SUNMONTUEWEDTHUFRISAT
Conferences, Lectures, & Seminars
Events for January
-
CS Colloq: Dr. Kenji Sagae
Thu, Jan 14, 2010 @ 03:30 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Talk Title: Practical analysis of natural language syntax, semantic roles and discourse structure with shift-reduce algorithmsSpeaker: Dr. Kenji SagaeHost: Prof. Kevin KnightABSTRACTAutomatic analysis of the structure of natural language through syntactic parsing techniques has long been considered of great potential value in the study of language, the development of language-enabled systems and interfaces, and the application of language technologies (such as machine translation, question answering and text mining) to the rapidly growing body of information in the form of machine readable text. However, for many years parsing systems suffered from lack of robustness and efficiency to deal with large-scale tasks. Recent research on linear-time parsers that learn from annotated data has opened new possibilities for how these and other issues in practical parsing technologies can be addressed.In this talk I will first present a simple and effective parsing framework that addresses the main challenges in the deployment of parsing technologies in practical tasks. I will show how the combination of machine learning and a parsing approach inspired by Knuth's deterministic LR algorithm produces parsers that are fast, robust and accurate. I will also present extensions of this framework that allow for linear-time analysis of semantic roles and discourse structure, and discuss the application of the resulting data-driven shift-reduce parsing approach in areas as diverse as child language analysis, biomedical text mining, and virtual human dialogue systems.BIOKenji Sagae is a research scientist in the Institute for Creative Technologies at the University of Southern California, where he works on natural language processing for virtual humans and related systems.
Before joining ICT as a research associate in June of 2008, he was a research associate in the Computer Science department of the University of Tokyo, where he worked on the connection of data-driven parsers to theoretically-motivated syntactic models, and the application of natural language processing to information extraction in bioinformatics. He received a PhD in Language Technologies from the School of Computer Science at Carnegie Mellon University in 2006. His dissertation research focused on automatic syntactic analysis of transcripts of dialogues between children and adult caregivers. He is currently the Information Officer for SIGPARSE, the international interest group on parsing technologies, and his parsing software is used by several research groups in the areas of child language and information extraction for biomedical text.Location: Seaver Science Library (SSL) - 150
Audiences: Everyone Is Invited
Contact: CS Front Desk
-
CS Colloq: Dr. Jelena Mirkovic
Tue, Jan 19, 2010 @ 03:30 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Talk Title: Combatting spoofing in a realistic InternetSpeaker: Dr. Jelena MirkovicHost: Prof. John Heidemann ABSTRACT:IP spoofing - forging a sender's IP address - exacerbates many security threats, such as denial of service and intrusions. It is also means for conducting reflector attacks where spoofed service requests lead legitimate servers to swamp the victim with replies.
Although many networks have deployed ingress filtering as means of spoofing prevention, legacy networks can still be used to spoof at will and at large. Six approaches to spoofed packet filtering have been proposed to date. Each shows promise under wide deployment (around 20% of the ISPs) but such deployment is unrealistic. The first part of my talk will tackle the problem of evaluating defense performance under realistic, sparse deployment. I will show that such performance depends strongly on the underlying Internet's topology and routing, and remains fairly constant regardless of the topology/routing sources and evolution trends. This evaluation concludes that three defenses would bring significant spoofing protection to all Internet users, and across multiple dimensions, if deployed systematically at top 18 tier-1 ISPs. Only one defense is effective under isolated deployment, and it only protects against spoofed but not against reflected traffic. The second part of my talk focuses on the three defenses that were effective in our evaluation. Each associates a source with some routing-dependent parameter and uses this information for filtering.
An open research problem is how to learn and update parameter values in presence of asymmetric routing, multipath routing and route changes, all of which are common in today's Internet. I will present our design and evaluation of the Clouseau system, which autonomously harvests the needed information from transit traffic and updates it promptly upon a route change. The information is inferred by filters applying randomized drops to TCP data traffic and observing subsequent retransmissions. No communication is required with packet sources or other filters, which makes Clouseau suitable for partial deployment. NS-2 simulations and experiments with a Clouseau prototype indicate that the operation cost is reasonable, the impact on legitimate traffic is minimal and the inferred information is accurate and robust to attacks by a smart adversary. BIO: Jelena Mirkovic is a Computer Scientist at the USC Information Sciences Institute, which she joined in 2007. Prior to this she was an Assistant Professor at the Computer and Information Sciences Department, University of Delaware, 2003-2007. She received her M.S. and Ph.D. from UCLA, and her B.S. in Computer Science and Engineering from the School of Electrical Engineering, University of Belgrade, Serbia. Her current research is focused on scientific cyber security experimentation, safe sharing of network data, denial-of-service attacks and IP spoofing. Her research is funded by the National Science Foundation, the Department of Homeland Security and the Infosys Corporation.
Location: Seaver Science Library (SSL) - 150
Audiences: Everyone Is Invited
Contact: CS Front Desk
-
CS DLS: Prof. Mary Vernon
Thu, Jan 21, 2010
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Location: SSL 150Title: Quantitative System DesignSpeaker: Prof. Mary Vernon (University of Wisconsin)Hosts: Prof. Leana Golubchik and Prof. Ramesh GovindanAbstract:
This talk will provide a 20-year perspective on the use of analytic models to design of a wide range of commercially important architectures and systems with complex behavior. These systems include resources with highly bursty and/or correlated packet arrivals, communication protocols with complex routing and blocking of messages, resources that are configured for a very high probability (e.g., 0.9999) of providing immediate service to each arriving client, and complex large-scale Grid/Internet applications.
The examples illustrate some guiding principles for model development, and show that the models can be relatively easy to develop.
More importantly, the models can be highly accurate -- often more accurate than simulation, and sometimes more accurate than the system implementation!
The examples also illustrate that the models can provide unique insight into system design as well as significant new system functionality.
In other words, analytic models are a key tool for competitive systems engineering. Time permitting, the talk will include some important observations about workload models, and some ways to avoid key pitfalls in simulation.Bio:
Mary K. Vernon received a B.S. degree with Departmental Honors in chemistry and the Ph.D. degree in computer science from the University of California at Los Angeles.
In 1983 she joined the Computer Science Department at the University of Wisconsin-Madison, where she is currently Professor of Computer Science and Industrial Engineering and Chair of the Computer Sciences Department.
Her research interests include performance analysis techniques for evaluating high performance computer/communication system design tradeoffs, Internet transport protocols, optimized CMP hardware/software co-design, and storage system design. She has co-authored over 80 technical papers including seven award papers - most recently one of three "Fast Track to ToN" papers at Infocom 2004, and the Best Paper Award at the 2005 USENIX Security Symposium. Prof. Vernon has served on the editorial board of the IEEE Transactions on Parallel and Distributed Systems, the 1999 NSF Blue Ribbon Panel for High Performance Computing, the NSF CISE Advisory Board, the CRA Board of Directors, the Board of Directors of the NCSA, and as Chair of the ACM SIGMETRICS. She received the NSF Presidential Young Investigator Award in 1985, the ACM Fellow award in 1996, the UW-Madison Vilas Associate Award in 2000 and the UW-Madison Kellett Mid-career Award in 2006. She is a member of the IFIP WG 7.3 on Information Processing System Modeling, Measurement and Evaluation.Location: Seaver Science Library (SSL) - 150
Audiences: Everyone Is Invited
Contact: CS Front Desk
