面向跨媒体信息的领域本体学习方法与应用研究

发布时间：2018-08-22 15:41

【摘要】：本体学习是一种本体的自动构建或对现有本体自动更新的方法,随着跨媒体信息时代的到来,互联网信息的内容与结构日趋复杂,使得领域本体的构建与自动更新成为新的研究热点。本文在分析跨媒体信息特征及民航突发事件应急管理领域本体应用特点的基础上,引入隐狄利克雷(Latent Dirichlet Allocation,LDA)主题模型,围绕领域本体学习方法及应用进行了相关的研究。在深入分析跨媒体信息结构、领域本体特点和本体学习方法等相关理论及应用的基础上,提出了一种基于LDA的领域本体学习方法,给出了方法的体系框架。利用自然语言处理方法获取领域本体概念相关的候选术语集,设计了领域本体的LDA主题模型,通过吉布斯采样进行了LDA模型训练与主题推断,完成了领域本体核心概念的相关术语自动提取。结合领域本体概念相关术语的主题概率分布,研究了领域本体概念间语义关系识别规则的构建方法,给出了概念及其相关术语的语义关系识别过程,在此基础上进一步研究了跨媒体信息与实例数据之间的映射关系,根据LDA主题推断实现了实例与概念的匹配。实验结果表明,基于LDA的领域本体学习方法可以有效解决大规模领域本体概念、关系及实例的自动更新问题,为大数据环境下民航突发事件领域本体在信息共享、推理与语义查询等应用方面提供了更好的数据支持。
[Abstract]:Ontology learning is a method to construct ontology automatically or update existing ontology automatically. With the arrival of cross-media information age, the content and structure of Internet information are becoming more and more complex. It makes the construction and automatic updating of domain ontology a new research hotspot. Based on the analysis of cross-media information features and the characteristics of ontology application in the field of emergency management of civil aviation emergencies, this paper introduces the topic model of (Latent Dirichlet allocation (LDA), and studies the method and application of domain ontology learning. On the basis of deeply analyzing the theory and application of cross-media information structure, domain ontology characteristics and ontology learning methods, a domain ontology learning method based on LDA is proposed, and the framework of the method is given. Using the natural language processing method to obtain the candidate terminology set of domain ontology concepts, the LDA topic model of domain ontology is designed, and the LDA model training and topic inference are carried out by Gibbs sampling. The related terms of domain ontology are extracted automatically. Based on the subject probability distribution of domain ontology concepts, this paper studies the construction method of semantic relation recognition rules between domain ontology concepts, and gives the process of semantic relation recognition of concepts and related terms. On this basis, the mapping relationship between cross-media information and instance data is further studied, and the matching between instance and concept is realized based on LDA topic inference. The experimental results show that the domain ontology learning method based on LDA can effectively solve the problem of automatic updating of large-scale domain ontology concepts, relationships and examples, which is the information sharing of civil aviation emergency domain ontology under big data environment. Reasoning and semantic query provide better data support.
【学位授予单位】：中国民航大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.1

【参考文献】