基于OMKast的术语关系获取和验证  

Term relationship acquisition and verification based on OMKast

作　　者：王娜娜[1,2] 黄运有 唐素勤[1] 王石[2] 曹存根[2] 

机构地区：[1]广西师范大学计算机科学与信息工程学院,广西桂林541000 [2]中国科学院计算技术研究所智能信息处理重点实验室,北京100190

出　　处：《计算机应用研究》2015年第11期3319-3323,共5页Application Research of Computers

基　　金：广西省自然科学基金资助项目(2011GXNSFA018159);国家自然科学基金青年基金资助项目(61103169);中国科学院计算技术研究所合作项目(国家自然科学基金(61173063)子课题);广西省研究生教育创新计划资助项目(YCSZ2014097)

摘　　要：为了从海量数据中获取知识,提出了术语间关系的抽取方法:定义了上下位关系和整分关系,在此基础上新增加了N条关系,从关系的定义域和值域、关系的限制条件、关系的公理等出发,准确地定义出关系的内涵,并根据关系的内涵定义出关系的语义特征。最后针对关系表达的灵活性,总结出了关系表述的语法特征和表述文法。结合上述语法特征和语义特征,编写了一种可执行的知识抽取程序——OMKast程序,并从原始文本语料中抽取关系。利用语义特征和统计学的方法验证抽取出的关系。实验结果表明,该方法具有良好的有效性。In order to obtain knowledge from big data,this paper proposed a method of term relationship extraction. Firstly, it defined the hyponymy and part whole relationship and added new relationship based on these relationship. Then accurately defined the connotation of the relationship by defining the relationship＇ s domain, range, constraints, axiom etc, and defined the semantic features of the relationship based on the connotation of the relationship. Finally it summarized the grammatical features and grammer of the expression of relationship in view of the flexibility of the expression of relationship. With the help of the grammatical features and the semantic features. A program was wrote and used to extract the relationship from the text. It used the semantic features and statistical method to validate the set extracted from corpus with above relationship. Experimental results demonstrate the feasibility of the method.

关 键 词：术语 术语间关系 关系公理 文本知识获取 术语关系获取 术语关系验证 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]