基于归纳逻辑程序设计的本体学习方法

项目名称： 基于归纳逻辑程序设计的本体学习方法

项目编号： No.60873153

项目类型： 面上项目

立项/批准年度： 2009

项目学科： 武器工业

项目作者： 高志强

作者单位： 东南大学

项目金额： 28万元

中文摘要： 本体是语义Web的核心。本项目旨在研究基于归纳逻辑程序设计的OWL DLP本体学习方法。OWL DLP是OWL Lite的子集和RDFS的超集，它覆盖了绝大部分语义Web本体。本项目的主要贡献如下：（1）提出一种改进的适于学习OWL DLP本体的归纳逻辑程序设计方法。利用数据集中的一元谓词（个体断言）和二元谓词（性质断言）集合构造实例空间，将关系路径看成正例的解释，在较小的假设空间中进行遍历搜索，从而得到高质量的规则集。（2）提出基于统计关系学习，特别是Markov逻辑网的直接学习OWL DLP本体的方法。它结合了概率图模型，特别是Markov网络能够处理非确定性数据的特点，可有效处理链接开放数据中的噪音。（3）针对本体中存在的不一致性问题，基于启发式策略，提出一种本体不一致推理和诊断方法。和其它方法相比，该方法可以在较短的时间内发现更多的MUPS。（4）开发了本体学习工具，构造了面向电信领域研发的语义搜索与分析系统，并得到初步应用。

中文关键词： 本体学习;归纳逻辑程序设计;OWL DLP;统计关系学习

英文摘要： Ontology plays an important role in Semantic Web. This project aims to study OWL DLP ontology learning approaches based on inductive logic programming. OWL DLP is the subset of OWL Lite and superset of RDFS. It is reported that OWL DLP covers most of the Semantic Web ontologies. Main contributions of this project are as follows: (1) An improved inductive logic programming algorithm is put forward, which is suitable for learning OWL DLP ontologies. Making using of unary (individual assertion) and binary (property assertion) predicates to construct instance spaces, and using relational paths as explanations for positive examples, we learn high quality rule sets by searching a much smaller hypothesis space. (2) Based on statistical relational learning, especially Markov logic network, we learn OWL DLP ontologies directly from data sets. This approach takes advantages of probabilistic graphical model, especially Markov network, which can deal with uncertainty naturally. As a result, this approach processes noises in Linked Open Data effectively and efficiently.（3）With respect to inconsistency in ontologies, an approach for inconsistent ontology reasoning and debugging has been put forward, which can find much more MUPS in less time by the heuristics strategy when compared to other approaches. (4) Several tools for ontology learning have bee developed, and we have implemented a semantic search and analysis system for research and development in telecommunication science and technology, which has been used for years.

英文关键词： Ontology Learning; Inductive Logic Programming; OWL DLP; Statistical Relational Learning