新闻情感倾向性识别算法的研究与应用  

Research and application of an algorithm about news′ emotional tendency recognition

作　　者：周邦定 曹海鹏[1] 张彦[1] 

机构地区：[1]中国科学技术大学信息科学技术学院,安徽合肥230026

出　　处：《微型机与应用》2015年第14期58-60,67,共4页Microcomputer & Its Applications

摘　　要：针对新闻情感倾向性识别问题,提出给单个倾向词分配倾向性、强度、极性和标志4个属性,通过依存句法找出倾向词之间的依存关系,再通过所提出的情感识别算法结合倾向词词典得出整个句子的情感值,最后将整篇新闻中关键句的情感值叠加,得到整篇新闻的情感值。实验结果表明,该方法具有很好的准确率和召回率。For the problem of news＇ emotional tendency recognition, this paper puts forward to allocate four properties-induding orientation, intensity, polarity and flag to a single tendency word. Find out dependency relationship between tendency words by dependency parsing. Through the emotion recognition algorithm proposed in this paper combined with tendency dictionary, whole sentence emotion value will be concluded. Finally, the emotion value of the whole news is got through adding every sentences emotion value. Experiment shows that the algorithm has good precision and recall rate.

关 键 词：依存语法 情感倾向性 浅层语义分析 倾向词搭配 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]