CAJ | 학술논문

文中旨在对中文微博文本中表达的情绪进行自动分析.目前,微博情绪分析的方法主要是平面型分类方法.该方法认为各个情绪类之间相互独立,相互并列,它们处在同一个平面层次上,只需要一次性构建一个分类器就可以完成情绪分类任务.事实上,Ekman六类情绪之间的关系并不完全独立.文中将Ekman六类情绪按照情感极性及情绪间的相互关系组织成三层树状结构,在此基础上提出了一种基于朴素贝叶斯模型的多层次中文微博情绪分析方法.实验结果表明,与传统的平面型朴素贝叶斯分类方法相比,文中提出的多层次微博情绪分析方法降低了各情绪类微博分布不平衡对分类结果造成的影响,提高了微博情绪识别的精度.
문중지재대중문미박문본중표체적정서진행자동분석.목전,미박정서분석적방법주요시평면형분류방법.해방법인위각개정서류지간상호독립,상호병렬,타문처재동일개평면층차상,지수요일차성구건일개분류기취가이완성정서분류임무.사실상,Ekman륙류정서지간적관계병불완전독립.문중장Ekman륙류정서안조정감겁성급정서간적상호관계조직성삼층수상결구,재차기출상제출료일충기우박소패협사모형적다층차중문미박정서분석방법.실험결과표명,여전통적평면형박소패협사분류방법상비,문중제출적다층차미박정서분석방법강저료각정서류미박분포불평형대분류결과조성적영향,제고료미박정서식별적정도.
Focus on automatic classifying emotion expressed in Chinese microblog. At present,most of the research on emotion analysis of microblog has focused on flat classification approach. The approach considers emotion categories are independent and coordinate of each other. They are in the same flat level and need only one classifier to achieve the emotion classification task. In fact,the relationship be-tween Ekman' s six emotion classes is not completely independent. Ekman' s six emotion classes are arranged into a tree structure accord-ing to the relationship between emotional polarity and mutual relations. Based on this tree structure,an approach of hierarchical emotion a-nalysis of Chinese microblog based on Na?ve Bayesian Model is proposed. Compared with the flat classification,the experimental results show that hierarchical classification method reduces the effect of the highly imbalanced distribution of the emotional categories on the classification results,and improves the accuracy of microblog emotion analysis.