Paradigm words selecting method based on word clustering for sentiments analysis
面向倾向性分析的基于词聚类的基准词选择方法*

This paper put forward a method of selecting paradigm words, which was based on the existing randomness and subjectivity issue.Firstly,it expanded words by a group of selected initial seed words;secondly,it obtained the second generation of seed words by means of hierarchical clustering.According to the similarity between two different expanded words, then it expanded and clustered the second generation seed words.At last it orderly iterated by same procedure to get the optimal clustering seed words as the final selected paradigm words.The experiment result indicates that the new method has a higher accuracy in selecting paradigm words while classifying the different emotional proclivities.