摘要
针对当前基于《知网》的词义相似度算法未充分考虑义原所具有的情感极性信息,以及未充分利用义原在义项表达式中的位置信息等问题,提出了一种改进的基于《知网》的词义相似度算法。首先,在计算义原相似度时,对现有的考虑义原距离以及义原深度的计算方法,引入义原情感极性作为新的参数,使得对包含情感色彩的词语的词义相似度计算结果更加精确;其次,对义原在义项表达式中的位置信息进行更加深入的分析,提出了一种新的位置权重分配方法,以增强词义相似度计算结果的合理性。实验结果表明,与已有方法相比,所提出的方法可有效提高词义相似度计算的精确度与合理性。
The existing algorithms of word semantic similarity based on HowNet fails to consider the affective polarity of sememes and fails to make full use of the position information of sememes in one expression of semantic item.To solve these problems,an improved algorithm of word semantic similarity based on HowNet was proposed.Firstly,affective polarity of sememe was introduced in calculating the sememe similarity by considering sememe distance and sememe depth,as a new parameter in combination with existing methods,which made the calculation of affective word semantic similarity more exact.Then,by analyzing the position information of sememes in an expression of semantic item,a novel method of assigning position weights of sememes was improved to enhance the rationality of word semantic similarity.Experimental results show that the improved algorithm can effectively improve and enhance the precision and rationality of word semantic similarity calculation compared with existing methods.
出处
《中国科技论文》
CAS
北大核心
2016年第2期202-207,共6页
China Sciencepaper
基金
高等学校博士学科点专项科研基金资助项目(20121102130001)
关键词
人工智能
知网
词义相似度
义原
情感极性
artificial intelligence
HowNet
word semantic similarity
sememe
affective polarity
