Talk by Vasilis Vassalos on DB Seminar

 

Score-consistent algebraic optimization of full text queries

 

Abstract

 
We discuss solutions to two important open problems for the execution of full-text search queries. First, we show how to correctly apply traditional database rewrite optimizations to full-text query plans with integrated scoring, and explain why existing techniques fail. Second, we show how our techniques are applied in a generic scoring framework that supports a wide class of scoring algorithms, including algorithms seen in the literature and user-defined scoring. These are the first steps towards the creation of a performant cost-based information retrieval engine.   
 

Bio

 
Vasilis Vassalos is an Associate Professor of Computer Science, serving at AUEB since 2004 and is leading the Web Information Management group of the Information Systems and Databases Laboratory. He is the recipient of a 2007 Marie Curie Outgoing International Fellowship and a Visiting Scholar at the CSE Department of UC San Diego for fall-winter 2007.