结合图片语义规则和机器学习的情感分类方法  

SENTIMENT CLASSIFICATION BASED ON IMAGE SEMANTIC RULES AND MACHINE LEARNING

作　　者：陈新元 谢晟祎[3] 陈庆强 张力 Chen Xinyuan;Xie Shengyi;Chen Qingqiang;Zhang Li(Department of Information Engineering,Fuzhou Melbourne Polytechnic,Fuzhou 350121,Fujian,China;College of Computer and Control Engineering,Minjiang University,Fuzhou 350121,Fujian,China;Experimental Training Center,Fujian Vocational College of Agriculture,Fuzhou 350181,Fujian,China;Information Science and Engineering College,Fujian University of Technology,Fuzhou 350118,Fujian,China;Liberal Arts College,Fuzhou Institute of Technology,Fuzhou 350506,Fujian,China)

机构地区：[1]福州墨尔本理工职业学院信息工程系,福建福州350121 [2]闽江学院计算机与控制工程学院,福建福州350121 [3]福建农业职业技术学院实验实训中心,福建福州350181 [4]福建工程学院信息科学与工程学院,福建福州350118 [5]福州理工学院文理学院,福建福州350506

出　　处：《计算机应用与软件》2021年第7期173-181,共9页Computer Applications and Software

基　　金：福建省教育科学“十三五”规划2020年度课题(FJJKCG20-402)。

摘　　要：传统微博情感分类方法往往忽略对大量表情符号和图片型表情的情感分析。在图片去重算法的基础上,利用表情词和表情符号的相关性构建表情图片情感词典,并加入对官方扩展表情包和图片型表情的情感识别与验证。由于单纯词典方法的性能有限,将词典抽取的规则情感特征与机器学习的基本特征模板融合,使用朴素贝叶斯等分类器,提高情感分类的性能表现。实验结果表明,图片情感词典的加入使方案的准确率有所提高,特别在处理情感词表达不规范的样本方面具有一定优势。A large number of emoticons and images with emotions are often left out in traditional researches on sentiment classification.On the basis of image deduplication algorithm,a dictionary of images with emotions was constructed with the help of the correlation between images and emoticons or emotional words.The emotion recognition and verification of official extended emoticons and images were added.The extracted emotional features and basic feature templates were integrated to break the limits of emotion classification approaches based on dictionaries.With Naive Bayes classifier,the accuracy of emotion classification was improved by cross validation.The experimental results show that the addition of image emotion dictionary enhances the scheme’s performance on accuracy,especially when dealing with the samples containing emotional words in deformity.

关 键 词：情感分类 图片情感 机器学习 朴素贝叶斯 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]