中文摘要：

基于微博的情感分析近几年获得了广泛的关注,但是通常是对微博上的电影或者产品的评论。我们的研究目标是针对微博上的热点事件的分析,将这些事件的评论分为正向或者负向,将微博用户的评论进行分类有利于辨别公众对于这个事件的普遍看法。本文提出了一个基于卷积神经网络的混合模型：CNN-SVM,用于对事件的评论进行分类。其次,根据微博用户的转发行为,我们提出了一种新的数据结构——转发树,用以解决评论中的一些分类混乱情况。实验结果表明CNN-SVM确实提高了情感分类的正确率,新提出的转发树结构在对真实世界的情感倾向性的逼近中也是十分有效的。

英文摘要：

Sentiment classification on weibo recently attracted research community the widespread attention. Most previous works focused on weibo comments about movies or products. In contrast, our study aims at the hot events on weibo. Comments of the events are considered either positive or negative on behalf of the user＇ s attitude to these events. Classification of user attitude helps to identify the opinions of the general public. In this paper, we put forward an innovative method based on convolution neural network, termed as CNN - SVM, to classify the comment. In addition, according to the forwarding behavior of users, we put forward a new data structure, repost tree for dealing with ambiguity in the comments. Extensive experiments demonstrated that the CNN-SVM method effectively improved the accuracy of events sentiment classification. The new data structure showed to be effective on steering the classification results towards real world sentiment tendency.