一种从术语定义句中自动抽取知识单元的方法

摘要

知识单元是对专业领域知识的细粒度表示，实现对知识的有序化、结构化描述。以术语定义句为语料，首先进行依存句法分析和语义角色的自动标注，获得知识单元的语义线索，然后制定抽取规则并设定权值，自动抽取出知识单元。实验结果表明，利用该方法进行知识单元的抽取F值为68.69%，效果良好，具有较好的领域适应性。该方法对于知识组织、术语库建设、知识挖掘等提供了有益的参考。%This paper presents a new method for knowledge unit automatic extraction, which is expected to describe structured knowledge in some certain domains. Using definitions of term, the framework of sentences will be obtained by dependency parsing and semantic roles annotation for knowledge units. Then, some rules and weights were made to extract knowledge units automatically. The experiment indi-cates that this method has better performance with f-measure 68. 69%, which can be used in different domains. This method provides a useful reference for knowledge organization, term bank construction and knowledge mining.