摘要
知网是一个英汉双语本体,含有丰富的语义知识。在综合考虑了知网中义原重合度、义原差度、层次深度等因素的基础上,提出了一种新颖的义原相似度计算方法,并在义原相似度计算的基础上,通过改进的匈牙利算法来计算基于知网的概念相似度,最后通过实验验证算法的有效性,与基于WordNet的方法相比,文中提出的基于知网的相似度计算一样可以取得较好的精确度。
HowNet is a Chinese-English bilingual ontology which provides plenty of knowledge. In this paper, we synthetically consider the factors of sememes superposition, sememes difference and hierarchical depth in HowNet and propose a novel approach to calculate the sememes similarity for HowNet. Base on the sememes similarity, we calculate the HowNet-based similarity between concepts using improved Hun- garian algorithm. The experiments are done to evaluate the validity of our approach. Compared with other method, our HowNet-based approach can measure the similarity between concepts as well as those Word- Net-based methods.
出处
《南京邮电大学学报（自然科学版）》
2011年第6期60-66,共7页
Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition
基金
江苏省博士后科研基金(1101087C)资助项目
关键词
本体
知网
义原相似度
概念相似度
匈牙利算法
ontology
HowNet
sememes similarity
concept similarity
improved Hungarian algorithm
