作者: Ivan Habernal , Tomáš Brychcín
DOI: 10.1007/978-3-642-40585-3_61
关键词:
摘要: This article presents a new semi-supervised method for document-level sentiment analysis. We employ supervised state-of-the-art classification approach and enrich the feature set by adding word cluster features. These features exploit clusters of words represented in semantic spaces computed on unlabeled data. test our three large datasets (Czech movie product reviews, English reviews) outperform current state art. To best knowledge, this reports first successful incorporation based local co-occurrence analysis task.