Metasearch via Voting

Metasearch engines are developed to overcome the shortcoming of single search engine and try to benefit from cooperative decision by combining the results of multiple independent search engines that make use of different models and configurations. In this work, we study the metasearch problem via voting that facilities multiple agents making cooperative decision. We can deem the source search engines as voters and all ranked documents as candidates, then metaseach problem is actually to find a voting algorithm to obtain group’s preferences on these documents(candidates). In addition to two widely discussed classical voting rules: Borda and Condorcet, we study another two voting algorithms, Black and Kemeny. Since Kemeny ranking problem is NP-hard, a new heuristic algorithm has been proposed for metasearch. Some experiments have been carried out on TREC2001 data for evaluating these metasearch algorithms coming from voting.