ICSI Research Review
Friday, October 12, 2012
2:00 - 5:00pm
Featured talks by ICSI research staff highlighting some of our latest results and new directions in computer science research. Talks will be given in the 6th floor lecture hall.
Agenda:
2:00 Overview of ICSI Research
Dr. Roberto Pieraccini
ICSI Director
2:15 Relationship Between UC Berkeley's EECS Department and ICSI
Prof. David Culler
Chair of EECS, UC Berkeley
2:30 Relationship Between Simons Institute and ICSI
Prof. Richard Karp
Algorithms Group Leader & Simons Institute Director
2:45 Brain Networks
Prof. Eric Friedman
Algorithms Group Member
3:15 Break
3:30 Drinking from a Multimedia Firehose: How to Deal with Billions of Daily Video Posts
Dr. Gerald Friedland
Speech Group Member
4:00 Beyond Technical Security: Developing an Empirical Basis for Socio-Economic Perspectives
Prof. Vern Paxson
Networking Group Member & UC Berkeley Professor
4:30 MetaNet: A Multilingual Metaphor Extraction, Representation, and Validation System
Prof. Srini Narayanan
AI Group Leader & UC Berkeley Professor
Abstracts:
"Brain Networks"
Analyses of connectomes (brain networks) have become an important tool in the understanding and diagnosis of a variety of brain disorders. This talk provides an overview of research in this area, describing the construction of connectomes from MRI scans, their analysis using modern network theory (tools developed for the study of social and computer networks), and their applications to several disorders, including Alzheimer's disease, multiple sclerosis, traumatic brain injuries, and agenesis of the corpus callosum, a birth defect. This talk will also cover some of the challenges and new directions in this area.
"Drinking from a Multimedia Firehose: How to Deal with Billions of Daily Video Posts"
Consumer-produced videos are the fastest-growing type of content on the Internet. YouTube claims that 72 hours of video are uploaded to its Web site alone every minute. Because the videos capture parts of the world, they are potentially useful for qualitative and quantitative empirical research on a larger scale than has ever been possible before. A major prerequisite to making social media videos usable for "field studies" is efficient and unbiased (e.g., keyword-independent) retrieval. More importantly, retrieval needs to go beyond simply finding objects to detecting more abstract concepts, such as “taking care of a car” or “winning a game that is not a card game.” Research on such a large corpus requires the creation of methods that exploit as many cues as possible from different modalities. ICSI has begun using novel acoustic methods to complement computer vision approaches. This talk summarizes ICSI’s research and progress in this area.
"Beyond Technical Security: Developing an Empirical Basis for Socio-Economic Perspectives"
Security is at once a technical property of a system and a socio-economic property of the environment in which it operates. While the vast majority of security research and practice focuses on the first of these, a perspective limited to technical considerations misses an entire half of the problem space: the human element. This talk sketches some recent work exploring security issues from a socio-economic perspective, which highlights both interesting new problems to tackle and the power that such approaches can potentially provide to defenders.
"MetaNet: A Multilingual Metaphor Extraction, Representation, and Validation System"
Metaphors are ubiquitous in language and thought and pose a critical barrier to progress in natural language understanding. This talk describes recent results from a new interdisciplinary project, MetaNet, whose goal is to be able to extract, analyze, interpret, and experimentally validate metaphors in four languages. This project is a multi-year collaborative effort led by ICSI with participants from UC Berkeley, UC San Diego, Stanford, University of Southern California, and UC Merced.