Approximate Ontology Matching Based on Structure Quantization

There is much implicit semantic information hidden in ontology structure, which hasn’t been used in ontology matching. In this paper, we analyse the network characteristics of ontology. Propose a set of semantic and theoretical criterions to measure the different characteristics of nodes and edges. Use these quantitative characteristics to identify core concept nodes and assign weight to edges. Then, convert the ontology matching to Labelled Weighted Graph Matching problem, and use convex relaxation algorithm to solve this quadratic programming problem. We implement our prototype and experimentally evaluate our approach on data sets. The evaluation results demonstrate that structure information significant effect matching result and our approach can achieve good precision and recall.