A General Framework For Graph Matching And Its Application In Ontology Matching

Graph matching (GM) is a fundamental problem in computer science. Two issues severely limit the application of GM algorithms. (1) Due to the NP-hard nature, providing a good approximation solution for GM problem is challenging. (2) With large scale data, existing GM algorithms can only process graphs with several hundreds of nodes.We propose a matching framework, which contains nine different objective functions for describing, constraining, and optimizing GM problems. By holistically utilizing these objective functions, we provide GM approximated solutions. Moreover, a fragmenting method for large GM problem is introduced to our framework which could increase the scalability of the GM algorithm.The experimental results show that the proposed framework improves the accuracy when compared to other methods. The experiment for the fragmenting method unveils an innovative application of GM algorithms to ontology matching. It achieves the best performance in matching two large real-world ontologies compared to existing approaches.