面向新闻的情感关键句抽取与极性判别  

News-oriented Emotional Key Sentence Extraction and Polarity Determination

作　　者：赵虹杰 刘华丽[1] 任巨伟[1] 林鸿飞[1] 

机构地区：[1]大连理工大学信息检索研究室,大连116023

出　　处：《山西大学学报（自然科学版）》2014年第4期588-594,共7页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金(60973068);辽宁省自然科学基金(201202031)

摘　　要：门户网站、博客和论坛中的新闻性文章很多具有倾向性,倾向性判别对了解社会动态和舆情状况有重要作用。在主题相关性的基础上,主要考虑了三类属性:位置属性、情感属性、特征词属性,提出了一种针对篇章级的情感关键句抽取方法,并通过集成学习判别情感关键句的极性。实验结果显示本文方法能够有效地挖掘出情感关键句并能对情感关键句进行较准确的极性判别,实现了情感关键句,抽取和极性判别的自动化,且具有较好的效果。The great majority of news articles have tendentious in portal websites,blogs and forums.Judging the tendentious of news articles plays an important role in understanding of the social dynamics and the public opinion situation.In this paper,based on the relevance of theme,we consider three kinds of attributes：location attribute,emotional attribute,characteristic words attribute.Then we propose a kind of emotional key sentence extraction method in chapter level and judge the polarity of emotional key sentence by integrated learning.The results of experiment show that the proposed method can effectively extract emotional key sentence and accurately determine the polarity of emotional key sentence.The proposed method achieves the automation of the emotional key sentence extraction and the polarity discrimination,and the results of the method have good effect.

关 键 词：新闻 情感关键句 情感判别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]