Developing graph-based approaches for to match large ontologies

Ontologies are domain-specific metadata that describes relationships between properties of a field, sample data, and properties developed for many different purposes. But, they can be described in different names or properties even within the same domain. Elimination of definition differences and finding the similarities between existing ontologies are provided by ontology matching algorithms. Ontology matching algorithms are used especially for data integration, reuse of ontologies and information extraction. Rapidly increasing data sizes and the development of information technologies have led to the growth of ontologies and so this development makes the ontology mapping algorithms difficult to work. Because of existing the small number of structural-based ontology matching algorithms among methods supports the idea of ​​developing graph-based ontology mapping architectures. In this study, a graph-based architecture was proposed to match large ontologies. With this architecture, it is primarily aimed to divide the large ontologies into small pieces and then matches them using sub-graph mining algorithms. Karger algorithm and clique percolation algorithm were used to divide large ontologies and both algorithms were applied to ontologies for the first time. In the next step of the thesis, these sub--parts obtained by using sub--graph mining algorithms are matched. GraMi and gSpan algorithms were selected as sub-graph mining algorithms and were first used in ontology matching field. The proposed model is compared to AML and Falcon-AO matching algorithms in terms of accuracy and time performance. According to obtained the results, it is seen that GraMi algorithm is better than matching algorithms for matching accuracy.