[Date Prev][Date Next][Date Index]
CSLI Calendar, Monday, 26 June 2000, vol. 15:34
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 June 2000 Stanford Vol. 15, No. 34
______________________________________________________________________
A weekly publication of the
Center for the Study of Language and Information (CSLI)
Stanford University, Ventura Hall, Stanford, CA 94305-4115
____________
ACTIVITIES FROM 28 JUNE TO 6 JULY 2000
WEDNESDAY, 28 JUNE 2000
3:00pm CSLI Talk
Ventura 17
A Reinforcement Learning Dialogue System
Michael Kearns
AT&T Labs
http://calendus.stanford.edu/CogSci/read/event_9663_CogSci_read.html
Abstract below
THURSDAY, 29 JUNE 2000
10:30am SRI AI Seminar Series
EJ228, SRI International
Experimental Results from Integrating Planning Systems
and Simulation Models
Roberto Desimone
DERA, UK
http://www.ai.sri.com/ai-seminars/
Abstract below
4:00pm Xerox PARC Forum
George Pake Auditorium at Xerox PARC
Objects, Devices, and Networks
Jim Waldo
Sun Microsystems, Inc.
http://www.parc.xerox.com/ops/projects/forum/
Abstract below
FRIDAY, 30 JUNE 2000
10:30am SRI AI Seminar Series
EJ228, SRI International
Pre-Processing Inner-Outer Belief Axioms
for Persuasive Discourse Plans
Max Garagnani
Computing, The Open University
http://www.ai.sri.com/ai-seminars/
Same talk different time/location from the CSLI
Coglunch below. See it for more information.
THURSDAY, 6 JULY 2000
12 noon CSLI Coglunch
Cordura 100
Pre-Processing Inner-Outer Belief Axioms
for Persuasive Discourse Plans
Max Garagnani
Computing, The Open University
Abstract below
4:00pm Xerox PARC Forum
George Pake Auditorium at Xerox PARC
Role of Research Laboratories
Masao Kato
FX Palo Alto Laboratory
http://www.parc.xerox.com/ops/projects/forum/
____________
ANNOUNCEMENT
This is one of the irregular summer calendars. Regular Calendars will
start again in late September or early October.
____________
CSLI TALK
on Wednesday, 28 June 2000, 3:00 pm
Ventura Hall, Room 17
A Reinforcement Learning Dialogue System
Michael Kearns
AT&T Labs
Spoken dialogue systems communicate with users via automatic
speech recognition (ASR) and text-to-speech (TTS) interfaces,
and mediate the user's access to a back-end database. Designers
of such systems face a number of nontrivial choices in dialogue
strategy, including user vs. system initiative (the choice between
soliciting relatively open-ended vs. constrained user utterances),
and choices in confirmation strategy (when to confirm or re-prompt
for an ambiguous utterance). System design has typically been done
in an ad-hoc manner, with subsequent improvements to dialogue
strategy being fielded sequentially.
In this work, we apply the formalism of Markov decision processes
(MDPs) and the algorithms of reinforcement learning to the problem
of automated dialogue strategy synthesis. In this approach, an MDP
is built from training data gathered from an initial "exploratory"
system. This MDP provides a state-based statistical model of user
reactions to system actions, and is used to simultaneously evaluate
many dialogue strategies and choose the apparent optimal among
them. At AT&T Labs, we have applied this methodology in a dialogue
system for accessing a database of information on activities in New
Jersey, and have run controlled user experiments to evaluate the approach.
In this talk, I will describe our results, which include statistically
significant improvements in system performance, and discuss the issues
we faced in making the methodology work.
This talk describes joint work with Satinder Singh, Diane Litman, and
Lyn Walker of AT&T Labs.
____________
SRI AI SEMINAR SERIES
on Tuesday, 29 June 2000, 10:30am
Conference Room EJ 228
http://www.ai.sri.com/ai-seminars/
Experimental Results from Integrating Planning Systems
and Simulation Models
Roberto Desimone
DERA, UK
We will present results from the implementation of a prototype for
integrating planning and simulation systems. We will describe exactly
what has been implemented and use a non-combatant military evacuation
scenario (NEO) to show the benefits of such integration. The lessons
learned will be discussed and, finally, a description of a revised and
improved architecture which we believe will address some of the
outstanding research issues, will be proposed.
____________
XEROX PARC FORUM
on Thursday, 29 June 2000, 4:00pm - 5:00pm
George Pake Auditorium, Xerox
http://www.parc.xerox.com/ops/projects/forum/
Objects, Devices, and Networks
Jim Waldo
Sun Microsystems, Inc.
The standard model of distributed computing allows the motion of data
from one networked machine to another, and the definition of objects
that can be found and used on the network. These networks connect what
we generally think of as computers--desktops and servers running one
of a variety of standard operating systems. This standard model has
been built up over a number of years, and is based on assumptions and
models that have become so ingrained that they are rarely made
explicit, much less questioned.
Recent work in both hardware and software has changed a number of the
rules that have informed this standard model of distributed
computing. The most obvious change is the increase in the size of
networks, both in their geographical scope and the numbers of machines
that are connected. The variety of machines, on both the low and the
high end, has increased dramatically. Finally, the ability to move
code as well as data around the network in ways that are both
efficient and safe has offered an opportunity to change our view of
distributed computing.
The Jini(tm) connection technology exploits the ability to have mobile
code in a system the offers real distributed objects. These objects
are not available in a distributed fashion (as in systems like CORBA
or DCE RPC) but really exist on multiple machines at the same
time. Such a system allows an approach to dealing with the changing
nature of large networks, and allows the use of standard
object-oriented techniques in distributed systems development.
Biography: Jim Waldo is a Distinguished Engineer with Sun
Microsystems, where he is the lead architect for Jini, a distributed
programming system based on Java. Prior to Jini, Jim worked in
JavaSoft and Sun Microsystems Laboratories, where he did research in
the areas of object-oriented programming and systems, distributed
computing, and user environments.
Jim is an adjunct faculty member of Harvard University, where he
teaches distributed computing in the department of computer science.
Jim received his Ph.D. in philosophy from the University of
Massachusetts (Amherst). He also holds M.A. degrees in both
linguistics and philosophy from the University of Utah. He is a member
of the IEEE and ACM.
____________
CSLI COGLUNCH
on Thursday, 6 July 2000, 12 noon
Cordura Hall, Room 100
http://www-csli.stanford.edu/events/Coglunch/
(Also SRI AI Seminar on 30 June 2000 at 10:30am in EJ228)
Pre-Processing Inner-Outer Belief Axioms
for Persuasive Discourse Plans
Max Garagnani
Computing, The Open University
Walton Hall, Milton Keynes - MK7 6AA, UK.
mailto:m.garagnani@open.ac.uk
http://mcs.open.ac.uk/mg343
A 'discourse' can be defined as a linearised structure of semantically
related statements generated by a source ('speaker') and conveyed to
an audience ('hearer') in order to achieve a specific 'communicative
intention' (or goal). Thanks to the hierarchical and decompositional
nature of the communicative goals, the AI planning technique has been
successfully adopted by many researchers as a solution to the problem
of automatic discourse generation.
However, planning a discourse that can persuade the audience about the
validity (or fallacy) of a certain proposition requires the speaker to
hypothesise and maintain an adequate model of the hearer's beliefs and
inference. Moreover, generating sophisticated discourse plans requires
the adoption of an expressive belief language, which usually contains
mutual belief expressions, uncertainty, belief grounding and
justification, and, in general, doxastic attitudes that are often
semantically related by logical formulae ('inner-outer' belief
axioms).
In this talk, I argue that a clear distinction between the formal
model of speaker-hearer beliefs and the planning machinery which makes
use of it is needed for the realisation of complex, real-world
persuasive discourse plans. The main results presented are:
1) a formal speaker-hearer belief language Lo (and related belief
system) that allows the representation of uncertainty, defeasible
justification, belief grounding, functional discourse relations
and communicative intentions;
2) a set of (abstract) discourse planning
operators (defined using Lo) encoding different strategies for
sincere belief persuasion and undermining;
3) a sound and linear pre-processing algorithm which can be used to
integrate 'inner-outer' domain axioms within a set of planning
operators.
I also show how this approach has been applied to pre-process a set of
inner-outer belief axioms into the given set of discourse planning
operators, and how the results have been used by the 'IPP' planner to
produce real examples of complex discourse plans. I conclude with
some considerations on the limits and possible extensions of the
solution presented, and by sketching possible future directions of my
work.
____________
END MATERIAL
The CSLI Calendar appears weekly on Wednesdays throughout the academic
year. Announcements, abstracts, and other information to appear in
the Calendar should be submitted to the editor, who reserves the right
to decide what does or does not go in the calendar
mailto:incalendar@csli.stanford.edu.
Requests to be added to the mailing list should be sent to
majordomo@csli.stanford.edu. With the lines in the body of the text
of either
subscribe csli-calendar
for the long form or
subscribe csli-short-calendar
for the short form. Problems with subscribing or unsubscribing should
be sent to owner-csli-calendar@csli.stanford.edu.
The full current issue is at
http://www-csli.stanford.edu/Archive/calendar/current.shtml
and the archives at
http://www-csli.stanford.edu/Archive/calendar/
People on most of the CSLI computers can type 'help csli-calendar' to
see the current issue.
The CSLI Calendar is also posted each week to
news://nntp-csli.stanford.edu/csli.bboard.
and
news://nntp.stanford.edu/su.events
Information about CSLI's research program is available at
http://www-csli.stanford.edu/
For maps to the Stanford University campus see
http://www.stanford.edu/home/visitors/maps.html
____________