摘要： 传统情感分析方法仅考虑单一文本,对长度短小且口语化严重的微博文本情感极性识别率较低。针对上述问题,提出一种结合上下文消息的情感分析方法。将微博情感分析问题看做标签序列学习任务,使用隐马尔可夫支持向量机把微博上下文语境融入微博情感分析问题中。实验结果表明,该方法较之于基于朴素贝叶斯或支持向量机的微博情感分析模型可以更好地分析微博情感极性。

关键词: 中文微博, 情感分析, 上下文语境, 文本表示, 标签序列学习

Abstract: Traditional sentiment analysis method only considers a single text,and the recognition rate of sentiment polarity is low when the micro-blog text is short and colloquial.In order to solve this problem,a kind of sentiment analysis method combined with linguistic context is proposed.Regarding micro-blog sentiment analysis problem as a tag sequence learning task,this paper uses Hidden Markov Support Vector Machine(SVMhmm) to blend the micro-blog context into micro-blog sentiment analysis.Experimental results show that this method can analyse the micro-blog sentiment polarity better than the Naive Bayes or Support Vector Machine(SVM) micro-blog sentiment analysis model.

Key words: Chinese micro-blog, sentiment analysis, linguistic context, text representation, label sequence learning

中图分类号: 

• TP391