摘要
《知网》是一部比较详尽的语义知识词典,含有丰富的语义信息和世界知识,研究其特定的语言描述方法是科学合理利用《知网》的关键.《知网》的知识系统描述语言方式也在不断改进和改善,已经从过去的线性方式发展为现在的立体、可嵌套的方式.因此,本文对《知网》2008版本所使用的描述方式进行介绍,并基于此描述方式实现了两种词语相似度计算方法.其中,方法一的设计思想源于《知网》2000版本的计算方法;方法二是对现有《知网》2008版本的计算方法进行了适当改进.实验结果表明,与现有《知网》2008版本的词语相似度计算方法进行比较,运用本文所实现两种方法,一些词语的相似度计算结果更为合理,从而更加符合人们的主观判断.
＂HowNet＂ is a detailed semantic knowledge dictionary, which contains rich semantic information and world knowledge. Making full use of its specific description is the key issue to use the ＂HowNet＂. The Knowledge Database Mark-up Language of the ＂HowNet＂has been constantly improved, which has alreadybeen changed from original linear way to stereoscopic and nested description method. Therefore, this paper introduces the description method of ＂HowNet＂2008 version, and achieved two kinds of word similarity computation methods based on the description. Among them, the first calculation method designed from the ＂HowNet＂ 2000 versions, and the second method is properly improved of the ＂HowNet＂ 2008 version. The experimental results show that the similarity calculation results of some words are more reasonable than the existing ＂HowNet＂ version 2008, which are more in line with the sub- iective judgment of people.
出处
《小型微型计算机系统》
CSCD
北大核心
2015年第8期1728-1733,共6页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61371194)资助
北京市自然科学基金项目(4152012)资助
北京市重点建设学科"计算机应用技术"和"通信与信息系统"资助
