[Date Prev][Date Next][Date Index]
CSLI Calendar, Wednesday, 5 March 2003, vol. 18:23
CSLI CALENDAR OF PUBLIC EVENTS
______________________________________________________________________
5 March 2003 Stanford Vol. 18, No. 23
______________________________________________________________________
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 5 MARCH 2003 TO 14 MARCH 2003
WEDNESDAY, 5 MARCH 2003
3:45pm Psychology Department Colloquium
Jordan Hall 420:041
"Categorization as Informed by Culture and Expertise"
Doug Medin
Northwestern University
http://www-psych.stanford.edu/news.html
4:15pm EE380: Computer Systems Lab Colloquium
Gates B03 (NEC Auditorium)
"The Future of Systems Research"
John Hennessy
President, Stanford University
(this is a video repeat of a talk on 3 October 2001, the
originally scheduled talk was cancelled due to illness)
http://ee380.stanford.edu/contents.html
THURSDAY, 6 MARCH 2003
12:15pm Stanford Networking Seminar
Gates 104
"Touch-Based Personal Area Networks"
Kurt Partridge
University of Washington
http://netseminar.stanford.edu/
Abstract below
1:00pm Berkeley International Computer Science Institute
ICSI, Rm 607 (UC Berkeley)
"Phenotropics, or Prospects for Protocol-adverse Computing"
Jaron Lanier
http://www.icsi.berkeley.edu/talks/
Abstract below
4:00pm PARC Forum
George Pake Auditorium at PARC
"Sustaining Success in a Challenging Economy: It's All About Results"
Eric Schmidt
Chairman and CEO, Google, Inc.
http://www.parc.com/forum/
4:00pm UC Berkeley CIS Seminar
Soda Hall 380 (UC Berkeley)
"Being There:
Capturing and Rendering the Realistic Appearance of the Visual World"
Richard Szeliski
Microsoft Research
http://www.cs.berkeley.edu/~eyal/cis-seminar
Abstract below
4:00pm Berkeley BISC Seminar
320 Soda Hall, Berkeley
"From Search Engines to Question-Answering System --
The Need for New Tools"
Lotfi A. Zadeh
BISC, CS Division, EECS Dept. UC Berkeley
http://buffy.eecs.berkeley.edu/Seminars/
Abstract below
4:15pm CSLI Seminar on Computational Learning and Adaptation
Cordura Hall, room 100
"Statistical Modeling of Large-Scale Scientific Simulation Data"
Tina Eliassi-Rad
Center for Applied Scientific Computing
Lawrence Livermore National Laboratory
http://cll.stanford.edu/scla.html
Abstract below
4:15pm SSP10: Symbolic Systems Forum
Bldg. 380:380C (Math Corner)
"Knowing Your Own Mind"
Krista Lawlor
Philosophy, Stanford
http://www.stanford.edu/~klawlor/
http://symsys.stanford.edu:8081/ssp-dynamic/servlet/ssp_events
Abstract below
4:15pm Neurobiology of Disease Seminar Series
Munzer Auditorium, Beckman B060
"Growth Factor Modulation of Neuronal Survival and Function in the
Adult CNS"
Mark Tuszynski
UC San Diego
http://www-med.stanford.edu/sbrc/calendar/
5:30pm Linguistics Department Colloquium
Margaret Jacks Hall 460:126
"Against Iconicity and Markedness"
Martin Haspelmath
Max-Planck-Institut fur evolutionare Anthropologie, Leipzig
http://www-linguistics.stanford.edu/colloq/
Abstract below
FRIDAY, 7 MARCH 2003
11:00am Berkeley Institute of Cognitive and Brain Seminar
Tolman 5101 (Berkeley)
"Constructions, Frequency Effects and Two Types of Syntactic
Processing"
Joan Bybee
Linguistics, University of New Mexico
http://socrates.berkeley.edu:4247/seminar.html
12 noon Logical Methods in the Humanities
Bldg. 380:383N (math corner)
"A User-Friendly Decision Procedure for the Probability Calculus,
with Some Applications to Bayesian Philosophy of Science"
Branden Fitelson
Philosophy, SJSU
http://www-logic.stanford.edu/seminars.html
Abstract below
12:30pm CS547: Human-Computer Interaction Seminar
Gates B03
"Art Therapy and Tele-Mediated Communication in Behavioral Telehealth"
Kate Collie
University of British Columbia
http://www-pcd.stanford.edu/cs547/
Abstract below
2:15pm NLP Reading Group
Math Corner, 380:383P
"The Limits of N-Gram Translation Evaluation Metrics"
Presented by: Susanne Riehemann and Chris Culy, SRI
http://www-nlp.stanford.edu/nlp/nlpgroup.html
Abstract below
3:15pm Philosophy Department Colloquium
Bldg. 90:92Q
"Carnap, Completeness, and Logical Consequence"
Erich Reck
University of California, Riverside
http://www-philosophy.stanford.edu/ce.html
4:15pm CS545: Database Seminar
Gates B12
"Internal Architectures for Object-Relational DBMS Engines"
Paul Brown
IBM Research (Almaden)
http://www-db.stanford.edu/dbseminar/
MONDAY, 10 MARCH 2003
11:00am SRI AI Seminar Series
EJ228, SRI International
"On the utility of conceptual semantics in translating event
descriptions across languages"
Robert Belvin
http://www.ai.sri.com/seminars/
Abstract below
3:30pm Social Lab
Jordan Hall 420:050
"Effect Propensity"
Itamar Simonson
Business, Stanford
http://www-psych.stanford.edu/news.html#social_lab
4:00pm Berkeley Slavic Department Colloquium
219 Dwinelle (Berkeley)
"'Even' in Discourse: The Interaction of Lexical Semantics and
Interpretation Strategies"
Igor Boguslavskii
(talk is in Russian)
http://linguistics.berkeley.edu/lingdept/Current/events.html
4:15pm Broad Area Colloquium in AI,
Geometry, Graphics, Robotics, and Vision
TCSeq 201
Title to announced
James Duncan
Yale
http://www.eng.yale.edu/faculty/vita/duncan.html
http://robotics.stanford.edu/ba-colloquium/
5:00pm Berkeley Linguistics Department Colloquium
182 Dwinelle (Berkeley)
"Ditransitive constructions in the world's languages:
alignment types, alignment splits, and inverse patterns"
Martin Haspelmath
Max Planck Institute for Evolutionary Anthropology, Leipzig
http://linguistics.berkeley.edu/lingdept/Current/events.html
Abstract below
TUESDAY, 11 MARCH 2003
12 noon Linguistics Department Colloquium
Margaret Jacks Hall 460:126
"Quantifying Ambiguity Resolution with Information Theory"
John Hale
Johns Hopkins University
http://www-linguistics.stanford.edu/colloq/
Abstract below
4:00pm Berkeley BISC Seminar
373 Soda Hall, Berkeley
"Concept-Based Information Retrieval and Search Engine"
Tomohiro Takagi
Computer Science Department, Meiji University, Japan
http://buffy.eecs.berkeley.edu/Seminars/
Abstract below
4:15pm Computer Science Faculty Candidate Talk
Packard 202
"Query Processing in Sensor Networks"
Sam Madden
UC Berkeley
Abstract below
4:15pm SNRC Industry Seminar
Packard 101
"IPv6 Networks and P2P Solutions"
Kazuho Miki
Director & Lab Manager, Hitachi America Ltd.
http://snrc.stanford.edu/events/industry-seminar/
Abstract below
6:30pm Emerging Technology Group
Cubberley Community Center, H-1, 4000 Middlefield, Palo Alto
"Fundamental Problems in Interactive Storytelling"
Chris Crawford
http://www.sdforum.org/
(there is a fee for non-SDForum members, see web page)
Abstract below
WEDNESDAY, 12 MARCH 2003
3:00pm Reuters Foundation Digital Vision Program Seminar
email for location mailto:digitalvision@csli.stanford.edu
Hossein Eslambolchi
President, AT&T Laboratories
http://reuters.stanford.edu/seminar_speakers.html
3:45pm Psychology Department Colloquium
Jordan Hall 420:041
Title to be announced
David Amaral
UC Davis
http://www-psych.stanford.edu/news.html
4:15pm EE380: Computer Systems Lab Colloquium
Gates B03 (NEC Auditorium)
"Bytecode-to-bytecode adaptive optimization for Smalltalk
Compilation and execution architecture for late-bound
object-oriented programming languages"
Eliot Miranda
Visualworks Engineering, Cincom
http://ee380.stanford.edu/contents.html
5:30pm Syntax Workshop
Margaret Jacks 460, Terrace Room
"Why Are Natural Languages So Ambiguous?"
Tom Wasow
Stanford University
http://www-linguistics.stanford.edu/sssg/
Abstract below
THURSDAY, 13 MARCH 2003
12 noon CSLI CogLunch
Cordura Hall, Room 100
"Constructive perception: a skill for coordinating perceptual
discovery and conceptual generation"
Masaki Suwa
http://www-csli.stanford.edu/events/Coglunch/
12:15pm Stanford Networking Seminar
Gates 104
Title to be announced
Victor Bahl
Microsoft
http://netseminar.stanford.edu/
4:00pm Personality Seminar
Jordan Hall 420:100
Anda Gershon and Ying Wong,
Stanford
http://www-psych.stanford.edu/news.html#person_lab
4:00pm SRI AI Seminar Series
EJ228, SRI International
"Conflict Resolution Strategies and Their Performance Models
for Large-Scale Multiagent Systems"
Hyuckchul Jung
University of Southern California
http://www.ai.sri.com/seminars/
Abstract below
4:00pm UC Berkeley CIS Seminar
Soda Hall 380 (UC Berkeley)
"Computer Assisted Forensic Identification"
Gene Myers
UC Berkeley
http://www.cs.berkeley.edu/~eyal/cis-seminar
4:15pm CSLI Seminar on Computational Learning and Adaptation
Cordura Hall, room 100
"Mining Molecular Fragments:
Finding Relevant Substructures in Sets of Molecules"
Michael R. Berthold
Tripos Data Analysis Research Lab
http://cll.stanford.edu/scla.html
Abstract below
4:15pm SSP10: Symbolic Systems Forum
Bldg. 380:380C (Math Corner)
"Empirical Investigations in Social Choice Theory"
Raja Shah
Symbolic Systems, Stanford
http://symsys.stanford.edu:8081/ssp-dynamic/servlet/ssp_events
Abstract below
4:15pm Neurobiology of Disease Seminar Series
Munzer Auditorium, Beckman B060
"Modeling Human Neurodegenerative Diseases in Drosophila"
Mel Feany
Harvard Medical School
http://www-med.stanford.edu/sbrc/calendar/
FRIDAY, 14 MARCH 2003
all day Fourth Annual Semantics Fest
Cordura 100
"The Construction of Meaning"
http://www-linguistics.stanford.edu/semgroup/semfest.html
11:00am Berkeley Institute of Cognitive and Brain Seminar
Tolman 5101 (Berkeley)
"Language learners and Language change:
The regularization of inconsistencies by adults and children"
Carla Hudson
Psychology, UC Berkeley
http://socrates.berkeley.edu:4247/seminar.html
12:30pm CS547: Human-Computer Interaction Seminar
Gates B03
"Technology Demonstrations: What are they for?"
Wally Smith
Edith Cowan University, Western Australia
http://www-pcd.stanford.edu/cs547/
Abstract below
3:15pm Philosophy Department Colloquium
Bldg. 90:92Q
"Qualia and Intentionality"
Brian Loar
Rutgers University
http://philosophy.rutgers.edu/FACSTAFF/BIOS/loar.html
http://www-philosophy.stanford.edu/ce.html
4:15pm CS545: Database Seminar
Gates B12
"Thinking Relational: Using SQL for spatial data access the way god
intended-- it's sets stupid!"
Jim Gray
Microsoft Research
http://www-db.stanford.edu/dbseminar/
____________
Stanford Blood Center status: critical shortage of O-; shortage of A-,
A+, and O+. For an appointment: http://bloodcenter.stanford.edu/ or
call 650-723-7831. It only takes an hour of your time.
____________
MEDIA-X ANNOUNCEMENT
Media X funds ten new research projects
Media X is pleased to announce the funding of ten new, year-long
research projects, bringing to 16 the total number of projects funded
directly by Media X since its launch last April.
Two campus-wide requests for proposals issued over the Christmas period
generated a total of 38 submissions, from which faculty committees
selected ten for funding.
"The number of submissions far exceeded our current level of funding,"
commented Keith Devlin, a member of the Media X Executive Committee,
which overseas the program. "Media X is funded entirely from the fees
paid by industrial partners to be members of Media X. Our development
plans call for a fairly aggressive membership drive to secure many
more industry members. When we reach our target, we will be able to
fund many more projects than at present. Right now, however, with the
program still in its early development stage, we are having to say no
to many excellent proposals."
As part of its efforts to secure industry partners, last fall Media X
hired a full time Director for its Industry Partners Program, Todd
Logan. Todd, a veteran Silicon Valley businessman, occupies an office
in the small Media X suite in Wallenberg Hall. He is currently heading
up a drive to secure a total of ten strategic partners, who each
commit $1 million for a five year partnership, and twenty new
affiliates, who pay $50,000 a year for their membership.
The ten newly funded Media X projects are:
Chris Chafe (CCRMA): What's the Delay? Audio Latency and its Effect on
Networked Musical Performance
Parvati Dev & Carla M Pugh (SUMMIT, School of Medicine): Quantitative
Evaluations of Clinical Performance
B. J. Fogg (CSLI) & Byron Reeves (Communication): Persuasive
Narratives Delivered via Mobile Devices
Cliff Nass (Communication): Psychology of Teaching Agents: Designing
Embodied Speaking Agents to Maximize Learning and Engagement
Dan Schwartz (Education): A Teachable Agent for Management Training
Tom Wasow (Linguistics) Emily Bender, Dan Flickinger, Stephan Oepen
(CSLI): Learning English via Robust Conversation
Helmut Krawinkler & Renate Fruchter (Engineering): DIVAS:
Digital-Video-Audio-Sketch * Re-use of Rich Contextual
Gesture-Discourse-Sketch Knowledge
Christopher Manning (CS) & Pat Langley (CSLI): Interactive
Computational Assistants for Video Segmentation and Classification
Roy Pea & Michael Mills (SCIL): Accelerating the Usefulness of Video
Libraries through HyperDiving
Stanley Peters (Linguistics): Structuring Video Content with
Assistance from the Sound Track
____________
STANFORD NETWORKING SEMINAR
on Thursday, 6 March 2003, 12:40pm (lunch 12:15pm)
Gates 104
http://netseminar.stanford.edu/
"Touch-Based Personal Area Networks"
Kurt Partridge
University of Washington
Personal Area Networks (PANs) are useful for many new mobile computing
applications including object identification, context awareness, and
appliance customization. But the several-meter range of today's local
wireless technologies, such as 802.11 and Bluetooth, leads to
ambiguities when connecting wearable devices with ubiquitous devices
in multi-user and multi-object environments. Intrabody communication,
a technology that uses low-frequency, low-power electric fields and
the conductive nature of the human body, could solve this problem by
only allowing communication between two simultaneously touched
devices. I will describe the implementation and performance of our
Intrabody Communication system and its applicability to ubiquitous
computing applications. This is joint work with Gaetano Borriello and
Sarah Newman.
About the speaker: Kurt Partridge is a PhD candidate in the University
of Washington's Computer Science and Engineering Department. His
research interests include technology to support novel user
interaction mechanisms for wearable and ubiquitous computing
devices. He received a BS in computer science from the University of
California at Berkeley and an MS in computer science from the
University of Washington.
____________
BERKELEY INTERNATIONAL COMPUTER SCIENCE INSTITUTE
on Thursday, 6 March 2003, 1:00pm - 2:00pm
Room 607, ICSI, 1947 Center Street, Sixth Floor, Berkeley
http://www.icsi.berkeley.edu/talks/
"Phenotropics, or Prospects for Protocol-adverse Computing"
Jaron Lanier
Whenever we pass a variable to a function, or send a message to an
object, we're simulating the sending of pulses down a wire. The way
that works is the sender and receiver agree in advance on a format
that makes the pulses interpretable, also known as a protocol.
Protocols aren't the only way information can travel between places,
however. When a physical coffee mug sits on a table, it's possible to
imagine that there's a protocol that exists between the two things,
but it's an awkward way to think. And yet that's what we often do
when we try to build scalable simulations of the world. We can end up
with a "coffee mug module" connected to a "table module" via a
protocol. In the early years of computing, many researchers wished
that the world was a little more like a protocol, so that would be
easier to interface computers to it. Early natural language
researchers, for instance, were unhappy to find that it wasn't so.
What happened instead was that processors eventually became powerful
enough to run pattern classification algorithms that could gather
information even though the world didn't agree with us in advance on a
format. Some examples are face recognition and feature tracking,
voice recognition, and scene understanding. The idea of phenotropics
is to use similar pattern recognition techniques to connect software
modules together inside the computer. Hopefully systems built in that
way will display more informative failure modes, and therefore be more
amenable to adaptive improvement. Another potential benefit is that
scientific simulations might not be distorted by protocols (as in the
example of the coffee mug on the table), and might be more easily
integrated into a new iteration of the scientific method in which they
could be usefully published, tested, and reused. A potential early
application in surgical simulation will be discussed.
About the speaker: Jaron Lanier is a computer scientist best known for
coining the term "Virtual Reality". Most recently he served as Chief
Scientist for Advanced Network and Services, the engineering office of
Internet2, a coalition of American research universities sharing an
experimental next generation network. While there, he lead the Nation
Tele-immersion Initiative, which was responsible for providing the
"driver" applications for Internet2. Tele-immersion is an extension
of Virtual Reality, in which people in different cities are given the
illusion that they are in the same room.
____________
UC BERKELEY CIS SEMINAR
on Thursday, 6 March 2003, 4:00pm-5:30pm
Soda Hall 380 (UC Berkeley)
http://www.cs.berkeley.edu/~eyal/cis-seminar
"Being There:
Capturing and Rendering the Realistic Appearance of the Visual World"
Richard Szeliski
Microsoft Research
The struggle to capture the lifelike appearance of the visual world
has long preoccupied artist, photographers, cinematographers, and
special effects supervisors. The advent of computers has made it
possible to create interactive experiences that give an even greater
sense of "being there" through the use of navigation, viewpoint
control, and interaction with animated objects. Image-based modeling
and rendering are now commonly used tools to facilitate the creation
and display of photorealistic models, but are commonly limited to
static scenes. The ability to process video to create dynamic visual
experiences, i.e., video-based rendering is the next frontier.
In this talk, I review a number of image-based modeling and rendering
systems and discuss some of the representations and estimation
algorithms that they use. I also present our work in video-based
rendering, in which we synthesize novel video from short sample clips
by discovering the hidden (quasi-repetitive) temporal structure.
Image-based and video-based rendering can be combined to create
compelling interactive photorealistic experiences. I demonstrate some
of these experiences with a particular focus on visiting remote sites
of interest, i.e., Virtual Tourism.
About the Speaker Richard Szeliski is a Senior Researcher in the
Vision Technology Group at Microsoft Research, where he is pursuing
research in 3-D computer vision, video scene analysis, and image-based
rendering. His current focus is on constructing photorealistic 3D
scene models from multiple images and video. He received a
Ph.D. degree in Computer Science from Carnegie Mellon University,
Pittsburgh, in 1988. He joined Microsoft Research in 1995. Prior to
Microsoft, he worked at Bell-Northern Research, Schlumberger Palo Alto
Research, the Artificial Intelligence Center of SRI International, and
the Cambridge Research Lab of Digital Equipment Corporation.
Dr. Szeliski has published over 100 research papers in computer
vision, computer graphics, medical imaging, and neural nets, as well
as the book Bayesian Modeling of Uncertainty in Low-Level Vision. He
was a Program Committee Chair for ICCV'2001, and is on the Editorial
Board of the International Journal of Computer Vision. He has served
as co-chair of the SPIE Conferences on Geometric Methods in Computer
Vision, the 1999 Vision Algorithms Workshop, and as an Associate
Editor of the IEEE Transactions on Pattern Analysis and Machine
Intelligence.
____________
BERKELEY BISC SEMINAR
on Thursday, 6 March 2003, 4:00pm-5:30pm
320 Soda Hall (UC Berkeley)
http://buffy.eecs.berkeley.edu/Seminars/
"From Search Engines to Question-Answering System --
The Need for New Tools"
Lotfi A. Zadeh
BISC, CS Division, EECS Dept. UC Berkeley
Search engines, with Google at the top, have many remarkable
capabilities. But what is not among them is the deduction capability -
the capability to synthesize an answer to a query by drawing on bodies
of information which are resident in various parts of the knowledge
base. It is this capability that differentiates a question-answering
system, Q/A system for short, from a search engine. Upgrading a search
engine to a Q/A system is a complex, effort-intensive, open-ended
problem. Semantic Web and related systems for upgrading quality of
search may be viewed as steps in this direction. But what may be
argued, as is done in the following, is that existing tools, based as
they are on bivalent logic and probability theory, have intrinsic
limitations. The principal obstacle is the nature of world knowledge.
The centrality of world knowledge in human cognition, and especially
in reasoning and decision-making, has long been recognized in AI. The
Cyc system of Douglas Lenat is a repository of world knowledge. The
problem is that much of world knowledge consists of perceptions. More
specifically, perceptions are f-granular in the sense that (a) the
boundaries of perceived classes are fuzzy; and (b) the perceived
values of attributes are granular, with a granule being a clump of
values drawn together by indistinguishability, similarity, proximity
or functionality. What is not widely recognized is that f-granularity
of perceptions put them well beyond the reach of computational
bivalent-logic-based theories.
Dealing with world knowledge needs new tools. A new tool which is
suggested for this purpose is the fuzzy-logic-based method of
computing with words and perceptions (CWP), with the understanding
that perceptions are described in a natural language. A concept which
plays a key role in CWP is that of Precisiated Natural Language (PNL).
It is this language that is the centerpiece of our approach to
reasoning and decision-making with world knowledge. A concept which
plays a key role in organization of world knowledge is that of an
epistemic (knowledge-directed) lexicon (EL). Basically, an epistemic
lexicon is a network of nodes and weighted links, with node i
representing an object in the world knowledge database, and a weighted
link from node i to node j representing the strength of association
between i and j. The name of an object is a word or a composite word,
e.g., car, passenger car or Ph.D. degree. An object is described by a
relation or relations whose fields are attributes of the object. The
values of an attribute may be granulated and associated with
granulated probability and possibility distributions. For example,
the values of a granular attribute may be labeled small, medium and
large, and their probabilities may be described as low, high and low,
respectively. Relations which are associated with an object serve as
PNL-based descriptions of the world knowledge about the object. For
example, a relation associated with an object labeled Ph.D. degree may
contain attributes labeled Eligibility, Length.of.study,
Granting.institution, etc. The knowledge associated with an object may
be context-dependent. What should be stressed is that the concept of
an epistemic lexicon is intended to be employed in representation of
world knowledge - which is largely perception-based - rather than Web
knowledge, which is not.
In conclusion, the main thrust of the fuzzy-logic-based approach to
question-answering which is outlined in this abstract, is that to
achieve significant question-answering capability it is necessary to
develop methods of dealing with the reality that much of world
knowledge is perception-based. Dealing with perception-based
information is more complex and more effort-intensive than dealing
with measurement-based information. In this instance, as in many
others, complexity is the price that has to be paid to achieve
superior performance.
____________
CSLI SEMINAR ON COMPUTATIONAL LEARNING AND ADAPTATION
on Thursday, 6 March 2003, 4:15pm to 5:30pm
Cordura 100
http://cll.stanford.edu/scla.html
"Statistical Modeling of Large-Scale Scientific Simulation Data"
Tina Eliassi-Rad
Center for Applied Scientific Computing, Lawrence Livermore
http://www.llnl.gov/CASC/people/eliassi-rad/
Simulations of complex scientific phenomena involve the execution of
large-scale computer programs. These simulations generate massive data
sets over the spatio-temporal space. Unfortunately, the sheer size of
the data has made efficient exploration of them impossible. Therefore,
constructing "queriable" models for such massive data sets is an
essential step in helping scientists discover new information from
their computer simulations.
At the Center for Applied Scientific Computing (CASC), we have
developed an ad-hoc query infrastructure, which reduces the data
storage requirements and query access times by (1) creating and
storing statistical and mathematical models of the data at multiple
resolutions, and (2) evaluating queries on the models of the data
instead of the entire data. In this talk, I focus on the modeling side
of our infrastructure. In particular, I present three simple but
effective statistical modeling techniques for simulation data. The
first modeling technique computes the "true" (unbiased) mean of
systematic partitions of the data. It makes no assumptions about the
distribution of the data and uses a variant of the root mean square
error to evaluate a model. The second statistical modeling technique
uses the Andersen-Darling goodness-of-fit method on systematic
partitions of the data. This method evaluates a model by how well it
passes the normality test on the data. Finally, the third statistical
modeling technique is a multivariate clustering algorithm, which
utilizes the cosine similarity measure to cluster the field variables
in a data set. To scale our multivariate clustering algorithm for
large-scale data sets, we take advantage of the geometrical properties
of the cosine similarity measure. This allows us to reduce the
modeling time from O(n*n) to O(f(u)*n), where n is the number of data
points and f(u) is a function of the user-defined clustering
threshold. We show that on average f(u) is less than n. Our
experimental evaluations on several scientific data sets illustrate
the value of using these statistical modeling techniques on multiple
resolutions of large-scale simulation data sets.
____________
SSP10: SYMBOLIC SYSTEMS FORUM
on Thursday, 6 March 2003, 4:15pm
Bldg. 380:380C (Math Corner)
http://symsys.stanford.edu:8081/ssp-dynamic/servlet/ssp_events
"Knowing Your Own Mind"
Krista Lawlor
Philosophy, Stanford
http://www.stanford.edu/~klawlor/
It can seem that knowing your own mind is the easiest thing in the
world to do. Are you hungry or not? Do you want to go to class this
afternoon, or does something else seem more appealing to you? These
questions are easily answered. Even when such questions are not easily
answered, it seems that you're the one to do the answering. Who else
could? You have a kind of authority when you report your attitudes,
emotions, pains and cravings, that others just cannot have, no matter
how good they are at knowing you. First-person authority is a robust
phenomenon, and it survives the recognition that we are fallible.
What is the source of first-person authority? How far does it really
extend? Just how much fallibility does it survive? I'll discuss one
recent philosophical theory of the source of first-person authority,
and some psychological evidence that casts doubt on the theory.
About the speaker: Professor Lawlor got her Ph.D. from the University
of Michigan in 1999, with a thesis on philosophical semantics. She
likes to think about issues in the philosophy of mind, like how mental
representations get their content, and how we know that content. She
also has a side-interest in epistemology, especially skepticism and
the epistemology of inference.
____________
LINGUISTICS DEPARTMENT COLLOQUIUM
on Thursday, 6 March 2003, 5:30pm
Margaret Jacks Hall, Room 460:126
http://www-linguistics.stanford.edu/colloq/
"Against iconicity and markedness"
Martin Haspelmath
Max-Planck-Institut fur evolutionare Anthropologie, Leipzig
The notions of iconicity and markedness have played important roles in
a diverse range of theoretical approaches, and as a result, the terms
have become multiply polysemous, in a way that is not often
acknowledged by theoretical linguists. While iconicity has most often
been appealed to by functionalists, the term markedness is employed
very widely.
In this talk, I argue that the two most commonly invoked kinds of
iconicity (iconicity as correspondence of complexity/markedness, e.g.
Givon 1995, Aissen 2003+, and iconicity of cohesion, e.g. Haiman 1983)
are unnecessary, because where they make right predictions, these
predictions also follow from the predictions made by the preference
for usage economy. But in many cases, these iconicity principles make
wrong predictions while economy makes right predictions, so that it is
clear that economy should replace iconicity.
For the term "markedness", I distinguish seven different senses, which
are related by family resemblances: markedness (i) as overt coding,
(ii) as specification for a feature (e.g. Trubetzkoy 1939), (iii) as
restricted cross-linguistic distribution, (iv) as a cluster of
correlating properties of meaningful categories ("typological
markedness", Greenberg 1966), (v) as dispreference for difficult
structures ("unnaturalness", Wurzel 1998), (vi) as rarity or
unexpectedness, (vii) as deviation from the default parameter setting.
I argue that apart from (i), none of these notions is required for an
explanatory approach to the structure of human languages. To replace
them, all that is needed are general processing preferences (economy,
distinctiveness, parsability) as well as general conceptual-pragmatic
preferences which give rise to frequency asymmetries. Finally, I
examine the use of the term "markedness" in the recent
optimality-theoretic literature (e.g. Aissen 1999, Bermudez-Otero &
Borjars 2003+), and I conclude that OT approaches simply adopt earlier
observations about "typological markedness" and "unnaturalness",
without offering a genuinely new perspective.
References
Aissen, Judith. 1999. "Markedness and subject choice in Optimality
Theory." Natural Language and Linguistic Theory 17: 673-711.
Aissen, Judith. 2003+. "Differential object marking: Iconicity vs.
economy." To appear in Natural Language and Linguistic Theory
Bermudez-Otero, Ricardo & Borjars, Kersti. 2003+. "Markedness in
phonology and in syntax: the problem of grounding." To appear in:
Honeybone, Patrick & Bermudez-Otero, Ricardo (eds.), Linguistic
knowledge: perspectives from phonology and form syntax. (Special
issue of Lingua)
Givon, T. 1995. "Markedness as meta-iconicity: distributional and
cognitive correlates of syntactic structure." In: Functionalism and
grammar. Amsterdam: Benjamins, 25-69.
Greenberg, Joseph. 1966. Language universals, with special reference
to feature hierarchies. (Janua Linguarum, Series Minor, 59.) The
Hague: Mouton.
Haiman, John. 1983. "Iconic and economic motivation." Language 59:781-819.
Trubetzkoy, Nikolaj. 1939. Grundzage der Phonologie. Guttingen:
Vandenhoeck & Ruprecht.
Wurzel, Wolfgang Ullrich. 1998. "On markedness." Theoretical
Linguistics 24.1: 53-71.
____________
LOGICAL METHODS IN THE HUMANITIES
on Friday, 7 March 2003, 12 noon-1:15pm
Math Corner 380:383N
http://www-logic.stanford.edu/seminars.html
"A User-Friendly Decision Procedure for the Probability Calculus,
with Some Applications to Bayesian Philosophy of Science"
Branden Fitelson
Philosophy Department, SJSU
I will present a simple, Mathematica-based decision procedure for a
rather broad class of arguments in the probability calculus. Some
background on the decision procedure will be presented, and then some
applications of the procedure to Bayesian philosophy of science will
be presented. The relevant background about Bayesianism will be
provided as well. The procedure in question has already been used to
solve dozens of problems in contemporary Bayesian confirmation theory
(these applications were reported in the author's dissertation, and in
several recent publications of the author as well).
____________
CS547: HUMAN-COMPUTER INTERACTION SEMINAR
on Friday, 7 March 2003, 12:30-2:00pm
Gates B03
http://www-pcd.stanford.edu/cs547/
"Challenging the Videoconferencing Gold Standard:
Art Therapy and Tele-Mediated Communication in Behavioral Telehealth"
Kate Collie
University of British Columbia
In the world of behavioral telehealth, high-quality video-conferencing
is usually assumed to be the best way for a psychiatrist or therapist
in one location to communicate with a patient or client in another
location. This assumption is being challenged by work being done at
the University of British Columbia (UBC) by an interdisciplinary team
of researchers, led by Kate Collie, who are looking at
Internet-supported art therapy as a form of psychotherapy that may be
particularly well suited for distance delivery when the modes of
communication are speech and hand-drawn images.
Art therapy is psychotherapy that employs non-verbal communication and
creative expression. Among other things, it is used with both children
and adults to treat the effects of trauma, including trauma associated
with life-threatening illness. The UBC team's research is aimed at
using the Internet to make art therapy services more available to
people with mobility limitations due to illness or disability. The
research has included the creation of customized audiographic software
for distance art therapy and the development of communication
protocols for online therapy. Although the modes of distance
communication (speech and hand-drawn images) were originally chosen
because of their low bandwidth requirements and the possibility of
reaching people in their homes this way, the research has highlighted
additional advantages (over video-conferencing) of using audiographic
systems for distance psychotherapy.
In this presentation, Kate will give an overview of the work being
done at UBC as a starting point for discussing research from a wide
range of fields in which modes of mediated and non-mediated
communication have been compared on measures related to task
completion and relationship formation. She will highlight key
differences between visual and verbal, written and spoken, and
synchronous and asynchronous communication that have been identified,
and examine why video-conferencing is often preferred, in spite of the
demonstrated superiority of audio-only communication for most types of
distance interpersonal interactions.
About the speaker: Kate Collie is doctoral candidate at the Institute
of Health Promotion Research at the University of British
Columbia. She is an artist who has exhibited nationally and
internationally and an art therapist with a specialty in trauma and
major physical illness. For the last five years, she has been
conducting research about Internet art therapy-for the purpose of
expanding access to this kind of service, particularly to people with
mobility limitations due to illness or disability. This telehealth
research has included examinations of strengths and weaknesses of
different forms of tele-mediated communication.
____________
NLP READING GROUP
on Friday, 7 March 2003, 2:15pm
Math Corner, 380:383P
http://www-nlp.stanford.edu/nlp/nlpgroup.html
"The Limits of N-Gram Translation Evaluation Metrics"
Susanne Riehemann and Chris Culy
SRI
N-gram measures of translation quality, such as BLEU and the related
NIST metric, are becoming increasingly important in machine
translation, yet their behaviors are not fully understood. In this
paper we examine the performance of these metrics on professional
human translations into German of two literary genres, the Bible and
Tom Sawyer. The most surprising result is that some machine
translations out-score some professional human translations. In
addition, it can be difficult to distinguish some other human
translations from machine translations with only two reference
translations; with four reference translations it is much easier. Our
results lead us to conclude that much care must be taken in using
n-gram measures in formal evaluations of machine translation quality,
though they are still valuable as part of the iterative development
cycle.
Note that there will be no nlp reading group on March 14 due to
SemFest, but there will be a special spring break meeting with a
presentation by Effi Georgala.
____________
SRI AI SEMINAR SERIES
on Monday, 10 March 2003, 11:00am
EJ228, SRI International
http://www.ai.sri.com/seminars/
"On the utility of conceptual semantics
in translating event descriptions across languages"
Robert Belvin
A problem for any automated language translation system is handling
translation divergences. A typical example of a divergence occurs when
something that is expressed by one word in the source language can
only be expressed by a phrase in the target language. In this talk we
will look at a particular kind of divergence, which arises when an
event description in one language will not submit to a very literal
translation, because the event concepts are not packaged into words in
the same way in the two languages. I will briefly describe event
decomposition in a conceptual semantic framework, and show how it
provides an apparatus that can exploit systematic differences in how
languages partition the conceptual elements of an event into lexical
units. In this way, one is spared having to specify transfer rules
between the languages for individual phrasal constructions. The
downside of this approach is that it suffers from the knowledge
acquisition bottleneck problem; in order to make use of conceptual
semantic representations, one has to acquire them in large numbers. I
will discuss methods which may allow for at least the partial
automation of acquiring these representations. I will also discuss how
this framework and attendant formalism may be useful as a way of
augmenting language models (e.g. for use in a speech recognizer).
Note for Visitors to SRI: Please arrive at least 10 minutes early in
order to sign in and be escorted to the conference room. SRI is
located at 333 Ravenswood Avenue in Menlo Park. Visitors may park in
the visitors lot in front of Building E, and should call extension
2592 to be escorted to the meeting room.
____________
BERKELEY LINGUISTICS DEPARTMENT COLLOQUIUM
on Monday, 10 March 2003, 5:00pm
182 Dwinelle (Berkeley)
http://linguistics.berkeley.edu/lingdept/Current/events.html
"Ditransitive constructions in the world's languages:
alignment types, alignment splits, and inverse patterns"
Martin Haspelmath
Max Planck Institute for Evolutionary Anthropology, Leipzig
In this talk I report on an ongoing cross-linguistic study of
ditransitive constructions, based on over 200 languages. By
"ditransitive constructions", I mean constructions of three-place
verbs taking an agent, a theme and a (macro-)recipient (including
addressee and benefactive) argument. I start with Dryer's (1986)
observation that the grammatical behavior of recipient and theme is in
many ways analogous to the behavior of (macro-)agent and
(macro-)patient in monotransitive constructions, and I pursue the
analogy further.
The basic alignment types are defined by the argument-marking
patterns, i.e. case-marking and indexing ("agreement") patterns (word
order is largely ignored). Corresponding to the basic monotransitive
types (accusative, neutral, ergative), there are three ditransitive
types: indirective (treating theme like patient), neutral (treating
both theme and recipient like patient), and sec=FAndative (treating
recipient like patient). A further logically possible type (neither
theme nor recipient treated like patient) is unattested. All types are
found both in case-marking and indexation, but case-marking heavily
favors indirective alignment, whereas indexation favors sec=FAndative
alignment. I will discuss explanations for the correlations, and I
will show a world map of the different patterns, demonstrating that
their geographical distribution is far from random. (This is based on
a map forthcoming in the World Atlas of Language Structures, Dryer et
al. (to appear).)
Like monotransitive constructions, ditransitive constructions
sometimes show animacy-based alignment splits. For instance, in Yimas
and French first and second person pronouns show neutral alignment,
whereas third person NPs show indirective alignment. Even more common
are alignment splits depending on lexical classes of verbs, but other
types of split which are attested in monotransitives (conditioned by
tense/aspect or subordination) do not seem to occur. Again, I ask
whether explanations proposed for monotransitive alignment splits can
be extended to.
____________
LINGUISTICS DEPARTMENT COLLOQUIUM
on Tuesday, 11 March 2003, 12 noon
Margaret Jacks Hall, Room 460:126
http://www-linguistics.stanford.edu/colloq/
(note special day and time)
"Quantifying Ambiguity Resolution with Information Theory"
John Hale
Johns Hopkins University
Ambiguities about grammatical category and syntactic structure
permeate natural language. Explaining human comprehenders'
performance in the face of such confusion has been called the central
problem in sentence processing (Tabor & Tanenhaus, 2001). How is it
that human sentence understanders are able to recognize combinatory
relationships, from an infinite range of possibilities, to arrive at a
meaningful interpretation of a sentence?
This talk argues that an answer lies in formalizing the idea that
comprehenders search the space of grammatical analyses in a way
constrained by the words they hear. Comprehenders are constantly
engaged in ambiguity resolution, and the more ambiguity is resolved,
the longer they take.
To make this intuition fully explicit, ambiguity resolution will be
given a precise interpretation in terms of information theory.
The general theory is tested using explicit grammar fragments that are
probabilistic versions of Generalized Phrase Structure Grammars
(Gazdar, Klein & Sag 1985) and Minimalist Grammars (Stabler 1997).
The theory will be shown to derive a range of well-documented
processing phenomena including garden-path sentences,
center-embedding, and the Accessibility (or Obliqueness) Hierarchy of
relativized grammatical functions.
____________
BERKELEY BISC DISTINGUISHED LECTURE SERIES
on Tuesday, 11 March 2003, 4:00pm - 5:30pm
373 Soda (Berkeley)
http://buffy.eecs.berkeley.edu/Seminars/
"Concept-Based Information Retrieval and Search Engine"
Tomohiro Takagi
Computer Science Department, Meiji University, Japan
Since a fuzzy set is defined by enumerating its elements and the
degree of membership of each element, we can use it to express word
ambiguity by enumerating all possible meanings of a word, then
estimating the degrees of compatibilities between the word and the
meanings.
Based on this approach, we have proposed using conceptual fuzzy sets
(CFSs) to represent the various meanings of a concept that change
dynamically depending on the context. A CFS (is realized as neural
networks in which a node represents a concept and a link represents
the strength of the relation between two (connected) concepts. The
activation values agreeing with the grades of membership are
determined through this associative memory. In a CFS, the meaning of a
concept is represented by the distribution of the activation values of
the other nodes. The distribution evolves from the activation of the
node representing the concept of interest.
This talk will start with my motivation to propose CFSs and algorithm
to generate CFSs. It will describe how it works to represent the
context dependent meaning of a word and to measure a conceptual
distance between documents. Next, information filtering and image
search (Google-Based Search Engine for Multimedia Data) will be
introduced as its applications to information retrieval using
capability of conceptual matching. Finally we will introduce our
approach to enhancing CFSs based on brain architecture.
____________
COMPUTER SCIENCE FACULTY CANDIDATE TALK
on Tuesday, 11 March 2003, 4:15pm - 5:15pm
Packard EE 202
"Query Processing in Sensor Networks"
Sam Madden
UC Berkeley
Many of the emerging applications for sensor networks are focused on
data collection and monitoring in remote environments. Unfortunately,
existing tools for building such applications require users of these
networks, who often aren't trained computer scientists, to write
low-level, embedded C code. Such deployments frequently become mired
in the difficulties of coding power-management, routing, and storage
features in these volatile distributed environments.
In this talk, I will discuss how many of these difficulties can be
overcome by providing users with a simple declarative interface where
short, SQL-like queries are pushed into the network. Such queries
concisely express a user's data needs, freeing him or her from the
details of implementation and execution. In additional to
dramatically simplifying the task of sensor-network programming, this
approach enables the system to transparently optimize in-network query
execution to minimize overall power consumption in ways that even
sophisticated programmers may miss.
I will summarize the query processing features of TinyDB, a query
processor for sensor networks we have developed at Berkeley, focusing
on a framework for executing and optimizing aggregation queries. I
will discuss current deployments that are underway at Berkeley, along
with new features that are being incorporated to accommodate these
deployments. I will include a brief demonstration of the system.
About the speaker: Sam Madden is a 4th year Ph.D. student in the
Database Group in the Computer Science Department at the University of
California, Berkeley. He received his M.Eng. and B.S. from MIT in
1999.
____________
SNRC INDUSTRY SEMINAR
on Tuesday, 11 March 2003, 4:15pm
Packard 101
http://snrc.stanford.edu/events/industry-seminar/
"IPv6 Networks and P2P Solutions"
Kazuho Miki
Director & Lab Manager, Hitachi America Ltd.
IPv6 is one of promising technologies for the next generation
network. It solves many current and future issues such as address
shortage that becomes matter of concern especially in Asia. The future
need for IPv6 has become gradually recognized, and the present is the
first stage of commercial introduction and deployment. IPv6 is a
technology holding the potential for a wide reaching influence because
of its Peer-to-peer (P2P) characteristic, so there could be many new
P2P solutions based on IPv6 networks.
About the speaker: Kazuho Miki is Director & Lab Manager of Network
Systems Research Lab of Hitachi America Ltd. in Brisbane,
California. The lab focuses on researching core technologies for Next
Generation Network Systems and Solutions. Previously, he was a
researcher of Central Research Lab in Hitachi Ltd. in Tokyo, Japan. He
has a Master of Engineering in Electronics and Communications from
Waseda University in Tokyo, Japan.
____________
EMERGING TECHNOLOGY GROUP
on Tuesday, 11 March 2003, 6:30pm
Cubberley Community Center, H-1, 4000 Middlefield, Palo Alto
http://www.sdforum.org/
(there is a fee for non-SDForum members, see web page)
"Fundamental Problems in Interactive Storytelling"
Chris Crawford
Interactive Storytelling presents us with a challenge going far beyond
the problems we encounter in games. Three fundamental problems
obstruct our efforts to create genuine interactive storytelling
technology. First, we must integrate all three primary cultures (art,
science, and business) into a unified whole capable of building
interactive storytelling systems, for all three cultures are necessary
to the creation of a viable interactive storytelling
technology. Second, we must concentrate our attention on the verb set
rather than the environment or the characters; as Aristotle noted,
character is revealed through the choices that characters make, so
without a large palette of such choices, we cannot reveal much in the
way of character. Third, we must devise tools that artists can use to
create interactive storyworlds. Whatever technologies we build for
interactive storytelling, they will be large and complex; without
well-designed tools, artists will be unable to cope with the immense
complexity of such products.
About the Speaker: Chris Crawford started out writing games for the
Atari Home Computer System in the early 1980's. After writing such
classics as Eastern Front (1941) and Legionairre, he founded the Games
Research Group. While there, he wrote The Art of Computer Game Design
and created Excalibur, a game about the Arthurian legends.
He continued to design and create games throughout the 1980s and early
1990s (Balance of Power, Patton Versus Rommel, Trust & Betrayal, Guns
& Butter, Balance of the Planet, and Patton Strikes Back). Along the
way he also wrote two more books about game design (most recently
Understanding Interactivity), founded The Journal of Computer Game
Design, and started the Computer Game Developer's Conference.
His current project, now nine years in the making, is a technology for
interactive storytelling and a development environment that permits
nontechnical artists to control the technology.
____________
SYNTAX WORKSHOP
on Wednesday, 12 March 2003, 5:30pm
Margaret Jacks Hall, fourth floor, Terrace Room
http://www-linguistics.stanford.edu/sssg/
"Why Are Natural Languages So Ambiguous?"
Tom Wasow
Stanford University
Every explicit theory of language processing, both psycholinguistic
and computational, entails that ambiguity makes processing harder.
This makes intuitive sense, since disambiguation is an additional
task, and ambiguity increases the probability of misinterpretation.
For this reason, the artificial languages of mathematics and computer
science are unambiguous by design. Yet natural languages are massively
ambiguous, exhibiting frequent lexical, syntactic, and scopal
ambiguities.
Psycholinguistic experiments and corpus studies provide little
evidence that speakers use the grammatical resources of their
languages to avoid ambiguity. My investigations of heavy NP shifting
and the dative alternation in English indicate that ambiguity
avoidance is not a significant factor in the choice of one constituent
ordering over the other. This is surprising, in light of the
widespread agreement that ambiguity impedes processing.
There are a number of explanations that might be given for the
ambiguity of natural language. One might deny the premise, pointing
out that people rarely notice any but the intended interpretations.
But this just begs the question: how do people do it? What makes us so
good at disambiguation, when all of our models of language processing
make it seem so hard? Perhaps ambiguous languages have some
advantages. For example, they might allow more concise expression,
helping us to overcome what Levinson calls "the
bottleneck...constituted by the remarkably slow transmission rate of
human speech". And they might convey certain communicative advantages,
such as facilitating interactions with people speaking different
dialects, or permitting speakers to convey different messages to
different audiences with the same utterance.
Speculations like these suggest that the pervasiveness of ambiguity in
natural language may have certain benefits, as well as its well-known
costs. I am part of a team that has begun investigating these
trade-offs through computational modeling of the evolution of
language. We represent the language of an individual as a matrix, with
symbols on one dimension and their denotations on the other; in each
cell, we put the probability that that symbol will be used to denote
that meaning. A fully unambiguous language would have only ones and
zeros in the cells, and degree of ambiguity can be measured by how
closely a matrix approaches that limit. Our simulations involve
interactions among individuals, followed by adjustments of the values
of the probabilities in the cells, based on the success or failure of
the interactions. We run populations of such matrices through
thousands of generations of interactions to see what happens to the
ambiguity of their language. Various parameters can be adjusted to
simulate some of the costs and benefits of ambiguity. These studies
are still in their early stages, but they have already yielded some
interesting results.
____________
SRI AI SEMINAR SERIES
on Thursday, 13 March 2003, 4:00pm - 5:30pm
EJ228, SRI International
http://www.ai.sri.com/seminars/
"Conflict Resolution Strategies and Their Performance Models for
Large-Scale Multiagent Systems"
Hyuckchul Jung
USC
Distributed, collaborative agents are promising to play an important
role in large-scale multiagent applications. Such collaborative agents
may enter into conflicts over shared resources, joints plans, or
tasks, e.g., one domain motivating our work is distributed sensor
networks where agents must resolve conflicts over shared sensors.
While efficient conflict resolution techniques to enable fast
convergence to a solution are required in a large-scale multiagent
system, conflict resolution strategies in such a system have not been
systematically investigated. In this talk, we first introduce conflict
resolution strategies based on the notion of local cooperativeness
which is defined by the flexibility given towards neighboring agents.
We formalize a set of cooperative conflict resolution strategies in
the context of DCSP (distributed constraint satisfaction problem), and
investigate the performance of the strategies with systematic
experiments. Second, we introduce formal models to analyze the
performance of the conflict resolution strategies so as to predict the
right strategy to adopt in a given domain. In general, performance
modeling has not received significant attention in the multiagent
literature. Our performance models are based on a distributed POMDP
(partially observable Markov decision process) framework. To address
scale-up issues in a large-scale multiagent system, we introduce
small-scale models called "building blocks" that represent the local
interaction among a small group of agents, and investigate several
ways to combine the building blocks. By modeling and combining
building blocks, we are able to predict the performance of different
conflict resolution strategies.
About the Speaker: Hyuckchul Jung received his B.S. degree and
M.S. degree in Computer Science from Seoul National University, Korea,
in 1995 and 1998 respectively. He is currently a Ph.D. student at the
Computer Science department of the University of Southern California,
and is expected to defend his dissertation titled "Conflict Resolution
Strategies and Their Performance Models for Large-Scale Multiagent
Systems" in Summer 2003, under the supervision of Prof. Milind Tambe.
Note for Visitors to SRI: Please arrive at least 10 minutes early in
order to sign in and be escorted to the conference room. SRI is
located at 333 Ravenswood Avenue in Menlo Park. Visitors may park in
the visitors lot in front of Building E, and should call extension
2592 to be escorted to the meeting room.
____________
CSLI SEMINAR ON COMPUTATIONAL LEARNING AND ADAPTATION
on Thursday, 13 March 2003, 4:15pm to 5:30pm
Cordura 100
http://cll.stanford.edu/scla.html
"Mining Molecular Fragments
Finding Relevant Substructures in Sets of Molecules"
Michael R. Berthold
Tripos Data Analysis Research Lab
Joint work with Christian Borgelt and Heiko Hofer
We present an algorithm to find fragments in a set of molecules that
help to discriminate between different classes of, for instance,
activity in a drug discovery context. The resulting molecular
fragments are connected subgraphs that appear frequently in molecules
of the class of interest and appear less frequent in molecules of
other classes. The presented method is based on an algorithm from
association rule mining and a new approach to generate fragments by
embedding them in all appropriate molecules in parallel, which results
in substantially faster searches by eliminating the need for frequent,
computationally expensive re-embeddings. In addition, the search tree
is pruned based on a local ordering of atoms and bonds, which reduces
the number of redundant fragments drastically and makes this type of
search feasible also for real world data sets.
In this talk I will present the underlying algorithm, explain how
incorporation of expert knowledge about chemical structures results in
a few extensions that result in further speed-ups, and illustrate the
usefulness of the new algorithm by demonstrating the discovery of
activity-related groups of chemical compounds in the National Cancer
Institute's HIV-screening dataset.
____________
SSP10: SYMBOLIC SYSTEMS FORUM
on Thursday, 13 March 2003, 4:15pm
Bldg. 380:380C (Math Corner)
http://symsys.stanford.edu:8081/ssp-dynamic/servlet/ssp_events
"Empirical Investigations in Social Choice Theory"
Raja Shah
M.S. Candidate, Symbolic Systems Program
Social choice theory is the formal theory of democracy, and was given
a mathematical footing in Kenneth Arrow's celebrated "impossibility
theorem" (1950). Arrow showed that no procedure for making collective
choices -- other than *dictatorship* -- could satisfy a small set of
reasonable-looking criteria. I will present a set of experiments I
have conducted with Todd Davies, which test whether people -- when
deciding what is best for a group -- act in accordance with one of
Arrow's key criteria: the "independence of irrelevant alternatives"
(IIA). This condition requires that all of the information necessary
for a collective choice between two alternatives must be contained in
individuals' pairwise preferences among those two options, so that
their relative rankings with respect to other alternatives cannot have
any influence. The need to impose IIA has been called into question
on philosophical grounds by various authors, and has recently come
under renewed fire. Arrow himself once allowed that IIA is "stricter
than desirable" (1967). Our findings indicate that subjects violate
IIA and the related criterion known as "regularity" (which implies
IIA) in very strong numbers, even when the presentation format only
shows pairwise individual preferences. One critique of Arrow has been
that his philosophical arguments for IIA are actually arguments for
another criterion that does not lead to impossibility. We define this
weaker notion as "independence of unavailable alternatives" (IUA). A
set of studies still in progress is examining whether IUA has more
intuitive appeal than IIA itself. I will discuss the relevance of
experiments such as ours to the search for defensible democratic
principles.
____________
CS547: HUMAN-COMPUTER INTERACTION SEMINAR
on Friday, 14 March 2003, 12:30-2:00pm
Gates B03
http://www-pcd.stanford.edu/cs547/
"Technology Demonstrations: What are they for?"
Wally Smith
Edith Cowan University, Western Australia
On the surface, a demonstration of information technology is a
straightforward case of 'look before you buy'. However, the
demonstration is markedly different from most other parts of systems
development and procurement. Much of what is transacted between
demonstrator and demonstratee is left implicit, and little is ever
recorded. Demonstrations are generally regarded as important for a
project's success, but reasons given for this vary. At one extreme, a
demonstration is a dry evaluation of requirements, while at another
extreme, it is a drama in which a performer charms, or fails to charm,
his or her audience through a brief glimpse into a technological
potential.
The talk reports an investigation of demonstrations based on
interviews with experienced practitioners in various IT areas. The aim
is to better understand the function of demonstrations and to examine
what makes them a success or failure. Issues raised are the complexity
of organizational communications surrounding demonstrations, the
ever-present potential for trickery and deception, and the consequent
importance placed on trust and understanding between possible future
partners. The analysis draws on a range of informational and
dramaturgical concepts including Erving Goffman's (1974) frame
analysis.
About the speaker: Wally Smith has interests in human-computer
interaction, organizational decision-making and training, and
knowledge management. Recent research has involved the development of
a software tool to assist the design and re-design of disaster
scenarios for training in emergency management. An ongoing project is
investigating the role of technology demonstrations in systems design
and associated decision-making. Related to this project, Wally is an
amateur magician and member of the Magic Circle. He holds a PhD in
Psychology from the University of London and is currently a Senior
Lecturer in Information Systems at Edith Cowan University in Western
Australia. Previously he has held positions at University College
London, City University London and the University of Western
Australia.
____________
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
____________