Abstract: In ontology matching it’s difficult for the implicit information contained in structure to be correctly represented and fully used. As for this problem, this article introduces the quantization of ontology structural characteristics into ontology matching. According to the similarity between ontology and complex network, the article analyzes the network characteristics of ontology and brings forward a series of theories based on structural and semantic characteristics as well as quantization criteria for nodes and edges. These theories and quantization criteria are applied to the selection of core nodes and measurement of edge weight. In this article, the problem of ontology matching is transformed into the matching of weighed label graphics, which are based on individual characteristics and organizational structure. The matching of approximately optimization is acquired through quadratic programming. Results of experiments have showed that the structural characteristics of ontology have great influence on ontology matching and the close coupled with the matching of core nodes will significantly improve the accuracy of matching.

Key words: complex networks, structure characteristics quantization, core node, semantic distance, approximate ontology matching, combinatorial optimization

摘要： 针对本体匹配中结构蕴含的隐式语义信息难以正确表示和充分使用问题，提出将本体结构特征量化引入本体匹配。根据本体与复杂网络的相似性，分析本体具有的网络特征，提出了一系列基于结构和语义特征的理论和节点、边的量化标准，并将其用于核心节点的选取和边权重的度量。将本体匹配转换为基于元素个体特征和整体组织结构的有权标签图匹配问题，通过二次规划方法求取近似最优匹配。实验证明本体拓扑结构特征对本体匹配具有较大影响，其与核心节点匹配的紧密藕合能够显著提高匹配的准确性。

关键词: 复杂网络, 结构特征量化, 核心节点, 语义距离, 本体近似匹配, 组合优化