基于关联规则的本体相似度综合计算方法  

Comprehensive method of computing ontology similarity based on association rules

作　　者：李华[1] 苏乐[1] 

机构地区：[1]重庆大学计算机学院,重庆400044

出　　处：《计算机应用》2012年第9期2472-2475,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(71102065)

摘　　要：目前较为流行的最小风险的本体映射(RiMOM)框架通过采用"多策略"的思想虽然取得了一定的效果,但其框架比较臃肿庞杂,且采用的计算结构相似度的选择策略存在一定的局限性。针对上述问题,提出一种基于关联规则的本体相似度综合计算方法。首先,构造关联规则的结构"树"模型,得出相应事务集;其次,进行关联规则的挖掘,根据关联规则计算概念结构的相似性;然后,计算概念的实例、属性、名称的相似度;最后,对多个特征相似度进行综合加权处理,实现本体相似度的最优计算。实验结果表明,该方法较RiMOM在查全率、查准率方面均有较大提高;同时该方法省去了策略选择的步骤,有效降低了时间复杂度。The recent Risk Minimization based Ontology Mapping （ RiMOM）, based on multi-strategy, in spite of certain improvements, has the shortcomings in complexity and redundancy, and limitations in similarity selection of structure characteristics. Therefore, a comprehensive method of computing ontology similarity based on association rules was proposed, First, the ＂ tree＂ model of association rules was constructed to compute the structure similarity. Then the similarity of instances, properties, and names were also computed. Finally, a comprehensive method was used to achieve the optimal calculation of the ontology similarity. The experimental results show that this method has higher recall and precision ratio and less time complexity than RiMOM.

关 键 词：本体 数据挖掘 关联规则 本体映射 语义分析 

分 类 号：TP311.11[自动化与计算机技术—计算机软件与理论] TP182[自动化与计算机技术—计算机科学与技术]