【作者】 张亮； 尹存燕； 陈家骏；

【Author】 Zhang Liang~(1,2),Yin Cunyan~1,Chen Jiajun~1 1.State Key Laboratory for Novel Software Technology,Nanjing University,Nanjing 210093 2.Jiangsu Police Institute,Nanjing 210000

【机构】 南京大学计算机软件新技术国家重点实验室； 江苏警官学院公安科技系；

【摘要】 基于语义资源Hownet的词语相似度计算是近年来的研究热点,但大多数研究都是对中科院计算所刘群提出的计算方法的改进和完善。本文充分分析和利用新版Hownet(2007)的概念架构和语义多维表达形式,从概念的主类义原、主类义原框架以及概念特性描述三个方面综合分析词语相似度,并在计算中区分语义特征相似度和句法特征相似度。实验结果理想,与人的直观判断基本一致。更多还原

【Abstract】 Chinese words similarity computing based on Hownet has become a hot point at present,but most research is the improvement and refinement of Liu-Qun’s method[l].Based on the new Hownet(2007),this paper make the best use of Hownet concept frame and semantic multi-dimension expression form,and proposes a new method which analyzes and processes Chinese words similarity from the three dimensions:the main sememe,the main sememe frame and concept characteristic description.Furthermore and the method distinguishes semantic similarity and syntax similarity.The result of experiment shows that the method has a good performance.更多还原