The Department of Computer Science at the University of Cyprus cordially invites you to the Colloquium entitled:

Mining Compressed Web Search Usage Patterns


Speaker: Dr. Michail Vlachos
Affiliation: IBM Research Zurich, Switzerland
Category: Colloquium
Location: Room 148, Faculty of Pure and Applied Sciences (FST-01), 1 University Avenue, 2109 Nicosia, Cyprus (directions)
Date: Monday, November 29th, 2010
Time: 10:30-11:30 EET
Host: Demetris Zeinalipour (dzeina AT

Analysis of historical search patterns holds great importance for web search engines, because it can help to better understand the users' search behavior. Capturing the user search preferences over time can provide useful insights in applications such as discovery of news events, keyword recommendation and personalized ad targeting. A major bottleneck in analyzing historical sequential data is the growing size of data repositories. Therefore, there is a need to enable search and mining operations directly on the compressed data. In this talk we will present how to facilitate efficient search over compressed sequential data, with specific focus on weblog query patterns. Our approach guarantees optimally tight distance bounds, while at the same time being efficient and lightweight. This helps drastically reduce the search time compared to previous state-of-the-art techniques. Additionally, we will explicate how to support other types of knowledge discovery operations, such as burst detection, query-by-burst and query-by-periodicity. We will demonstrate extensions and applications of the presented technique for a multitude of areas.

Short Bio:
Michalis Vlachos is a Research Member at IBM Zurich Research Laboratory. Previously he was with IBM Research, NY, and has also visited Microsoft Research, Seattle. Dr Vlachos’ research interests include data mining, machine learning, databases, time-series analytics and data visualization. For his contributions at IBM he has received two Research Division Awards and three Invention Plateau Awards. He holds or has applied for 15 patents.

  Mailing List:

Sponsor: The CS Colloquium Series is supported by a generous donation from Microsoft