基于图的DL-Lite本体不一致性处理方法的研究

发布时间：2018-03-14 17:10

本文选题：DL-Lite　切入点：本体调试　出处：《东南大学》2016年博士论文　论文类型：学位论文

【摘要】：互联网的高速发展带来了海量的数据,其中大量的非结构和半结构的数据不利于自动化处理。语义Web为计算机能够理解的结构化数据提供了技术支持,促进了人机协同工作。作为语义Web技术的重要组成部分,基于描述逻辑的本体语言OWL为其提供了严谨的逻辑基础。在描述逻辑的语言体系中,DL-Lite能提供多项式时间内的推理服务,同时又具备UML以及E-R模型的表达能力。DL-lite下的应用受到了学术界以及工业界的广泛关注。在实际应用中,本体的构建与维护是一项复杂的过程。此过程常导致本体出现不一致性问题,从而使得标准推理服务失效。因此本体中的不一致性是本体工程中必须要处理的一类问题。针对本体中的不一致性问题,通常有两种处理方法,一是找出并消除本体中的不一致：二是应用非标准的推理方法,在不一致存在的情况下实现有意义的推理。本文重点关注DL-Lite中的不一致性问题并且结合以上两种方法开展工作。具体内容如下：(1)在本体调试方面,提出了一种基于图的本体调试方法。将DL-Lite本体转换成有向图,通过路径遍历的方式来计算本体中的最小不可协调保持子集。基于图的调试方法摆脱了对描述逻辑推理机的依赖,对比实验结果表明新的方法有更高的执行效率和稳定性。(2)在本体修正方面,提出了一种基于图的修正方法。新的修正方法首先通过修正状态将不一致木体划分成保留部分和待判别部分。在修正状态的基础上定义出一种新的满足最小改变原则的修正算子。本文采用两种方式实现了新的修正算子,一种是基于评分函数的,另一种基于碰集树的方法。从对比实验结果来看,基于评分函数的修正方法执行效率要高于基于碰集树的修正方法。(3)在不一致容忍语义方面,定义了两种新的不一致容忍语义。新定义的不一致容忍语义与经典的不一致容忍语义相比避免了计算整个ABox关于TBox的闭包,同时保证了原语义的表达能力,有效地提高了不一致容忍下推理服务的效率。(4)在不一致查询应答方面,提出了一种基于不一致容忍语义的查询方法,新的查询方法将本体和查询分别以不同的规则转换成图,这避免了经典查询重写方法导致的查询项冗长的问题,并且优化了查询的效率。
[Abstract]:The rapid development of the Internet has brought a great deal of data, among which a large number of unstructured and semi-structured data are not conducive to automated processing. Semantic Web provides technical support for structured data that can be understood by computers. As an important part of semantic Web technology, OWL, an ontology language based on description logic, provides a rigorous logic basis for it. DL-Lite can provide reasoning services in polynomial time in the language system of description logic. At the same time, the application of UML and E-R model. DL-lite has been widely concerned by academia and industry. The construction and maintenance of ontology is a complex process, which often leads to inconsistency of ontology. Therefore, inconsistency in ontology is a kind of problem that must be dealt with in ontology engineering. One is to identify and eliminate inconsistencies in ontology, and the other is to apply non-standard reasoning methods. This paper focuses on the inconsistency in DL-Lite and combines the above two methods to carry out the work. The content of this paper is as follows: 1) in ontology debugging, In this paper, a graph-based ontology debugging method is proposed. The DL-Lite ontology is transformed into a directed graph, and the minimal uncoordinated holding subset in the ontology is calculated by path traversal. The graph-based debugging method gets rid of the dependence on the description logic inference engine. The experimental results show that the new method has higher execution efficiency and stability. In this paper, a graph based correction method is proposed. Firstly, the inconsistent wood body is divided into reserved and discriminant parts by modifying the state. Based on the modified state, a new method is defined to satisfy the minimum change. In this paper, a new modified operator is implemented in two ways. One is based on scoring function, the other is based on collision set tree. From the results of comparative experiments, the efficiency of the modified method based on score function is higher than that based on collision set tree. Two new inconsistency tolerance semantics are defined. Compared with the classical inconsistency tolerance semantics, the new inconsistency tolerance semantics avoids the computation of the entire ABox closure on the TBox, and at the same time ensures the expressive ability of the original semantics. The efficiency of reasoning service under inconsistent tolerance is improved effectively. In the aspect of inconsistent query response, a query method based on inconsistency tolerance semantics is proposed. The new query method transforms ontology and query into graphs with different rules, respectively. This avoids the problem of query term verbosity caused by classical query rewriting method and optimizes query efficiency.
【学位授予单位】：东南大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.1

【参考文献】