room 041
The Manifest Image
Bas van Fraassen, Princeton Philosophy
FRIDAY, 21 APRIL
12:30 - HCI Seminar
Skilling Auditorium
Amazing Animation
Shannon Halgren, Claris
3:30 - Linguistics Colloquium
Cordura Hall, Room 100
Title to be announced
Joseph Greenberg, Stanford Linguistics
MONDAY, 24 APRIL
12:15 - CSLI Lecture/Tutorial
Turing Auditorium
Walking the Fine Line to Becoming a Webmaster
Christine Quinn, Stanford Networking and Communications
Abstract below
____________
The CSLI Calendar appears on Thursday of each week 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 the Web, URL
<http://www-csli.stanford.edu/>. The Calendar, with available abstracts, is
also posted each week to the csli.bboard newsgroup.
____________
SEMINAR ON COMPUTATIONAL LEARNING AND ADAPTATION
on Wednesday, 12 April
4:00 p.m., Cordura Hall, Room 100
Flexible Discriminant and Mixture Models
Trevor Hastie
Stanford Statistics
<trevor@playfair.stanford.edu>
Fisher's linear discriminant analysis (LDA) is a useful and graphical tool for
multi-group classification. With a large number of predictors, one can find a
reduced number of discriminant coordinate functions that are ``optimal'' for
separating the groups. With two such functions one can produce a
classification map that partitions the reduced space into regions that are
identified with group membership, and the decision boundaries are linear.
However, often this formulation is inadequate because:
* the linear decision boundaries can be too restrictive to capture
the class separation;
* if there are too many (correlated) predictors (e.g., grey-scale
pixel values from a digitized image), LDA tends to overfit the
training data, and the discriminant functions are too noisy to
be interpretable.
These deficiencies occupy opposite sides of the spectrum: in one case LDA is
too restrictive, in the other it is too flexible.
In this talk we represent LDA as an optimal scaling problem in multiple
regression, which we generalize to forms of nonparametric regression suitable
for the two scenarios outlined above. A natural extension of LDA is to model
each class by a mixture of Gaussians. Our approach adapts seamlessly, and it
can fit a mixture model, confined to an optimal lower dimensional subspace, by
regression methods.
This talk describes joint work with Rob Tibshirani and Andreas Buja.
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>.
____________
CSLI SEMINAR
SERIES ON INTELLIGENT AGENTS
on Thursday, 13 April
12:15 p.m., Cordura Hall, Room 100
Meaning and Mechanism: Methodological Issues in Agent Design
Stan Rosenschein
Teleos and Stanford Computer Science
<stan@teleos.com>
In the early days of AI, before symbolic computing, cybernetic models of
information and control reigned supreme. During that era, robot experimenters
built small devices, with sensors and effectors, that wandered around,
avoiding obstacles and moving from place to place. Beginning in the late
1950s, when symbolic computing began to displace cybernetic models as an
intellectual framework, experimentalists discovered the power of symbolic
tools for simulating reasoning and planning, and theoreticians discovered the
power of the *idea* of symbolic computing in explaining how the gap between
mental content and physical process might be bridged. Yet, when it comes to
building physical agents closely coupled to rich environments, even after
forty years of work, researchers are still building simple feedback-controlled
wandering robots. What explains the difficulty in applying symbolic models of
intelligence to physical robots? Does symbolic AI offer anything of value to
robot designers? If not, what other conceptual frameworks might be explored?
(This talk may get started a bit late since the speaker is lecturing
immediately before the talk elsewhere on campus.)
NOTE: A complete schedule for the CSLI seminar series on intelligent agents is
available at <http://www-csli.stanford.edu/>. These talks are scheduled for
12:15 (noonish) on Thursdays throughout the Spring quarter at Stanford, in
Cordura Hall, Room 100. If you order beforehand, we will provide a sandwich
and beverage (for $4.50, paid at the door; quarters and small bills
preferred). Send sandwich orders (turkey, ham, roastbeef, or vegetarian) to
<bocata@csli.stanford.edu> some time on or before Tuesday of the week of the
given talk.
____________
CSLI SEMINAR
SERIES ON VISUAL PROGRAMMING
on Thursday, 13 April
2:15 p.m., Cordura Hall, Room 100
Rendering By Example
Ravi Krishnamurthy
Hewlett-Packard Laboratories
<krishnamurthy@hpl.hp.com>
User friendly interfaces for databases have been widely considered an
important topic of research but has lacked its due attention from database
researchers for lack of meta understanding of the problem. Here we present a
linguistic abstraction, namely a visual language, for an important facet of
the problem.
Typical database application can be broadly partitioned into three repeatable
phases: (1) user input phase; (2) query computation phase; and (3)
post-interaction with the computed data. This post interaction phase is
typically so application specific that it requires significant programming as
evidenced in PC applications. These applications not only present the data in
an appropriate visual form but also allow for subsequent interaction with the
data by the user, in order to communicate/update the computed answer
effectively to/by the user. Such a presentation of data with subsequent
interaction secmantics is termed "rendering" of data.
The traditional individualistic applications did not lend to meta
understanding of the problem of rendering in an application independent
manner. Further, the statement of what to render and how-to-do-it (i.e., the
associated efficiency/implementation issues) were inextricably fused together
in these applications, analogous to the traditional problems observed in the
context of file management applications a few decades ago. As before, we
propose a declarative language for describing a chosen rendering of a computed
answer.
Rendering By Example is a declarative domain calculus language, quite similar
to the Query-By-Example, that can be used even by naive users with no
programming background. The declarative (i.e., what-you-want) aspect of the
language enables ease of statement, modification and reuse as well as
providing the separation of system (i.e., optimization, compilation, error
recovery etc.) issues from the functionality issues, all of which were proven
useful in the development of DB technology.
This work was done jointly with Moshe Zloof in the context of Picture
Programming project at HP Labs.
____________
SYMBOLIC SYSTEMS FORUM
on Thursday, 13 April
4:15 p.m., Building 60, Room 61-F
The Verbmobil Project: Machine Translation
in the Late 20th Century
Dan Flickinger
CSLI
<danf@csli.stanford.edu>
Since the early days of computers, machine-assisted translation from one human
language to another has been a tempting application of the technology. Much
of the research and development has been funded by large-scale government
projects, with periods of enthusiasm alternating with disillusionment. In the
past ten years, building on significant advances in hardware and in speech
recognition, Japan and Europe have launched massive efforts to build systems
that will translate ordinary spoken language.
The German government is currently funding one such effort, called Verbmobil,
a four-year project to design and build a prototype of a box that can sit on a
table between two people, and translate sentences or phrases upon demand, from
German or Japanese into English. Stanford's Center for the Study of Language
and Information is developing one small piece of this project, providing an
implementation of a grammar for English to be used in generating the results
of the translation process. In the talk, I describe the grand Verbmobil
scheme, the current application domain of scheduling a meeting time, and the
nature of the grammar we're building at CSLI, with the help of the Symbolic
Systems crew.
DAN FLICKINGER is a research engineer at CSLI, managing the Verbmobil English
grammar project which began in 1994 under the direction of Professor Ivan Sag.
Dan moved to CSLI from Hewlett-Packard Labs, where he worked for eleven years
managing a research group developing natural language understanding software,
and most recently a project on handwriting recognition. He received his Ph.D.
in linguistics from Stanford in 1987, writing a dissertation on the use of
inheritance and lexical rules to express generalizations in the lexicon.
____________
SEMINAR ON HUMAN-COMPUTER INTERACTION
on Friday, 14 April
12:30 p.m., Skilling Auditorium
A Social Science Theory is a Practical Thing: Social Responses
to Communication Technology and Microsoft's Bob
Cliff Nass and Byron Reeves
Stanford Communication
<nass@leland.stanford.edu>
For the past five years, we have been collaborating on a new paradigm in
human-computer interaction called "Social Responses to Communication
Technology" or "SRCT." Essentially, SRCT is a theory and methods that argues
that human's interactions with computers and other technologies is
fundamentally social and natural. Via a series of experiments, we have
demonstrated that one can directly apply theories and methods taken from the
social sciences to inform how people will respond to computers. The theory,
results, and methods of this research program were used in the development of
Microsoft's Bob. In the first part of our talk, we will discuss SRCT. In the
second part of our talk, we will demonstrate how SRCT was used to inform the
design of Bob.
PROFESSOR CLIFFORD NASS is an associate professor of Communication at Stanford
University, with appointments in Science, Technology, and Society, Sociology,
and Symbolic Systems. He received his B.A. cum laude In Mathematics and his
Ph.D. in Sociology, both from Princeton University. He has worked as a
computer scientist for the IBM Research Center in Yorktown Heights and Intel
Corp. He has been a principal investigator on grants sponsored by US West
Advanced Technologies, National Science Foundation, the Center for Integrated
Facilities Engineering, the Center for the Study of Language and Information,
and Stanford University. He has consulted for such organizations as American
Electronics Association, Amoco, Apple, Hewlett-Packard, IBM, Microsoft,
Northern Telecomm, and the Smithsonian Institute. He has published over
twenty journal articles and book chapters concerning technology and
statistical methodology.
PROFESSOR BYRON REEVES is the Paul C. Edwards Professor of Communication, and
Director of the Institute for Communication Research at Stanford University.
He is a nationally recognized expert on the psychological processing of media
in the areas of attention, emotions, learning, and physiological responses.
He has been a consultant to Capital Cities-ABC, The Disney Channel, The
Advanced Television Test Center, The Federal Communication Commission, Apple,
Hewlett-Packard, Microsoft, IBM, US West, Northern Telecomm, and Congress. He
has published extensively in communication, psychology, and neuroscience. His
academic background is in graphic design and music (B.F.A., Southern Methodist
University) and communication (Ph.D., Michigan State University).
____________
PHILOSOPHY COLLOQUIUM
on Friday, 14 April
3:15 p.m., Building 90, Room 91-A
Laws of Nature and Natural History
in the Description of the World
James McAllister
University of Leiden Philosophy
(Netherlands)
____________
LINGUISTICS COLLOQUIUM
on Friday, 14 April
3:30 p.m., Cordura Hall, Room 100
So What Good are Linguists Anyway If They Can't
Make Computers Translate?
Martin Kay
Stanford Linguistics and Xerox PARC
<kay.parc@xerox.com>
Machine translation gave the initial inspiration to computational linguistics
and continues to motivate much of the work. That is surely fair enough since
the problem is clearly computational and obviously linguistic. But forty
years of money and effort has brought us hardly any closer to the answer.
Some people think linguists are part of the problem when they should be part
of the answer. Jelinek, of IBM speech and machine translation fame at IBM,
said his project took a step forward whenever he fired a linguist. The world
continues to pour money down the same rat hole with little discernable
progress, with or without the linguists. The German government is giving it a
new twist: "Notice how we never seem to get anywhere on machine translation?
And, notice how we never seem to get anywhere on speech recognition? Well.
Tell you what ..." You guessed it. Verbmobil was born---speech to speech
translation research at a cost of $10,000,000 per annum. Mark my words:
before you know it, they will be bad mouthing us again.
So I think we should get together and organize against the oppressor. Lest
you may be in any doubt, I will tell you why they could not make the machines
translate and why it is not our fault.
____________
CSLI SEMINAR
on Tuesday, 18 April
2:15 p.m., Ventura Hall, Room 17
Geometric Data Analysis
Henri Rouanet
University of Paris
____________
1995 IMMANUEL KANT LECTURE I
on Tuesday, 18 April
8:00 p.m., Building 420, Room 041
What is Empiricism, and What Could It Be?
Bas van Fraassen
Princeton Philosophy
____________
SEMINAR ON COMPUTATIONAL LEARNING AND ADAPTATION
on Wednesday, 19 April
4:00 p.m., Cordura Hall, Room 100
Feature Construction in Decision-Tree Induction
Cesar Montes Gracia
Technical University of Madrid
____________
STASS SEMINAR
on Thursday, 20 April
10:00 a.m., Cordura Hall, Room 100
Situation Theory and the Liar Paradox
John Etchemendy
Stanford Philosophy and Symbolic Systems
<etch@csli.stanford.edu>
I will present an overview of the solution to the liar paradox presented in
Jon Barwise's and my book, _The Liar: An Essay and Truth and Circularity_.
____________
CSLI SEMINAR
SERIES ON INTELLIGENT AGENTS
on Thursday, 20 April
12:15 p.m., Cordura Hall, Room 100
Reacting, Planning, and Learning in an Autonomous Agent
Nils Nilsson
Stanford Computer Science
<nilsson@cs.stanford.edu>
We present an autonomous agent architecture and its component subsystems that
integrate important abilities needed for robust, flexible performance in
dynamic environments. These abilities involve appropriate reaction to
environmental situations given the agent's goals; selective attention to
multiple, competing goals; planning new action routines when innovation beyond
designer-provided routines is necessary; and learning the effects of actions
so that the planner can use them to build ever more reliable plans. The
teleo-reactive format allows actions to be closely coupled to continuous
environmental feedback and is also especially compatible with conventional AI
planning and learning mechanisms. The workings of the architecture and its
subsystems are illustrated in a simulated robot domain.
This is joint work with Scott Benson <sbenson@cs.stanford.edu; http://
robotics.stanford.edu/users/sbenson/bio.html>. Papers describing this work
can be downloaded from Benson's www home page and from mine.
NOTE: A complete schedule for the CSLI seminar series on intelligent agents is
available at <http://www-csli.stanford.edu/>. These talks are scheduled for
12:15 (noonish) on Thursdays throughout the Spring quarter at Stanford, in
Cordura Hall, Room 100. If you order beforehand, we will provide a sandwich
and beverage (for $4.50, paid at the door; quarters and small bills
preferred). Send sandwich orders (turkey, ham, roastbeef, or vegetarian) to
<bocata@csli.stanford.edu> some time on or before Tuesday of the week of the
given talk.
____________
PHONOLOGY WORKSHOP
on Thursday, 20 April
7:30 p.m., Building 460, Room 146
Rule Ordering and Constraint Interaction in OT
Young-Yee Yu Cho
Stanford Asian Languages
<yucho@csli.stanford.edu>
Rule ordering was one of the most powerful tools for phonological analysis
prior to the introduction of OT. In particular, numerous cases have been
reported of dialects or historical stages of a language that contain the same
underlying representations and the same rules, but differ simply by virtue of
the ordering of the rules. I argue that when equipped with two further
assumptions involving markedness and association of structure, OT has not only
the same kind of descriptive coverage in dealing with dialectal variation as
derivational theories but it also handles cases where the latter make
incorrect predictions. Two cases of where the constraint interaction of OT
diverges from the rule ordering of operational theories (Korean and Klamath)
will be discussed in depth.
____________
1995 IMMANUEL KANT LECTURE II
on Thursday, 20 April
8:00 p.m., Building 420, Room 041
The Manifest Image
Bas van Fraassen
Princeton Philosophy
____________
SEMINAR ON HUMAN-COMPUTER INTERACTION
on Friday, 21 April
12:30 p.m., Skilling Auditorium
Amazing Animation
Shannon Halgren
Claris
<shannon_halgren@claris.com>
____________
LINGUISTICS COLLOQUIUM
on Friday, 21 April
3:30 p.m., Cordura Hall, Room 100
Title to be announced
Joseph Greenberg
Stanford Linguistics
<josephg@csli.stanford.edu>
_____________
CSLI LECTURE / TUTORIAL
on Monday, 24 April
12:15 p.m., Turing Auditorium
Walking the Fine Line to Becoming a Webmaster
Christine Quinn
Stanford Networking and Communications
<quinn@tied-house.stanford.edu>
This will be a general introduction to the World Wide Web, including an
introduction to various Web browsers, illustrations of how the Web works, an
introduction to HTML (what you use to design Web pages), principles for
designing Web pages, and information about specific Web resources at Stanford.
_____________
NEW VISITOR
GODEHARD LINK (<link@csli.stanford.edu>, Cordura 214) is professor of logic
and philosophy of science at the University of Munich, Germany. In recent
years he has been doing extensive research in the logic and linguistic
semantics of natural language, pursuing an algebraic approach to semantic
structure. Currently he is working on relating mereological ideas in
natural-language semantics to problems in formal ontology, and also on a
unifying perspective of the semantics of conditionals. He has been a visitor
at CSLI in the academic year 1985--86, and again during the 1987 Linguistic
Institute. Dates of visit: April--October 1995.
_____________