中文句子倾向性分析

Opinion analysis of the Chinese sentence

针对句子的倾向性进行判断,采用SentiWordNet构建中文倾向性词表,通过剔除停用词等降低句子向量的维数,以此来提高句子向量化速度,然后利用支持向量机分类器进行句子倾向性判断,最后提出两种新的置信度计量方法对倾向性句子进行排序.实验结果表明,构建的识别系统在一定程度上能有效识别倾向性句子.

To judge the sentence＇ s opinion, we adopt SentiWordNet to build Chinese opinioned vocabulary, eliminate stop words to reduce the dimension of the sentence vector, use this method to enhance the speed of structuring sentence＇ s vector, apply support vector machine （SVM） classifier to determine whether the sentence having opinion, propose two new methods of measuring confidence and sort the opinioned sentences by the confidence score. The experimental result shows that our system can effectively detect opinion sentences in some extent.