摘要
A Kullback-Leibler（KL）distance based algorithm is presented to find the matches between concepts from different ontologies. First, each concept is represented as a specific probability distribution which is estimated from its own instances. Then, the similarity of two concepts from different ontologies is measured by the KL distance between the corresponding distributions. Finally, the concept-mapping relationship between different ontologies is obtained. Compared with other traditional instance-based algorithms, the computing complexity of the proposed algorithm is largely reduced. Moreover, because it proposes different estimation and smoothing methods of the concept distribution for different data types, it is suitable for various concepts mapping with different data types. The experimental results on real-world ontology mapping illustrate the effectiveness of the proposed algorithm.
提出了一种基于Kullback-Leibler(KL)距离的本体映射方法.该方法将本体中每个概念抽象为一个概率分布,并通过相应的实例数据对其进行估计;对于不同本体的2个概念,通过计算相应概率分布之间的KL距离而求得其相似度.进而求得本体间概念的映射关系.该方法与传统的方法相比,极大地降低了计算的复杂度,并且此算法针对不同的数据类型提出了不同的概念分布的估计和平滑方法,所以能够适用于各种数值类型的概念映射.通过试验,证明了此方法的有效性.
基金
Foundation of Next Generation Internet of China
