Abstract: Sentiment classification focuses on learning a classification model from labeled opinion text, which can predict the sentiment polarity in other opinion text. In this field, one important research topic is to combine the prior knowledge with a generative classifier to construct a new model. By studying the domain character and weight of sen¬ti¬ment lexicons, this paper proposes an approach to automatically construct domain-specific sentiment lexicons which can be reformulated as a generative prior, and then combine it with a generative model. Experimental results show that the proposed generative model performs significantly and consistently better than some of the-state-of- the-art do-main-independent methods.

Key words: sentiment classification, prior knowledge, domain-specific sentiment lexicon, sentiment lexicon

摘要： 情感分类是通过分析数据中的情感信息, 来预测数据所传递的情感倾向。其中结合语言学词典与产生式分类器构造带有先验知识的分类模型, 是一类重要的研究课题。通过研究情感词的领域性和不同权重的特性, 提出了一种新的融入情感先验知识的情感分类方法。通过自动分析构造领域相关的情感词及其权重信息, 将其作为情感先验知识, 融入到产生式分类模型中, 得到更适合特定领域的分类模型。实验结果表明, 该方法在分类性能上, 显著并一致地优于其他结合了领域无关先验知识的产生式分类模型。

关键词: 情感分类, 先验知识, 领域情感词, 情感词