MPP-MLO: Multilevel Parallel Partitioning for Efficiently Matching Large Ontologies

Abstract

The growing usage of Semantic Web has resulted in an increasing number, size and heterogeneity of ontologies on the web. Therefore, the necessity of ontology matching techniques, which could solve these issues, is highly required. Due to high computational requirements, scalability is always a major concern in ontology matching system. In this work, a partition-based ontology matching system is proposed, which deals with parallel partitioning of the ontologies at multilevel. At first level, the root based ontology partitioning is proposed. Match able sub-ontology pair is generated using an efficient linguistic matcher (IEI-Sub) to uncover anchors and then based on maximum similarity values, pairs are generated. However, a distributed and parallel approach of Map Reduce-based IEI-sub process has been proposed to efficiently handle the anchor discovery process which is highly time-consuming. In second level partitioning, an efficient approach is proposed to form non-overlapping clusters. Extensive experimental evaluation is done by comparing existing approaches with the proposed approach, and the results shows that MPP-MLO turns out to be an efficient and scalable ontology matching system with 58.7% reduction in overall execution time.