摘要
为了更加客观准确地判断微博的情感倾向,提出表情符向量化算法.首先,该算法将初始化表情符向量从随机产生改进为包含表情符语义信息的向量;然后,用随机产生的负向样本提高泛化能力.通过定性和定量分析可知:该算法能够保留表情符的语义信息;相对于忽略表情符的纯文本情感分析,在微博文本中融入表情符信息的微博情感分析能够提高微博情感分类的精度.
In order to judge the emotional orientation of Weibo more objectively and accurately, an emoticonvectorization algorithm is proposed. Firstly, the initialization emoticon vector isimproved from random generation to a vector containing emoticon semantic information;Secondly, the randomly generated negative samples are used to improve the generalization performance. Through qualitative and quantitative analysis, the algorithm can preserve the semantic information of emoticons. Compared with the plain text sentiment analysis that ignores emoticons, sentiment analysis of Weibo incorporating emoticon information in Weibo text can improve the accuracy of Weibo sentiment classification.
作者
吴晨茜
陈锻生
WU Chenxi;CHEN Duansheng(College of Computer Science and Technology, Huaqiao University, Xiamen 361021, China)
出处
《华侨大学学报（自然科学版）》
CAS
北大核心
2019年第3期399-404,共6页
Journal of Huaqiao University(Natural Science)
基金
国家自然科学基金资助项目(61370006)
福建省科技计划重点资助项目(2015H0025)
关键词
表情符
表情符向量
卷积神经网络
情感分析
微博
emoticon
emoticon vector
convolutional neural network
sentiment analysis
Weibo
