|
March 24, 2006
13:00-14:00
|
Wei Feng
|
Title Holistic Schema Matching for Web Query Interfaces Abstract
One significant part of today's Web is Web databases, which can dynamically provide information in response to user queries. To help users submit queries to different Web databases, the query interface matching problem needs to be addressed. To solve this problem, we propose a new complex schema matching approach, Holistic Schema Matching (HSM). By examining the query interfaces of real Web databases, we observe that attribute matchings can be discovered from attribute-occurrence patterns. For example, First Name often appears together with Last Name while it is rarely co-present with Author in the Books domain. Thus, we design a count-based greedy algorithm to identify which attributes are more likely to be matched in the query interfaces. In particular, HSM can identify both simple matching, i.e., 1:1 matching, and complex matching, i.e., 1:n or m:n matching, between attributes. Our experiments show that HSM can discover both simple and complex matchings accurately and efficiently on real data sets. |
|
|
March 31st, 2006
13:00-14:00
|
Dr.
Yan Liu
|
Title: Automatic Feature Selection in Video Understanding
Abstract:
The rapid growth and wide applications of digital video data have led to a significant need for video understanding. Unfortunately, the gap between high-level concepts and low-level features and the high time cost of video analysis are two important obstacles of efficient video data management. Feature selection, which can improve the prediction performance of the predictors, provide faster and more cost-effective predictors, and provide better understanding of the data, is introduced to address these two problems. But applying existing automatic feature selection algorithms to video data is impractical because of the unrealistic amount of computer time. So far, most feature selection technologies in video applications are based on researchers' intuition although human interaction can't satisfy the dramatic increase of video data and the multiple requirements of different users.
In this talk, I propose and demonstrate four low time cost feature selection algorithms to address the problem of high dimensionality, complex hypotheses, and sparse training data in classification, with special applications in video segmentation, categorization and retrieval. Compared to existing methods, in both artificial datasets and real problems in several different domains, these methods meet or exceed existing methods' accuracy while running at several orders of magnitude faster. |
Bio: Dr. Liu, Yan is an assistant professor in the Department of Computing at The Hong Kong Polytechnic University. Her research interests include machine learning, multimedia understanding and E-learning. She received her PhD in Computer Science
Department from Columbia University and
her M.S. in Business School from Nanjing University. |
|
April 21st, 2006
13:00-14:00
|
Dr.
Yufei Tao
|
Title:
Privacy Preserving Publication with Generalization
Abstract:
Privacy preservation is a serious concern in publication of personal data, and has attracted considerable attention from the database community. In the first part of the talk, we will review "k- anonymous generalization", an important methodology underlying the existing solutions in the literature. We also explain the drawbacks of this methodology. In particular, k-anonymous generalization exerts the same amount of privacy preservation for all persons, without catering for their concrete needs. The consequence is that we may be offering insufficient protection to a subset of people, while applying excessive privacy control to another subset.
The rest of the talk will be devoted to introducing a novel generalization ramework based on the concept of "personalized anonymity". This technique performs the minimum generalization for satisfying everybody's requirements, and thus, retains the largest amount of information from the microdata. We also present the results of our theoretical study that provide valuable insight into the behavior of alternative solutions. These results mathematically reveal the circumstances where the previous work fails to protect privacy, and establish the superiority of the proposed technique.
The content of the talk will appear in SIGMOD 2006. |
Bio: Dr. Yufei Tao is an Assistant Professor in the Department of Computer Science,
the City University of Hong Kong. He holds a PhD from the Hong Kong University of Science and Technology, and did his post doc at the Carnegie Mellon University, USA. Yufei received the Hong Kong Young Scientist Award in 2002. He enjoys solving database problems, particularly those related to privacy preservation, and spatial/temporal databases. |
|
April 28th, 2006
13:00-14:00
|
David
Yang
|
Title: An Efficient Approach to Support Querying Secure Outsourced XML Information Abstract:
Data security is well-recognized a vital issue in an information system that is supported in an outsource environment. However, most of conventional XML encryption proposals treat confidential parts of an XML document as whole blocks of text and apply encryption algorithms directly on them. As a result, queries involving the encrypted part cannot be efficiently processed. This talk focuses on XQEnc, a novel approach to support querying encrypted XML. XQEnc is based on two important techniques of vectorization and skeleton compression. Essentially, vectorization, which is a generalization ofcolumns of a relational table, makes use the basic path of an XML tree to label the data values. Skeleton compression collapses the redundant paths into a multiplicity attribute. Our analysis and experimental study shows that XQEnc achieves both better query efficiency and more robust security compared with conventional methods. As an application, we show how XQEnc can be realized with relational techniques to enable secure XML data outsourcing. |
|
|
July 28th, 2006
13:00-14:00
|
Prof. Ben Choi
|
Title: Building a Better Search Engine Abstract:
This
presentation will outline several new techniques for building a better search
engine. New algorithms have been developed for
automatically
classifying and clustering Web pages to allow users to quickly find
information on the Web. A new text indexing scheme has
been developed for
speeding up keyword search. And, a new parallel and distributed architecture
has been developed to handle massive
search requests.
Several techniques under development will also be outlined, including Web
crawling methods for finding and updating Web
pages and page
ranking methods for ordering search results. These techniques have been
implemented and tested on an object-oriented
database, the
experience of which will also be shared. In addition, two patent applications
have been filed based on some the new
techniques.
|
Bio: Prof. Ben Choi, Ph.D. & Pilot: He is an Associate Professor in Computer Science at Louisiana Tech University and is also a pilot of airplanes and helicopters. He worked in the computer industry as a System Performance Engineer at Lucent Technologies and as a Computer Consultant. He got his Bachelor, Master, and Ph.D. degrees all from The Ohio State University. His areas were Electrical Engineering, Computer Engineering, and Computer Science. His works included associative memory, parallel computing, and machine learning. His research interests are developing software and hardware methods for building intelligent machines and abstracting the Universe as a Computer. |
|
August 4th, 2006
13:00-14:00
|
Yiping Ke
|
Title:
Mining Quantitative Correlated Patterns Using an Information-Theoretic Approach
Abstract:
Existing research on mining quantitative databases mainly focuses on mining associations. However, mining associations is too expensive to be practical in many cases. In this paper, we study mining correlations from quantitative databases and show that it is a more effective approach than mining associations. We propose a new notion of Quantitative Correlated Patterns (QCPs), which is founded on two formal concepts, mutual information and all-confidence. We first devise a normalization on mutual information and apply it to QCP mining to capture the dependency between the attributes. We further adopt all-confidence as a quality measure to control, at a finer granularity, the dependency between the attributes with specific quantitative intervals. We also propose a supervised method to combine the consecutive intervals of the quantitative attributes based on mutual information, such that the interval combining is guided by the dependency between the attributes. We develop an algorithm, QCoMine, to efficiently mine QCPs by utilizing normalized mutual information and all-confidence to perform a two-level pruning. Our experiments verify the efficiency of QCoMine and the quality of the QCPs. |
|