语义Web中的本体自动映射  

Automatic Ontology Mapping in Semantic Web

作　　者：唐杰[1] 梁邦勇[1] 李涓子[1] 王克宏[1] 

机构地区：[1]清华大学计算机科学与技术系知识工程组,北京100084

出　　处：《计算机学报》2006年第11期1956-1976,共21页Chinese Journal of Computers

基　　金：国家自然科学基金(90604025)资助

摘　　要：分布式语义信息集成是语义Web面临的六大挑战之一.本体映射是语义集成的关键.文章基于贝叶斯决策理论提出最小风险的本体映射模型:RiMOM(Risk Minimization based Ontology Mapping).RiMOM将映射发现问题转换成风险最小化问题,提供了一个多策略的本体映射方法.该方法不仅在1∶1的映射上取得了较好的效果,还实现了n∶1映射.实验表明在几个公开的数据集上,RiMOM可以取得比同类方法更高的查准率和查全率.Interoperability over distributed ontologies is one of the six challenges for Semantic Web. Ontology mapping is the key point to each the interoperability. In this paper, based on Bayesian decision theory, the authors propose an approach called RiMOM （Risk Minimization based Ontology Mapping） to automatically discover the mapping between ontologies. RiMOM treats the mapping problem as a decision problem and formalizes mapping discovery as that of risk minimization. Based on multiple strategies, RiMOM deals with not only 1：1 mapping, but also n：1 mapping. Experiments on several public data sets show that RiMOM can outperform existing methods in terms of precision and recall.

关 键 词：语义WEB 本体映射 最小风险决策 本体集成 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]