基于有监督分词方法的维吾尔文情感分析

摘要

To improve the accuracy of Uyghur sentiment analysis,a design method of supervised word segmentation based on information gain was presented.The supervised word segmentation method was used in sentiment analysis,which avoided the curse of dimensionality and meaningless feature.Results of experiment show that the feature space is smaller and the model trained on the feature space can get higher accuracy using the proposed feature extraction method.%为提高雏吾尔文情感分析的准确率,提出一种基于信息增益的有监督雏吾尔文分词方法,并将其用在情感分析中,避免传统空格分词方法造成的雏数灾难和特征项语义不完整等问题.实验结果表明,用该分词方法得到的特征空间规模更小,在此特征空间上训练出来的模型性能更好,能够有效提高雏吾尔文情感分析的准确率.