面向微博电影评论的情感分类研究

发布时间：2018-05-12 20:08

本文选题：电影评论 + 情感分析　；参考：《云南财经大学》2014年硕士论文

【摘要】：随着web2.0的发展，微博的出现不断地改变着人们的生活方式。由于其强大的影响力和渗透力，现在越来越多的人喜欢通过微博发表电影评论。微博电影评论是观众对电影好坏的一种情感表达，对这些信息进行情感分类研究，不仅有助于观众决策，选择好的电影，同时也能够使制片商及时获取大众对电影的反应，调整相应的营销策略，从而提高电影票房成绩。微博电影评论是电影评论在社交网络平台上存在的一种新模式。现有的电影评论研究大多数是针对传统电影评论。传统电影评论主题单一，且篇幅较长，微博电影评论与其不同。因此本文根据微博电影评论的特点，对其情感分类进行研究，并开展了以下几个方面的研究内容：一、通过对大量微博电影评论统计和分析，在知网词典的基础上，构建一个电影领域情感词典，用于微博电影评论情感分类。二、根据主题发散这一特点，提出一种基于主题情感句提取的微博电影评论情感分类方法。该方法分为三步：第一步，提取主题相关句，将主题无关的句子从评论文本中剔除；第二步，，主客观分类，即从余下的主题相关内容中，找出主题情感句，去除客观性的句子；第三步，情感分类，采用机器学习方法对最终的评论文本进行分类，获得其情感倾向。同时在这一过程中，对零指代句子利用依存句法分析方法进行消除。三、提出了一种基于主动学习和协同训练的半监督情感分类方法。在网上获取未标注的微博电影评论语料很容易，但想要得到大量的标注语料，则需要消耗很多的人力和时间。为了减小人工标注的工作量，文本采用半监督方法，并在协同训练框架的基础上，引入主动学习思想，从而改善分类器的性能，提升分类的准确率。
[Abstract]:With the development of web2.0, the appearance of Weibo is changing people's way of life. Because of its powerful influence and penetration, more and more people now like to post movie reviews through Weibo. Weibo film review is a kind of emotional expression of good or bad movies. The research on the emotional classification of this information not only helps the audience to make decisions and choose good movies, but also enables the producers to obtain the public's response to the movies in time. Adjust the corresponding marketing strategy, so as to improve the film box office results. Weibo movie review is a new mode of movie review on social network platform. Most of the current research on film reviews is aimed at traditional film reviews. The traditional film review has a single theme and long length, and Weibo film review is different from it. Therefore, according to the characteristics of Weibo movie review, this paper studies its emotional classification, and carries out the following research contents: Firstly, based on the statistics and analysis of a large number of Weibo movie reviews, a film domain emotion dictionary is constructed on the basis of the Chih-net Dictionary, which is used to classify the emotion of Weibo movie reviews. Secondly, according to the characteristic of subject divergence, a method of emotion classification of Weibo film review based on subject emotion sentence extraction is proposed. The method is divided into three steps: first, the topic related sentences are extracted, and the topic independent sentences are removed from the comment text, the second step is the subjective and objective classification, that is, the topic emotion sentences are found out from the remaining subject related contents, and the objective sentences are removed. The third step is emotion classification. Machine learning is used to classify the final comment text to obtain its emotional tendency. At the same time, the zero-anaphora sentence is eliminated by the method of syntactic analysis. Thirdly, a semi-supervised emotion classification method based on active learning and cooperative training is proposed. It is easy to get the untagged Weibo review data on the Internet, but it takes a lot of manpower and time to get a large amount of annotated data. In order to reduce the workload of manual annotation, the text adopts semi-supervised method, and on the basis of collaborative training framework, the active learning idea is introduced to improve the performance of classifier and improve the accuracy of classification.
【学位授予单位】：云南财经大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.092

【参考文献】