CAJ | 학술논문

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.
제출료일충기우Kullback-Leibler(KL)거리적본체영사방법.해방법장본체중매개개념추상위일개개솔분포,병통과상응적실례수거대기진행고계;대우불동본체적2개개념,통과계산상응개솔분포지간적KL거리이구득기상사도.진이구득본체간개념적영사관계.해방법여전통적방법상비,겁대지강저료계산적복잡도,병차차산법침대불동적수거류형제출료불동적개념분포적고계화평활방법,소이능구괄용우각충수치류형적개념영사.통과시험,증명료차방법적유효성.
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.