摘要
[目的/意义]微博作为国内的主要社交媒体,对其评论文本进行情感分析有助于微博的深度挖掘。[方法/过程]针对目前文本情感分析中应用较广的机器语言在处理含有连接词句子时所存在的缺陷,对中文连接词制定了处理规则,将表情符号纳入特征向量,并结合情感词典计算情感决策得分,提出了基于语言规则和情感得分的增强监督学习改进模型。[结果/结论]通过实例验证,结果表明改进后的模型可显著提高文本分类有效性。
[Purpose/Significance]As the main social media in China,micro-blog has much influence on public opinion.It is of significance to sentiment analysis on micro-blog reviews since it is useful for deep mining of micro-blog.[Method/Process]This paper aims at the defects existing in the machine language which is widely used in text sentiment analysis to deal with the sentences containing the conjunction,the processing rules of Chinese connectives are formulated,the emoticons are incorporated into the feature vectors,and the emotional decision scores are calculated with emotion dictionaries.[Result/Conclusion]A system improvement measure is proposed,and its effectiveness is verified by an example.
作者
付月
史伟
Fu Yue;Shi Wei(Business School,Huzhou University,Huzhou 313000)
出处
《情报杂志》
CSSCI
北大核心
2018年第12期130-134,167,共6页
Journal of Intelligence
基金
浙江省教育厅资助项目"新常态下高校网络舆情的监测与预警研究"(编号:Y201533615)
浙江省社会科学联研究课题"新常态下高校突发事件网络舆情的监测与预警研究"(编号:2017B49)
浙江省自然科学基金资助项目"大数据背景下基于情感本体的中文微博挖掘:情感分析的视角"(编号:LY15G030023)
浙江省高校重大人文社会科学攻关计划项目"社交化短文本大数据挖掘研究:情感分析的视角"(编号:2016QN003)
关键词
微博文本
意见挖掘
情感分析
增强监督学习
Micro-blog text
opinion mining
affective analysis
enhanced supervision learning
