本体推理机求解Mups的性能评测研究

发布时间：2018-05-23 21:42

本文选题：不一致术语集 + 极小不可满足保持子术语集　；参考：《计算机学报》2017年06期

【摘要】：求解极小不可满足保持子术语集(Mups)是不一致术语集调试的核心工作.在构建术语集依赖关系图模型基础上,从概念之间的依赖关系角度出发,定义语义依赖度、语义簇、依赖度分布3个指标反映本体术语集的复杂程度;通过讨论不可满足概念数目、冲突公理集基数和冲突公理基数对Mups问题求解难易的影响,定义冲突公理集最大基数和冲突公理最大基数两个指标反映不一致本体术语集的数据复杂程度;基于这些复杂性指标,设计针对Mups问题的不一致本体数据标准测试集(Mups Benchmark,MupsBen)来评测Pellet、Hermit、FaCT++、JFact和TrOWL这5种推理机在黑盒算法下求解Mups的性能.评测实验显示,所定义的复杂度指标能够有效反映Mups求解问题的数据复杂程度.对于特定推理机,其性能随测试数据的结构复杂程度的增大而降低;对于不同推理机,由于其内在推理机制与优化策略的差别,在不同复杂度指标下表现出不同的性能差异.
[Abstract]:Solving minimal unsatisfiable retention subsets (Mups) is the core of inconsistent terminology set debugging. On the basis of constructing dependency graph model of terminology set, from the angle of dependency relation between concepts, three indexes, semantic dependency degree, semantic cluster and dependency degree distribution, are defined to reflect the complexity of ontology terminology set. By discussing the number of unsatisfiable concepts, the effects of conflict axiom set cardinality and conflict axiom cardinality on the difficulty of solving Mups problem, Defining the maximum cardinality of the conflicting axiom set and the maximum cardinality of the conflicting axiom reflect the data complexity of the inconsistent ontology terminology set; based on these complexity metrics, A standard test set of inconsistent ontology data for the Mups problem is designed to evaluate the performance of five inference machines (Pellet Hermitt FaCT JFact and TrOWL) for solving Mups under the black box algorithm. The evaluation results show that the defined complexity index can effectively reflect the data complexity of the Mups problem. For a specific inference engine, its performance decreases with the increase of the complexity of the structure of the test data. For different inference machines, due to the difference between their internal reasoning mechanisms and optimization strategies, the performance varies under different complexity indicators.
【作者单位】：吉林大学计算机科学与技术学院;吉林大学符号计算与知识工程教育部重点实验室;
【基金】：国家自然科学基金(61133011,41172294,61170092,61272208) 吉林省科技发展计划(201201011)资助~~
【分类号】：TP18

【参考文献】