Skip to Main Content
In this paper, we take morphemes as the basic tokens and present a fine-to-coarse strategy for Chinese sentence-level sentiment classification. This study involves three parts. First, we employ morphological productivity to extract sentiment morphemes from a sentiment dictionary and to calculate their polarity intensity at the same time. Then, we apply the acquired morpheme-level sentiment information to predict the semantic orientation of sentiment words and phrases within an opinionated sentence. Finally, all the sentiment phrases and their polarity scores are combined to determine the semantic orientation of the sentence. The experimental results on NTCIR-6 OAPT test set show our system can achieve state-of-the-art performance.