Abstract: Text sentiment classification is used widely, such as information filtering, information security and information recommendation. This paper proposed an improved method based on semantic comprehension. In this paper, sentiment primitive was introduced into sentiment terms identification, and then semantic values of phrases were obtained.Furthermore, this paper further analysed adverbs and its influence on identification of text orientation in the semantic level and achieved the text sentiment classification. I}he experimental results show that the proposed approach is suitable for judging sentiment orientation.

Key words: Text sentiment classification, Sentiment orientation, Semantic comprehension