摘要： 汉语词语情感倾向自动判断避免了个人判断的影响，并提高了主观性词典创建效率。 讨论和分析汉语词语情感倾向判断技术，使用情感特征集合进行倾向性描述，建立基于二元语法依赖关系的情感倾向互信息特征模型。采用机器学习方式得到分类器，对词语的情感倾向进行自动判别，并进行比较和优化，性能得以提高，最好的SVM准确率达到95.47%，F值达到93.90%。采用特征集合描述情感倾向性，在建立的互信息特征模型上，使用机器学习方法自动判断词语情感倾向是有效的。

关键词: 自动判断, 特征选择, 机器学习, 情感分析, 倾向

Abstract: The Chinese word sentiment polarity automatic judgment can avoid artificial error and improve the efficiency of the subjective lexicon creation. The technology of the Chinese word sentiment polarity judgment is discussed and analyzed. The polarity is described by using the sentiment characteristics set. The model of the sentiment polarity mutual information characteristics is created based on the bigram dependency of POS tagging. The classifier is available by machine learning to automatically judge, compare and optimize the word sentiment polarity. All of these help to improve the properties, the highest accuracy of SVM reaches 95.47%, and the F value is up to 93.90%. So it is effective to describe the sentiment polarity by using characteristic set and to automatically judge the word sentiment polarity by machine learning and based on the mutual characteristics model.

Key words: automatic estimation, feature selection, machine learning, sentiment analysis, polarity

中图分类号: 

• TP311.52