基于改进的义原关联度算法的词语相关度计算

摘要

词语的语义相关度计算主要应用于语义消歧、信息检索等领域.语义相关度计算的研究一般基于知网,把义原和解释义原之间的横向关系作为义原关联度,进而得出语义相关度.基于以上研究,利用知网中解释义原所构成的横向联系,提出将解释义原之间的关系作为义原关联度,计算各义原的解释义原之间相似度,把其中的最大值作为义原关联度.实验结果表明,在运算量相当的情况下,算法得到的语义相关度与人们的直觉更加相符.%The algorithm of semantic relevancy is mainly applied to solve the problems existed in such fields likerndisambiguation and information retrieval. The research on the algorithm of semantic relevancy is generally based on HowNet, to take the transverse relationship between sememes and their interpretations as sememes' relevancy, thus leading to the semantic relevancy. Based on the above-mentioned research, the transverse relationship formed in interpreting sememes in HowNet was used to bring forward the notion that to take the relationships between the sememes' interpretations as sememes' relevancy, computing the similarities of the interpretations of each sememe then the maximum value is regarded as the sememes' relevancy. The experimental result indicates that with the operational volume being equal, the semantic relevancy computed by the stated method matches people's intuitions to a much larger scale.