关键词: 主观性句子识别, 观点挖掘, 概率主题模型, 弱监督

Abstract: Subjectivity sentence identification aims to detect the opinionated sentences in text. This paper proposes mixing topics and subjectivity sentence identification model based on probabilistic topic model. Through considering the topics， the model can detect the subjective sentences, and can also extract the subjective topics from texts simultaneously. The proposed model is a weaklysupervised generative model, which only needs a small set of domain independent subjectivity lexicon to modify prior of model. The experiment results demonstrate that the model can highly improve the sentence subjectivity identification recall and the Fvalue, and the extracted subjectivity topics are semantically informative.


Key words: subjectivity sentence identification, opinion mining, probabilistic topic model, weakly-supervised