An Analysis and Comparison of Three Methods for Document Semantic Orientation Recognition
三种文档语义倾向性识别方法的分析与比较*

This paper investigates and implements three methods for document semantic orientation recognition: one is based on the weighted emotional words; another is based on the analysis of the semantic pattern; and the third is based on text classification. The first method is try to use semantic orientation of feature words. The second method is to simplify the structure of the natural language syntax, in order to acquire the appropriate size patterns with semantics tendency. The third method is based on the direct use of traditional text classification method. Through experiments and the application in an analysis system of net-mediated public opinions, shortages and advantages of the three methods are proposed.