中文摘要：

原子分解是理解本体内部模块结构的有效途径。以局部化模块抽取为基本操作的原子分解方法可用于强表达力的SROIQ本体,但效率较低。基于有向超图的本体模型能够显式地反映原子的依赖关系,但只局限于弱表达力的EL本体。提出一种混合的原子分解算法,首先利用有向超图表示EL子本体,形成部分原子分解,利用模块抽取方法添加剩余非EL公理,得到本体的全部原子分解。以生物医学本体作为测试数据,实验表明,这种混合的原子分解算法能够有效减少运行时间。与传统的基于模块抽取的方法相比,原子分解效率平均提高6.7倍。

英文摘要：

Atomic decomposition is an important approach for understanding the inner structure of ontology.The traditional approach of atomic decomposition,in which the local module extraction is basic operation,is able to decompose expressive SROIQ ontology,however,this approach is inefficient.The direct hypergraph model for ontology represents the dependency relationship of atoms explicitly,it is restricted to an inexpensive EL ontology.In this paper,a hybrid approach is proposed for atomic decomposition.First,the EL subontology is presented in direct hypergraph which forms a partial atomic decomposition for ontology.Then,the remainder non-EL axioms are added into the existential atomic decomposition to obtain the complete decomposition.An empirical evaluation of the algorithm on biomedical ontologies confirms a significant improvement in running time.An average speedup of 6.7-fold is achieved compared to the traditional approach.