摘要
主要对跨领域中文评论句中的各个评价对象所对应的观点表达的情感倾向进行研究。在结合单一领域特别是产品领域中情感分类的常用算法以及结合跨领域评论观点表达的特殊性的基础上,提出了基于词典资源和有监督机器学习这两种方法来对跨领域中文评论句进行情感分类,探讨了跨领域中文评论在算法上与单一领域的异同,同时对两种方法进行了比较。实验结果表明,提出的方法具有较大的实用价值。
This paper aimed the research on opinions' polarity of each object on Chinese comments of universal areas. Based on the freqently-used sentiment analysis algorithm for single area ( especially products area) and the particularity of cross-do- main opinion, it proposed two methods to determine the polariy of opinions-dictionary resource method and supervised machine learning method. Meanwhile, it also discussed the algorithm difference between cross-domain and single area and compare the two oronosed methods. ExrJerimental results show that the proposed method has great practical value.
出处
《计算机应用研究》
CSCD
北大核心
2013年第3期736-741,共6页
Application Research of Computers
基金
国家社会科学基金资助项目(11CYY031)
国家自然科学基金资助项目(61170180)
关键词
跨领域
情感分类
知网
有监督机器学习方法
支持向量机
cross-domain
sentiment analysis
HowNet
supervised machine learning method
support vector machine ( SVM )