Wednesday, May 2, 2012 - 9:40am to 11:00am
KEC 1007

Speaker Information

Arash Termehchy
Ph.D. student
Department of Computer Science
University of Illinois at Urbana-Champaign

Abstract

Current data search and exploration paradigms fall short of providing usable, effective, efficient, robust, and economical search and exploration experiences. For instance, Web search engines do not effectively satisfy complex information needs. Although database query languages such as SQL are intended for expressing complex information needs, the languages are too hard to use, and the underlying databases are too maintenance-intensive, to be viable bases for an effective Big Data infrastructure. In my research, I have set forth solid theoretical foundations for a usable, effective, robust, and economical data search and exploration experience and developed large scale systems that provide such an experience.

In this talk, I will argue that keyword and natural language query interfaces should use deep meta-data information to effectively rank the results of queries. I also postulate that search results from keyword and natural language query interfaces should not depend on how the underlying data sets are organized. I show that current query interfaces do not adhere to these principles. I describe a series of large scale search and exploration systems that embody and validate these principles through extensive user studies over real world data.

Speaker Bio

Arash Termehchy is a Ph.D. student at the Department of Computer Science, University of Illinois at Urbana-Champaign under the supervision of Marianne Winslett. His research interest is in data and information management in a broad sense, including large scale data management, human centric data management, data trustworthiness, social data management, data mining, and semantic Web. He is the recipient of the ICDE'11 best student paper award, the ICDE'11 best papers selection, the Yahoo! Key Scientific Challenges award, and the Feng Chen Memorial award.