论文部分内容阅读
随着社交媒体和网络的迅速发展,互联网(如微博和论坛)上产生了大量用户参与的对于诸如人物、事件和产品等的评论信息,这些信息成为分析人们情感和观点的重要材料。本文将基于传统语言特征的分析方法探讨对这些非正式文本情感分析的局限性。受各种社会认知理论的启发,我们结合局部语言特征和全球性社会惯例来改善情感分析的方法。在没有使用任何额外标记数据的基础上,使用这种新方法对美国总统选举领域的多类型文本情感进行分析,其效果显著。
With the rapid development of social media and internet, a great number of users participate in the comment information such as people, events and products on the Internet (such as Weibo and Forum), which are important materials for analyzing people’s feelings and opinions. This article explores the limitations of emotional analysis of these informal texts based on the analysis of traditional language features. Inspired by various theories of social cognition, we improve the methods of sentiment analysis by combining local linguistic features and global social conventions. The use of this new method to analyze the multi-type text sentiment in the US presidential election is notable without the use of any additional markup data.