Simple May Be Best - A Simple and Effective Method for Federated Web Search via Search Engine Impact Factor Estimation

This paper reports our participation in the three tasks, i.e., vertical selection (VS), resource selection (RS) and results merging (RM) in TREC 2014 Federated Web Search track. In consideration of the connections between vertical and search engine (i.e., a vertical could contain multiple resources), we address the two tasks in an iterative way. Existing algorithms adopted relevance measures to calculate the semantic relatedness between query and resources or returned results. However they neglected the influence of search engine in itself. In this work, we propose a Search engine Impact Factor (SEIF) estimation approach to improve the performance of vertical and resource selection. The officially released results showed that our systems ranked 1st in RS task and 2nd in VS task.