CAJ | 학술논문

本体学习是当前本体研究的热点之一。概念的抽取和分类关系的构建是本体学习的关键。领域概念的分类方式很多，如何根据不同的应用目的，生成合适的分类结构，是当前本体学习系统应该着重考虑的问题。论文提出了一种有指导的领域本体概念体系结构学习方法，该方法以中文 Web 上蔬菜领域非结构化文本为语料，将浅层句法分析等语言学方法与 TFIDF 和 C‐值等统计学方法相结合进行概念抽取；在分类关系抽取时，基于目标本体的已知一个分支，采用余弦距离计算概念与已知分支概念的语义距离，并结合概念之间的共现频度来确定层次及上下位关系。与目前中文本体的代表性方法相比，文中提出的方法在查全率和查准率方面有明显的提高。
본체학습시당전본체연구적열점지일。개념적추취화분류관계적구건시본체학습적관건。영역개념적분류방식흔다，여하근거불동적응용목적，생성합괄적분류결구，시당전본체학습계통응해착중고필적문제。논문제출료일충유지도적영역본체개념체계결구학습방법，해방법이중문 Web 상소채영역비결구화문본위어료，장천층구법분석등어언학방법여 TFIDF 화 C‐치등통계학방법상결합진행개념추취；재분류관계추취시，기우목표본체적이지일개분지，채용여현거리계산개념여이지분지개념적어의거리，병결합개념지간적공현빈도래학정층차급상하위관계。여목전중문본체적대표성방법상비，문중제출적방법재사전솔화사준솔방면유명현적제고。
Concepts extraction is the basis of relationship learning ,axioms learning ,and other subtasks of ontology learning .In the context of the increasing interest in the use of Web data ,the underutilized structured information of Baidu Encyclopedia with the characteristic of wiki ,and the much lower concepts learning accuracy in agriculture than other do‐mains ,the methods of concepts extraction are researched for vegetable domain ontology .The experimental results reveal that the multi‐strategy of combing linguistics‐based techniques ,statistics‐based techniques and structured information of Baidu Encyclopedia raise the recall of concepts learning ,the hybrid statistical approach of the measures TFIDF and C‐value has im‐proved the precision remarkably .This research will promote progress in the area of agricultural ontology learning ,especially that of the Web‐based data in Chinese language .