[目的/意义]微博评论情感分类模型可以为相关舆情监管部门正确管控话题事件的发展状况和舆情提供一定的指导作用。[方法/过程]基于字词向量的多尺度卷积神经网络，运用多尺度卷积核改善微博评论中上下文信息有限的条件制约，构建基于字词向量的多尺度卷积神经网络微博评论情感分类模型；通过爬取"微博热搜整改"数据，对模型的可行性和优越性进行验证。[结果/结论]验证结果表明基于字词向量的多尺度卷积神经网络在微博舆情等上下文信息有限的短文本分类任务中表现良好。本文在理论层面为微博舆情情感分类提供了更为准确的情感分类理论模型及分类方法，在实践层面可以更好地指导舆情监管部门对舆情的情感倾向进行更好的引导和监管。
[Purpose/significance] The microblog commentary sentiment classification model can play a guiding role for the relevant public opinion supervision departments to correctly control the development of the topic events and the public opinion.[Method/process] Based on the multi-scale convolutional neural network of word vector, this paper used multi-scale convolution kernel to improve the conditional constraints of finite context information in microblog commentary, and constructed multi-scale convolutional neural network microblog commentary emotion classification model based on word vector. Finally, the feasibility and superiority of the model were verified by crawling the real data of "microblogging hot search and rectification".[Result/conclusion] Verification results show that the multi-scale convolutional neural network based on word vector performs well in the short text classification task with limited context information such as weibo public opinion. On the theoretical level, this paper provides a more accurate emotional classification theory model and classification method for microblogging public opinion emotion classification. In practice, it can better guide the public opinion supervision department to better guide and supervise the emotional sentiment of public opinion.
[1] 新浪科技.微博发布2018年第二季度未经审计财报[EB/OL].[2018-10-08]. https://tech.sina.com.cn/i/2018-08-08/doc-ihhkuskt9159883.shtml.
[2] NEVES-SILVA R, GAMITO M, PINA P, et al. Modelling influence and reach in sentiment analysis[J]. Procedia CIRP, 2016, 47:48-53.
[3] SOLEYMANI M, GARCIA D, JOU B, et al. A Survey of Multimodal Sentiment Analysis[J]. Image & vision computing, 2017,65(9):3-14.
[4] ÖZTURK N, AYVAZ S. Sentiment analysis on Twitter:a text mining approach to the Syrian refugee crisis[J]. Telematics & informatics, 2018,35(4):136-147.
[5] 唐晓波,兰玉婷.基于特征本体的微博产品评论情感分析[J].图书情报工作,2016(16):121-127.
[6] 马松岳,许鑫.基于评论情感分析的用户在线评价研究——以豆瓣网电影为例[J].图书情报工作,2016(10):95-102.
[7] 梁晓敏,徐健.舆情事件中评论对象的情感分析及其关系网络研究[J].情报科学,2018,V36(2):37-42.
[8] 唐慧丰,谭松波,程学旗.基于监督学习的中文情感分类技术比较研究[J].中文信息学报,2007,21(6):88-94.
[9] 张冬雯,杨鹏飞,许云峰.基于Word2vec和SVMperf的中文评论情感分类研究[J].计算机科学,2016,43(6):418-421.
[10] NIHAR M. R, RAJESH S. P. LFNN:Lion Fuzzy Neural Network-based evolutionary model for text classification using context and sense based features[J]. Applied soft computing,2018,71:994-1008.
[11] 冯兴杰,张志伟,史金钏.基于卷积神经网络和注意力模型的文本情感分析[J].计算机应用研究,2018,35(5):1434-1436.
[12] ZHANG X, ZHAO J, LECUN Y. Character-level convolutional networks for text classification[J]. Computer science, 2015(9):1-9.
[13] 庞俊.基于确定话题和情感极性的博客文本聚类研究[D].武汉:武汉理工大学,2010.
[14] 王晰巍,张柳,文晴.基于贝叶斯模型的移动环境下网络舆情用户情感演化研究——以新浪微博"里约奥运会中国女排夺冠"话题为例[J].情报学报,2018,37(12):1241-1248.
[15] ZHANG Q, GAO Z, LIU J. Research of Weibo short text classification based on Word2vec[J]. Netinfo security, 2017(1):57-62.
[16] 周瑛,刘越,蔡俊.基于注意力机制的微博情感分析[J].情报理论与实践,2018,41(3):89-94.
[17] 控杏,林庆.主观性文本情感分类研究综述[J].信息技术,2018(8):126-130.
[18] Word2vec.[EB/OL].[2018-10-08]. https://en.wikipedia.org/wiki/Word2vec.
[19] 周练.Word2vec的工作原理及应用探究[J].图书情报导刊,2015(2):145-148.
[20] 卢宏涛,张秦川.深度卷积神经网络在计算机视觉中的应用研究综述[J].数据采集与处理,2016,31(1):1-17.
[21] 吴鹏,刘恒旺,沈思.基于深度学习和OCC情感规则的网络舆情情感识别研究[J].情报学报,2017,36(9):972-980.
[22] REN Y, JI D, REN H. Context-augmented convolutional neural networks for twitter sarcasm detection[J]. Neuro computing, 2018,308(9):1-7.
[23] 于政.基于深度学习的文本向量化研究与应用[D].上海:华东师范大学,2016.
[24] 王毅,谢娟,成颖.结合LSTM和CNN混合架构的深度神经网络语言模型[J].情报学报,2018(2):194-205.
[25] SMITH L N. Cyclical learning rates for training neural networks[J]. Computer science, 2017:464-472.
[26] 2017年度十大网络用语.[EB/OL].[2019-02-22].https://baike.baidu.com/item/2017年度十大网络用语/22273910?fr=aladdin.
[27] 陈珂,梁斌,柯文德,等.基于多通道卷积神经网络的中文微博情感分析[J].计算机研究与发展,2018,55(5):945-957.
[28] 曲之琳,胡晓飞.基于改进激活函数的卷积神经网络研究[J].计算机技术与发展,2017,27(12):77-80.
