Comprehensive Information Based Semantic Orientation Identification

The identification of the texts´ attitude orientation is to identify the opinion on a special issue in a text is supporting or opposing. This paper introduces a method to identify the orientation of a Chinese text´s attitude using the comprehensive information acquired form the text. First, we label the text with information in several layers, which include information of segmentation, POS, semantic role and semantic emotional orientation. We add these information to a text step by step, and make use of them to identify the attitude orientation of the text in a given topic. Experiments show that, the orientation identifier which uses above comprehensive information from syntactic, semantic and pragmatic levels has better performance in this task than some others.