Hybrid Web Service Discovery Based on Fuzzy Condorcet Aggregation

The web service discovery is a major issue in service oriented computing. We distinguish several semantic service matchmakers that cover multiple matching criteria, these approaches may use crisp logic-based matching, token based similarity measures, and eventually machine learning that combines many individual rankings into a global matching list. This latter category aims to boost the discovery performance by using various matching algorithms, but it introduces additional difficulties, such as the weighting of the matching algorithms components, and the compensation between the matching criteria(inputs, outputs, pre-conditions, effects ... ). The purpose of this paper is to handle the aforementioned difficulties by introducing a majority vote based approach. This technique fuses five individual rankings (four textual similarity measures and a pure logic matching algorithm) into a global ranking. The different scores are aggregated according to the condorcet principle. More specifically we use a fuzzy dominance relationship, to compare the services, and thus we infer the condorcet order of the final ranking. We have tested our approach, on the OWLSTC benchmark, and the preliminary results are very encouraging.