公众文本之情感词典研究进展

Progress of generating sentiment lexicons from text in social media

针对海量的社交媒体数据进行情感分析,能够实时检测与跟踪公众对社会事件、政治活动、公司战略、重大决策等方面的观点,同时了解用户在其评论、博客、微博等文本中的情感倾向.本文首先论述了从作者角度出发、针对文本中的主观词等构建的文本情感词典研究现状,包括基于辞典的和基于语料库的生成方法及其典型应用.然后阐述了从读者角度出发、通过读者对文档的情感反馈而构建的公众情感词典研究进展,包括其数据来源,以及词层和主题层的公众情感词典生成方法和模型.

Sentiment analysis for big social media data could not only detect and track public opinions about social events, political movements, company strategies and decisions in real time, but also understand users emotions which are expressed through reviews, blogs, microblogs/tweets and other text. In this article, we first present the progress of generating sentiment lexicons in text with subjective words from the writer＇s perspective, including the thesaurus-based and corpora-based methods and typical applications. Then, the progress of generating social emotion lexicons by tile emotional responses of readers from the reader＇s perspective is reviewed, which includes tile data sources, word-level and topic-level models.