特定领域下关系模板的研究  

Research of Relation Pattern in Specific Domain

作　　者：张素香[1] 李蕾[1] 谭咏梅[1] 

机构地区：[1]北京邮电大学信息工程学院

出　　处：《北京邮电大学学报》2006年第5期79-83,共5页Journal of Beijing University of Posts and Telecommunications

基　　金：国家"863计划"项目(2001AA114210)

摘　　要：以公司人事变动领域为例,针对实体关系抽取课题,从知识自动获取角度出发,基于Bootstrapping思想提出了层次知识获取模型,利用内外2层模块相互嵌套自动获取知识,获得了实体关系分析所需要的领域专用词典和模板规则.结合全信息理论,对模板添加语义和语用标注,生成全信息知识库.在此基础上,完成关系抽取实验和评测.A new method of automatic entity relation extraction is proposed. Based on the Bootstrapping algorithm, the hierarchy knowledge extraction model can be designed. The inner specific word extraction model and outer pattern extraction model can be nested each other to extract automatically knowledge, so that the specific dictionary and pattern rules used for the entity relation extraction is achieved. Combined with the Comprehensive information theory, the semantic and pragmatic information can be added into the relation extraction patterns to generate the comprehensive information knowledge-base（CIKB）. Both the experiments of relation extraction and the evaluation have been done.

关 键 词：全信息理论 全信息知识库 层次知识获取 标量聚类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]