[Date Prev][Date Next][Date Index]
CSLI Calendar, Wednesday, 22 October 2003, vol. 19:8
CSLI CALENDAR OF PUBLIC EVENTS
______________________________________________________________________
22 October 2003 Stanford Vol. 19, No. 8
______________________________________________________________________
A weekly publication of the
Center for the Study of Language and Information (CSLI)
Stanford University, Ventura Hall, Stanford, CA 94305-4115
http://www-csli.stanford.edu/
____________
ACTIVITIES FROM 22 OCTOBER 2003 TO 31 OCTOBER 2003
WEDNESDAY, 22 OCTOBER 2003
12 noon Brain Lunch-Seminar
Bldg 380:383N
"On general architecture of Universal Learning Neurocomputers"
Victor Eliashberg
Universal Learning Systems
http://math.stanford.edu/seminars.htm
12 noon Psychology Developmental Brownbags
Bldg. 380:381U
"What's acquired in syntactic development: evidence from
crosslinguistic studies of word order"
Ulrike Hahn
Cardiff University
http://www-psych.stanford.edu/news.html#dev_brownbag
3:45pm Psychology Department Colloquium
Jordan Hall 420:041
"The Achievement Gap(s)"
Carol Dweck
Columbia University
http://www-psych.stanford.edu/news.html
4:15pm EE380: Computer Systems Lab Colloquium
Gates B03 (NEC Auditorium)
"Wireless Communication in a Post-Spectral World"
Andy Rappaport
August Capital
http://ee380.stanford.edu/contents.html
4:15pm CS300: Departmental Lecture Series
Gates B12
Leo Guibas
http://forum.stanford.edu/profile/guibas.html
http://forum.stanford.edu/students/cs300.php
5:00pm CS300: Departmental Lecture Series
Gates B12
Ron Fedkiw
http://forum.stanford.edu/profile/fedkiw.html
http://forum.stanford.edu/students/cs300.php
THURSDAY, 23 OCTOBER 2003
12:15pm Stanford Networking Seminar
Packard 202
"Apocrypha: Making P2P Overlays Network-aware"
Prasanna Ganesan
Stanford University
http://netseminar.stanford.edu/
Abstract below
3:30pm Carlos McClatchy Memorial Colloquium
McClatchy Hall, S-64 (screening room in sub-basement)
"Online Discussion and Democracy: The Electronic Dialogue Project"
Vincent Price
http://communication.stanford.edu/
4:00pm PARC Forum
George Pake Auditorium at PARC
"Ultra Wideband - the coming wave"
Rajeev Krishnamoorthy
TZero Technologies
http://www.parc.com/forum/
4:15pm CSLI Seminar on Computational Learning and Adaptation
Cordura Hall, room 100
"Reinforcement Learning in Multiagent Systems"
Junling Hu
http://cll.stanford.edu/scla.html
Abstract below
4:15pm SSP10: Symbolic Systems Forum
Bldg. 380:380C (Math Corner)
"Conversational Intelligence"
Stanley Peters
Linguistics, Stanford
http://symsys.stanford.edu:8081/ssp-dynamic/servlet/ssp_events
Abstract below
4:15pm CS300: Departmental Lecture Series
Gates B12
Monica Lam
http://forum.stanford.edu/profile/lam.html
http://forum.stanford.edu/students/cs300.php
5:00pm CS300: Departmental Lecture Series
Gates B12
Armando Fox
http://forum.stanford.edu/profile/fox.html
http://forum.stanford.edu/students/cs300.php
5:30pm Semantics and Pragmatics Workshop/Phonology Workshop
Bldg. 160:321 (Wallenberg Hall)
"The curious case of girl-swimming: morphology, lexical
semantics and the interpretation of compound"
Rochelle Lieber
University of New Hampshire
http://www-linguistics.stanford.edu/semgroup/
Abstract below
6:00pm SDForum Distinguished Speaker Series
George Pake Auditorium, PARC
"Ubiquitous Open Source: What it means for the software industry"
Mitch Kapor
Founder Lotus Development Corporation,
Electronic Frontier Foundation, and Open Source Application Foundation
http://www.sdforum.org/p/calEvent.asp?CID=1175&mo=10&yr=2003
(there is a fee for non-SDForum members, see web page)
7:00pm Global Mobility: Media and Minority Culture
A Stanford Humanities Center Conference
Kresge Auditorium
"Harnessing the Hive: How Social Software Transforms Global Media"
J.C. Herz
Journalist and Author of Joystick Nation and Surfing the Internet
http://shc.stanford.edu/global
FRIDAY, 24 OCTOBER 2003
all day Global Mobility: Media and Minority Culture
A Stanford Humanities Center Conference
Levinthal Hall, Stanford Humanities Center
http://shc.stanford.edu/global
11:00am Berkeley Institute of Cognitive and Brain Seminar
Tolman 5101 (Berkeley)
"The Cultural Origins of Human Cognition"
Michael Tomasello
Max Planck Institute for Evolutionary Anthropology, Leipzig
http://socrates.berkeley.edu:4247/seminar.html
12 noon Diversity in Language Seminar
Bldg. 260:252 German Studies Library
Papers for discussion:
Anna Wierzbicka, Preface to the second edition of
"Cross-Cultural Pragmatics"
Nick J. Enfield, "The theory of cultural logic: how
individuals combine social intelligence with semiotics to
create and maintain cultural meaning".
http://dlcl.stanford.edu/research/workgroups/diversity.html
RSVP to davidyo[at]stanford.edu
Information below
12 noon Ethics@Noon
Bldg. 100:101k
"Giving Money Away: Ethics, Public Policy, and Philanthropy"
Rob Reich
Political Science, Stanford
http://www.stanford.edu/dept/EIS/noon.htm
12 noon UC Berkeley Oxyopia Lecture
489 Minor Hall (UC Berkeley)
"What are the inputs to motion-selective striate-cortex cells?"
Russell DeValois, PhD
Psychology, and Optometry-Vision Science Programs, Berkeley.
http://spectacle.berkeley.edu/ucbso/oxyopia/oxy_current.html
12:30pm CS547: Human-Computer Interaction Seminar
Gates B01
"Smart Mobs: Mobile Communication, Pervasive Computing, and
Collective Action"
Howard Rheingold
http://www.rheingold.com/
http://www-pcd.stanford.edu/cs547/
Abstract below
3:15pm CLIR Minisummit
Cordura 100
"Adding Categorization, Geo-referencing and CLIR to AlertNet,
a news site for the International Relief Community~
Shuji Yamaguchi, Reuters Digital Vision Fellow, CSLI
"User-Assisted Query Translation in MIRACLE CLIR system"
Daqing He, Institute for Advanced Computer Studies, U. of Maryland
"Semantic Annotation and Ambiguity Resolution for a
Multilingual Medical Information System"
Dominic Widdows (et al.), CSLI
Abstract below
3:15pm Friday Cognitive Seminar
Jordan Hall 420:100
Title to be announced
Daniel Richarson
http://www-psych.stanford.edu/news.html#frisem
SATURDAY, 25 OCTOBER 2003
all day Global Mobility: Media and Minority Culture
A Stanford Humanities Center Conference
Levinthal Hall, Stanford Humanities Center
http://shc.stanford.edu/global
all day Biomedical Computation at Stanford 2003 (BCATS 2003)
TCSEQ
http://bcats.stanford.edu/
Information below
MONDAY, 27 OCTOBER 2003
10:30am Stanford Phonology Workshop
Margaret Jacks Hall 460:126
"Language-specific and context-specific markedness in lingual
consonant acquisition"
Mary Beckman
Ohio State University
http://www-linguistics.stanford.edu/pinterest/
(note unusual date and time)
Abstract below
4:15pm CS528: Broad Area Colloquium in AI,
Geometry, Graphics, Robotics, and Vision
TCSeq 200
"Dense Multiview Stereo Approaches for Handling Occlusions,
Highlights, Reflections, and Translucency"
Sing Bing Kang
Microsoft Research
http://graphics.stanford.edu/courses/cs528/
Abstract below
TUESDAY, 28 OCTOBER 2003
3:00pm CSLI Tea
Cordura Greenhouse
4:15pm Logic Seminar
Bldg. 380:380F (math corner)
"Report on decision methods for arithmetical
universal-existential sentences"
Jesse Alama and Patrick Girard
Stanford
http://www-logic.stanford.edu/seminars.html
Abstract below
WEDNESDAY, 29 OCTOBER 2003
12 noon Psychology Developmental Brownbags
Bldg. 380:381U
"How children infer unobserved causes"
Alison Gopnik
UC Berkeley
http://www-psych.stanford.edu/news.html#dev_brownbag
3:45pm Psychology Department Colloquium
Jordan Hall 420:041
"Auditory perception and auditory-visual interactions"
Gregg Recanzone
UC Davis
http://www-psych.stanford.edu/news.html
4:15pm EE380: Computer Systems Lab Colloquium
Gates B03 (NEC Auditorium)
"What's New in Python: Not your usual list of new features"
Guido van Rossum
Elemental Security and python.org
http://ee380.stanford.edu/contents.html
THURSDAY, 30 OCTOBER 2003
12 noon Semantics and Pragmatics Workshop
"The Agent Focus voice in Mayan languages"
Judith Tonhauser
Stanford
Margaret Jacks Hall 460:126
http://www-linguistics.stanford.edu/semgroup/
4:15pm SSP10: Symbolic Systems Forum
Bldg. 380:380C (Math Corner)
"Music Controllers: does the hardware of interaction affect the style"
Bill Verplank
Music, Stanford University
http://symsys.stanford.edu:8081/ssp-dynamic/servlet/ssp_events
4:15pm Information Systems Seminar
Packard 101
Title to be announced
Kannan Ramachandran
UC Berkeley
http://isl.stanford.edu/groups/seminar/
4:15pm Fundamental Themes in Neuroscience Seminar
Munzer Auditorium, Beckman B060
"Synaptic Circuits Serving Smell"
Jeffry Isaacson
Neuroscience, UC San Diego
http://www-med.stanford.edu/sbrc/calendar/
FRIDAY, 31 OCTOBER 2003
11:00am SRI STAR-Lab Seminars
EJ 124 (SRI International)
"Structured Language Models"
Ciprian Chelba
Microsoft
http://www.speech.sri.com/cgi-bin/run-cpp?private/seminars.html.cpp
Abstract below
11:00am Berkeley Institute of Cognitive and Brain Seminar
Tolman 5101 (Berkeley)
"Where the Action Is. Neural and Functional Considerations"
Anjan Chatterjee
Neurology and Center for Cognitive Neuroscience, U. of Pennsylvania
http://socrates.berkeley.edu:4247/seminar.html
12 noon UC Berkeley Oxyopia Lecture
489 Minor Hall (UC Berkeley)
"A relation between memory and auditory and visual attention"
Erv Hafter
http://spectacle.berkeley.edu/ucbso/oxyopia/oxy_current.html
12:30pm CS547: Human-Computer Interaction Seminar
Gates B01
"The Open Mind Initiative:
Large-scale knowledge acquisition from non-experts via the web"
David Stork
Ricoh Innovations
http://www-pcd.stanford.edu/cs547/
Abstract below
____________
Stanford Blood Center status: critical shortage of O+ and O-.
shortage of A+ and AB-. For an appointment:
http://bloodcenter.stanford.edu/ or call 650-723-7831. It only takes
an hour of your time.
____________
BRAIN-LUNCH SEMINAR
on Wednesday, 22 October 2003, 12 noon
Bldg. 380:383N
http://math.stanford.edu/seminars.htm
"On general architecture of Universal Learning Neurocomputers"
Victor Eliashberg
Universal Learning Systems
The talk will discuss the general architecture of Universal Learning
Neurocomputers and some possibilities of these systems as
computational metaphors for understanding the basic information
processing characteristics of the brain.
____________
STANFORD NETWORKING SEMINAR
on Thursday, 23 October 2003, 12:40pm (lunch 12:15pm)
Packard 202
http://netseminar.stanford.edu/
"Apocrypha: Making P2P Overlays Network-aware"
Prasanna Ganesan
Stanford University
A peer-to-peer(P2P) system consists of a dynamic set of nodes that
organizes itself in an overlay network. In general, this overlay
network may have nothing to do with the location of nodes on the
physical network. We propose a generic, distributed hill-climbing
mechanism called Apocrypha to make any P2P overlay "network-aware'',
and thus improve its quality. We demonstrate the applicability and
utility of Apocrypha on two different P2P systems: the Chord system
which uses a deterministic overlay topology, and the Gnutella system
which operates on an ad hoc topology. We also describe new, improved
routing protocols for the Chord system, and introduce the long-circuit
phenomenon in Gnutella.
This is joint work with Qixiang Sun and Hector Garcia-Molina.
About the Speaker: Prasanna Ganesan is a PhD student working with
Prof. Hector Garcia-Molina. His research interests include
Peer-to-Peer Systems, Databases and Algorithms. If you desperately
want to learn more about him, try Google.
Notes: Lunch will be available at 12:15. No drinks will be provided.
The talk itself will begin at 12:45.
____________
CSLI SEMINAR ON COMPUTATIONAL LEARNING AND ADAPTATION
on Thursday, 23 October 2003, 4:15pm to 5:30pm
Cordura 100
http://cll.stanford.edu/scla.html
"Reinforcement Learning in Multiagent Systems"
Junling Hu
Talkai, Inc.
In decentralized multiagent systems where agents pursue independent
goals, learning plays an important role for improving each agent's
performance. However, the task of learning in such environments is
much more formidable than that in single-agent systems. This is mainly
because other agents are learning at the same time, which causes the
constant change of the learning landscape. A successful learning
method therefore has to take such joint learning into account. In this
talk I will focus on one learning method we propose, which allows an
agent to learn its optimal actions based on its observations of joint
actions and its projections of other agents' future learning. We have
applied this method to several domains including E-commerce, where two
online companies engage in price competition. I show that our
multiagent reinforcement learning method gives considerably better
performance than using single-agent reinforcement learning.
____________
SSP10: SYMBOLIC SYSTEMS FORUM
on Thursday, 23 October 2003, 4:15pm
Bldg. 380:380C (Math Corner)
http://symsys.stanford.edu:8081/ssp-dynamic/servlet/ssp_events
"Conversational Intelligence"
Stanley Peters
Linguistics, Stanford University
http://www-csli.stanford.edu/~peters/
What knowledge and skills do people use in dialogue with other people?
They clearly need knowledge of a shared language, and it helps to know
something about the topics being discussed. Are additional skills
required, ones specific to communication, maybe even particular to
communicating in human language? I will sketch reasons for thinking
there are, ways of investigating the question further, and some
implications of our (still tentative) answer for technology that
converses with people.
About the Speaker: Professor Peters received his S.B. in Mathematics
from the Massachusetts Institute of Technology in 1963, and has since
held numerous academic positions. He spent 17 years at the University
of Texas-Austin, four of which were spent as the Chairman of their
Department of Linguistics. Since coming to Stanford in 1983, he
served as the Director of the Center for the Study of Language and
Information for from 1987 to 1990, and served as Chair of the
Department of Linguistics since 1996 to 2002. His current interests
include situation theory, mathematical properties of grammars, and
cooperative software.
____________
STANFORD PHONOLOGY WORKSHOP
on Thursday, 23 October 2003, 5:30pm
Bldg. 160:321 (Wallenberg Hall)
http://www-linguistics.stanford.edu/pinterest/
"The curious case of 'girl-swimming':
morphology, lexical semantics, and the interpretation of compounds"
Rochelle Lieber
University of New Hampshire
In this talk I will offer an analysis of compound interpretation for
both root and synthetic compounds in English. Such analyses in the
past have generally been couched in terms of syntax or argument
structure, but I will offer an analysis in terms of lexical semantic
representations based on the framework developed in my forthcoming
book Morphology and Lexical Semantics (Cambridge University Press, May
2004). In doing so, I will expand the range of data for which such
analyses must be responsible, challenging the 'facts' that have become
part of the repertoire of such analyses, specifically that it is never
possible for the first stem in a compound to receive a subject
interpretation.
____________
DIVERSITY IN LANGUAGE SEMINAR
on Friday, 24 October 2003, 12 noon
Bldg. 260:252 German Studies Library
http://dlcl.stanford.edu/research/workgroups/diversity.html
The second diversity group meeting will be Friday Oct. 24th, noon-1pm,
in 260-252 (German Studies Library), when we will discuss the two
papers below. We will provide food, so if you plan to attend **please
RSVP** to David Oshima, davidyo[at]stanford.edu, so that we know the
expected number of attendees.
Papers for discussion:
Anna Wierzbicka, Preface to the second edition of "Cross-Cultural
Pragmatics", and
Nick J. Enfield, "The theory of cultural logic: how individuals
combine social intelligence with semiotics to create and maintain
cultural meaning".
Copies of both papers will be made available in buildings 460, 50, and
260. The Enfield paper is also available online at:
http://www.stanford.edu/~sells/enfield.pdf
but note that it is large (10MB) and only displays/prints using the
latest version of Acrobat, 6.0.
Future mailings about our activities will *only* be sent to the
diversity-in-language mailing list. Here are the instructions for
getting on it, if you are not already:
Send mail to:
majordomo@lists.stanford.edu
with the following command in the body of your email message:
subscribe diversity-in-language
If you have problems, please contact David Oshima.
____________
CS547: HUMAN-COMPUTER INTERACTION SEMINAR
on Friday, 24 October 2003, 12:30-2:00pm
Gates B01
http://www-pcd.stanford.edu/cs547/
"Smart Mobs:
Mobile Communication, Pervasive Computing, and Collective Action"
Howard Rheingold
http://www.rheingold.com/
Smart mobs emerge when communication and computing technologies
amplify human talents for cooperation. The impacts of smart mob
technology already appear to be both beneficial and destructive, used
by some of its earliest adopters to support democracy and by others to
coordinate terrorist attacks.
The technologies that make smart mobs possible are mobile
communication devices and pervasive computing - inexpensive
microprocessors embedded in everyday objects and environments.
Already, governments have fallen, youth subcultures have blossomed
from Asia to Scandinavia, new industries have been born and older
industries have launched furious counterattacks.
Street demonstrators in the 1999 anti-WTO protests used dynamically
updated websites, cell-phones, and "swarming" tactics in the "battle
of Seattle." A million Filipinos toppled President Estrada through
public demonstrations organized through salvos of text messages.
The pieces of the puzzle are all around us now, but haven't joined
together yet. The radio chips designed to replace barcodes on
manufactured objects are part of it. Wireless Internet nodes in cafes,
hotels, and neighborhoods are part of it. Millions of people who lend
their computers to the search for extraterrestrial intelligence are
part of it. The way buyers and sellers rate each other on Internet
auction site eBay is part of it. Research by biologists, sociologists,
and economists into the nature of cooperation offer explanatory
frameworks.
The people who make up smart mobs cooperate in ways never before
possible because they carry devices that possess both communication
and computing capabilities. Their mobile devices connect them with
other information devices in the environment as well as with other
people's telephones. Dirt-cheap microprocessors embedded in everything
from box tops to shoes are beginning to permeate furniture, buildings,
neighborhoods, products with invisible intercommunicating
smartifacts. When they connect the tangible objects and places of our
daily lives with the Internet, handheld communication media could
mutate into wearable remote control devices for the physical world.
Media cartels and government agencies are seeking to reimpose the
regime of the broadcast era in which the customers of technology will
be deprived of the power to create and left only with the power to
consume. That power struggle is what the battles over file-sharing,
copy-protection, regulation of the radio spectrum are about. Are the
citizens of tomorrow going to be users, like the PC owners and website
creators who turned technology to widespread innovation? Or will they
be consumers, constrained from innovation and locked into the
technology and business models of entrenched interests?
About the Speaker: Howard Rheingold is the author of:
* Smart Mobs
* The Virtual Community
* Tools for Thought
was the editor of:
* The Whole Earth Review
* The Millennium Whole Earth Catalog
* HotWired
founded:
* Electric Minds
* Brainstorms
____________
CLIR MINISUMMIT
on Friday, 24 October 2003, 3:15pm
Cordura 100
"Adding Categorization, Geo-referencing and CLIR to AlertNet,
a news site for the International Relief Community~
Shuji Yamaguchi
Reuters Digital Vision Fellow, CSLI
This presentation talks about a project that will apply technologies
out of natural language processing research to AlertNet, a news site
run by Reuters Foundation. The project aims to add features to the
AlertNet site for the next 8 months. Such features include news
categorization, geo-reference and CLIR, all of which will be performed
semi-automatically in order to go beyond cost barrier of human editing
and to aggregate contents more extensively. CLIR for this project is
motivated by persistent "language divide" on Internet. System design
is discussed and approaches and technologies to be employed are
reviewed. As this is to explain a near future plan not a finished
research, suggestions are all welcome on the design and approaches.
"User-Assisted Query Translation in MIRACLE CLIR system"
Daqing He
Institute for Advanced Computer Studies
University of Maryland
The initiative of search process lies with the human searchers.
However human searchers understanding of what to look for and how to
find it are often inadequate at the start of a search session. Human
searchers in Cross Language Information Retrieval (CLIR) face a more
difficult challenge. Besides the difficulties encountered in
monolingual IR, they also must discover query terms that would lead
the system to translate into terms that might be found in relevant
documents. Moreover they must do so without even being able to read
those original documents due to language barriers.
In this talk, I will present our work on user-assisted query
translation in interactive CLIR, which aims at helping human searchers
to overcome some of those difficulties. The talk will concentrate
mainly on the design of several cues used in the query translation
process to help users to identify the meanings of translation
alternatives, the results of the experiments involving human subjects
to test the effectiveness of the query translation approach, and the
integration of user-assisted query translation into the MIRACLE
system, an interactive cross language search system developed during
Tides surprise language experiment.
~Semantic Annotation and Ambiguity Resolution
for a Multilingual Medical Information System"
Dominic Widdows
(Stanley Peters, Beate Dorow, Scott Cederberg and Chiu-Ki Chan)
CSLI
This talk describes a concept based information system for English and
German medical documents. Queries in one language are matched to
documents in another by using the Unified Medical Language System as a
semantic `metalanguage' (much as Latin terms are traditionally used
for medical concepts, independently of any one vernacular tongue).
This process relies on being able to automatically recognize medical
terms and map them to their appropriate UMLS codes, and on being able
to resolve the ambiguity which arises when such a mapping is multiply
defined. This problem is addressed using the structure of UMLS itself
(with no further human intervention) - notably, we use relationships
between concepts in an English knowledge source to resolve ambiguities
in both German and English documents. Compared with a baseline "vector
space translation" model, the semantic annotation produces much more
accurate results, and the automatic disambiguation component gives
further reliable (but slight) improvements. This raises interesting
questions about the value of detailed knowledge sources for cross
lingual information retrieval.
____________
BIOMEDICAL COMPUTATION AT STANFORD 2003 (BCATS 2003)
on Saturday, 25 October 2003, all day
TCSEQ
http://bcats.stanford.edu/
We are pleased to announce that on Saturday, October 25, 2003,
Stanford University will host the FOURTH ANNUAL BCATS (Biomedical
Computation at Stanford) SYMPOSIUM. It will provide an open and
interdisciplinary forum for Stanford Students, post-docs and
researchers to disseminate and share their latest research in widely
ranging fields involving biomedical computation. The symposium is
designed to expose the participants to up and coming researchers and
their ideas. The agenda will include two keynote addresses from
leaders in biomedical computation: Sean Eddy of Washington University,
and Peter Hunter of University of Auckland.
Another speaker to be announced shortly at http://bcats.stanford.edu/
Last year's symposium (BCATS 2002) was a huge success, thanks in large
part to the participation of the students and post-docs who made oral
and poster presentations. To ensure similar success, we would like to
encourage any Stanford students or post-docs whose research involves
biomedical computation to consider submitting abstracts of their
research for inclusion in the BCATS 2003 symposium and receive
feedback from the Stanford community and participants. As an
additional incentive we will continue the tradition of awarding prizes
for the best presentation and the best poster.
Technical Content will include:
* Informatics, Data Modeling and Biostatistics
* Biomechanical simulation and Modeling
* Structural Biology - Genetic and Evolutionary Computing
* Biomedical Image Acquisition and Processing
* Computer Assisted Interventions and Robotics
* Network and Computing Technology in Education
For more details, please visit http://bcats.stanford.edu.
Date abstract submission opens: September 1, 2003
Date abstract submission closes: September 22, 2003
Date registration opens: September 1, 2003
Please note that registration is free for Stanford affiliates.
In the meanwhile, mark your calendars, and we look forward to seeing
you at BCATS!
If you have any questions, please email us at bcats2003[at]email.com
____________
STANFORD PHONOLOGY WORKSHOP
on Monday, 27 October 2003, 5:30pm
Margaret Jacks Hall 460:126
http://www-linguistics.stanford.edu/pinterest/
(note unusual date)
"Language-specific and context-specific markedness
in lingual consonant acquisition"
Mary Beckman
Ohio State University
In his influential monograph on child language, aphasia, and
phonological universals, Jakobson (1941/1968) made several claims that
continue to shape how we think about acquisition today. The most
general claim was that there are universal substantive principles that
structure the phoneme inventories of all spoken languages, and that
these principles also determine how children acquire speech. More
specifically, he claimed that there is a clear discontinuity between
the rich set of consonant- and vowel-like sounds that children produce
randomly in their preverbal babbling and the initially impoverished
but reliable inventory of phonemes that they produce in their first
words. Moreover, the consonants and vowels that are first to be
mastered are the same for all children everywhere, and each child
expands this initial inventory in a rigid universal order. Subsequent
results from the large cross-language Stanford Child Phonology project
(e.g., Vihman et al., 1985) and other smaller projects on specific
languages show these specific claims to be quite wrong. At the same
time, the more general idea that there are fairly ubiquitous
markedness constraints is supported, at least for some consonant
features. Perhaps because of this support, accounts of acquisition in
mainstream North American phonological frameworks continue to assume
many of the specific universals that Jakobson posited. Most notably,
accounts of velar fronting in children acquiring Germanic languages
assume a universal ranking of lingual place of articulation features
whereby *DORSAL >> *CORONAL (e.g., Smolensky, 1996; Pater, 2003). In
this talk, I will first review the literature on acquisition of
phonation type contrasts, as a vehicle for suggesting what kind of
research is necessary to explore markedness hierarchies in
phonological acquisition. I will then review some pilot results from
the paidonlogos project (http://ling.osu.edu/~edwards) on lingual
obstruent acquisition. These results suggest that standard
assumptions about the earlier emergence of coronal consonants are
premature. Japanese- and Greek-acquiring children seem to master
velar consonant production first, except in the environment of front
vowels. The results support a richer account of markedness, one in
which language-specific phoneme frequencies play a role (see, e.g.,
Stemberger & Bernhardt, 1998) and consonant mastery is not viewed in
isolation from the phonetic constraints of neighboring vowels.
____________
CS528: BROAD AREA COLLOQUIUM FOR
AI-GEOMETRY-GRAPHICS-VISION-ROBOTICS
on Monday, 27 October 2003, 4:15pm
TCSeq 200
http://graphics.stanford.edu/courses/cs528/
"Dense Multiview Stereo Approaches for Handling Occlusions,
Highlights, Reflections, and Translucency"
Sing Bing Kang
Vision Technology Group, Microsoft Research
http://research.microsoft.com/users/sbkang/
While stereo matching was originally formulated as the recovery of 3D
shape from a pair of images, it is now generally recognized that using
more than two images can dramatically improve the quality of the
reconstruction. Unfortunately, as more images are added, the
prevalence of semi-occluded regions (pixels visible in some but not
all images) also increases. In addition, non-rigid effects in real
scenes such as highlights, reflections, and translucency further
complicate multiview stereo.
In this talk, I will describe how we progressively tackle the problems
of occlusion, highlights, reflections, and translucency. To handle
occlusion, we use a combination of shiftable windows and a dynamically
selected subset of the neighboring images to do the matches. To handle
highlights, we apply a color histogram differencing technique.
Finally, to take into account reflections and translucency, we model
the image formation as additive superposition of two layers at two
different depths, and solve for them iteratively. I will show results
for both synthetic and real image sequences as validation of these
approaches.
About the Speaker: Sing Bing Kang received his Ph.D. in robotics from
CMU in 1994. He is currently a researcher at Microsoft Corporation,
where he is working on environment modeling from images. His paper on
the Complex Extended Gaussian Image had won the IEEE Computer Society
Outstanding Paper award at CVPR'91. His IEEE Transactions on Robotics
and Automation paper on human-to-robot hand mapping had been awarded
the 1997 King-Sun Fu Memorial Best Transaction Paper award. Sing Bing
has published about 20 refereed journal papers and about 45 refereed
conference papers, and holds 11 US patents. He has also co-edited a
book on panoramic vision, which was published by Springer in 2001.
____________
LOGIC SEMINAR
on Tuesday, 28 October 2003, 4:15pm-5:30pm
Math Corner 380:380F
http://www-logic.stanford.edu/seminars.html
Note new room
"Decision methods for arithmetical universal-existential sentences"
Jesse Alama and Patrick Girard
Stanford
First order arithmetic is very expressive. Truth of very simple
classes of sentences is well known to be undecidable
(non-recursive). In this series of talks a much less known (and
potentially useful) fact is presented: the class of arithmetical
universal-existential sentences is decidable by an efficient
procedure. At the first meeting necessary definitions and main results
are stated. The only prerequisite is the knowledge of the arithmetical
language.
____________
SRI STAR-LAB SEMINAR
on Friday, 31 October 2003, 11:00am
EJ 124 (SRI International)
http://www.speech.sri.com/cgi-bin/run-cpp?private/seminars.html.cpp
"Structured Language Models"
Ciprian Chelba
Microsoft
The talk presents an attempt at using the syntactic structure in
natural language for improved language models for large vocabulary
speech recognition. The structured language model merges techniques in
automatic parsing and language modeling using an original
probabilistic parameterization of a shift-reduce parser. A maximum
likelihood re-estimation procedure belonging to the class of
expectation-maximization algorithms is employed for training the
model. Experiments on the Wall Street Journal, Switchboard and
Broadcast News corpora show improvement in both perplexity and word
error rate --- word lattice re-scoring --- over the standard 3-gram
language model. Further experiments investigate the portability of
syntactic structure across domains --- Wall Street Journal to Air
Travel Information Systems --- as well as the use of the structured
language model for information extraction from text. This work
investigates the use of frequency-localized temporal patterns of the
speech signal for developing robust front-end for Automatic Speech
Recognition (ASR).
____________
CS547: HUMAN-COMPUTER INTERACTION SEMINAR
on Friday, 31 October 2003, 12:30-2:00pm
Gates B01
http://www-pcd.stanford.edu/cs547/
"The Open Mind Initiative:
Large-scale knowledge acquisition from non-experts via the web"
David Stork
Ricoh Innovations and the Open Mind Initiative
http://rii.ricoh.com/%7Estork/
The Open Mind Initiative is a web-based collaborative framework for
collecting large knowledge bases from non-expert contributors. Such
knowledge bases are vital for a wide range of 'intelligent' software
such as speech and handwriting recognizers, commonsense reasoners, and
natural language understanding systems. This talk begins by examining
several important trends that underly Open Mind:
* the rise in open source software
* the expansion of opportunities for less-skilled users to
contribute knowledge
* the increase in scientific collaboration over the internet
* the growing need for large sets of 'informal' data from
non-experts
Next we contrast the Open Mind approach with traditional data mining,
and then describe ongoing projects collecting common sense, natural
language and handwriting recognition knowledge bases. Our largest
project, Open Mind common sense, has collected over a million simple
assertions from over tens of thousands of non-expert contributors.
Important considerations are speeding the collection of data (by
interactive learning techniques and motivating contributors) and
ensuring data quality (by identifying and filtering unreliable or even
'hostile' contributions). We discuss the importance of the
human-machine interface as well as the role of game interfaces.
The talk concludes with a vision of future directions and
opportunities.
About the Speaker: David G. Stork is Chief Scientist of Ricoh
Innovations as well as Consulting Professor of Electrical Engineering
and Visiting Lecturer in Art and Art History at Stanford
University. His primary interests lie in pattern recognition, machine
learning, neural networks and novel uses of the internet; he is the
creator and leader of the Open Mind Initiative. He sits on the
editorial boards of four international journals and his five books
include HAL's Legacy: 2001's computer as dream and reality (MIT Press)
for general audiences and the second edition of Pattern Classification
with R. Duda and P. Hart (Wiley).
____________
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 (i.e., no abstracts). 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://news.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
____________