Optimizing semantic web services ranking using parallelization and rank aggregation techniques

Abstract

The problem of combining many rank orderings of the same set of candidates, also known as the rank aggregation problem, has been intensively investigated in the context of Web (e.g meta-search) databases (e.g combining results from multiple databases), statistics (e.g. correlations), and last but not least sports and elections competitions. In this paper we investigate the use of rank aggregation in the context of Semantic Web services. More precisely we propose an optimization technique for ranking Semantic Web services based on non-functional properties by using parallelization and rank aggregation methods. Instead of using a ranking algorithm over the entire set of non-functional properties our approach splits the set of non-functional properties in multiple subsets, runs the ranking algorithm on each of the subsets and finally aggregates the resulting ranked lists of services into one unifying ranked list. Experimental results reported in this paper show improvements of our initial rank aggregation method both in terms of quality and processing time.