摘要
术语层次关系抽取是领域概念关系体系构建的重要基础。针对目前术语关系抽取中手工实现的问题,提出了基于词形规则模板匹配的术语层次关系抽取方法,实现从科技论文文本中抽取类属关系(IS-A)和整体部分关系(PART—OF)关系。利用复合术语的head和modifier特征,比较两个术语之间存在的边缘共用词汇,构造模板来确定它们之间的IS—A和PART-OF关系;提出泛化度指标,用于测量两个术语在概念层次树上的相对位置;提出相关度概念,用于测量两个术语之间在语义上的相关性。对不存在共用词汇和不匹配模板的术语采用泛化度差值和相关度来判断它们之间是否存在层次关系。实验从信息资源管理领域的论文文本中提取层次关系术语对1306对,准确率达到92.5%,证明提出的方法是有效的。
A term relationship extraction method was put forward integrating morphology rules and statistics analysis to extract two types of hierachical relations among term pairs : IS-A and PART-OF. In morphology rule analysis, five types of templates were designed to judge hierachical relations among terms with common left or fight sub-string using head or modifier feature of multi terms. A generation index was put forward to measure generation degree of a term and judge hieraehieal position of two terms; an association index was put forward to measure association degree of term pair and judge similarity relation of two terms in concept tree. Presented methods contains following process: term generation measure computation, term pair association degree computation, morphology rule analysis and template match, non-match term pair relationship judgement. In experiments, 1306 hierarchical relation term pairs were extracted from information resource management paper corpus, and the precision is 92.5%.
出处
《情报学报》
CSSCI
北大核心
2013年第7期708-715,共8页
Journal of the China Society for Scientific and Technical Information
基金
中国科学技术信息研究所预研项目“基于内容和链接的学术社交网络分析”(YY201221)
“十二五”国家科技支撑计划项目“面向外文科技知识组织体系的大规模语义计算关键技术研究”(2011BAH10804)“基于STKOS的知识服务应用示范”(2011BAH10B06)
中国人民大学明德学者科学研究基金(中央高校基本科研业务费专项资金资助)“知识工程背景下信息资源管理术语构建方法研究”项目(10XNJ052)资助
关键词
术语关系抽取
层次关系
词形规则
文本挖掘
term relation extraction, hierarchical relation, morphology rule, text mining