基于依存关联分析的情感词扩展  

Opinion Words Expansion Based on Dependency Parsing

作　　者：孙艳[1] 周学广[1] 付伟[1] 

机构地区：[1]海军工程大学电子工程学院,武汉430033

出　　处：《北京邮电大学学报》2012年第5期90-93,共4页Journal of Beijing University of Posts and Telecommunications

基　　金：国家自然科学基金项目(611100042)

摘　　要：无扩展的情感词识别方法对于倾向容易发生演化的社区媒体信息的召回率很低.基于此,提出了一种基于依存关联分析的情感词扩展识别方法,首先对文本进行新词和短语发现;再对句子的依存树进行剪枝和归并;然后按照给定的策略进行情感词和评价对象的扩展;最后得到情感词扩展的识别结果.该方法能同时扩展具有直接依存关系和间接依存关系的情感词.实验结果表明该方法有效.The forgoing experiments have shown that the method of opinion words identification with non-propagation is extremely limited on social medium information.A method of opinion words expansion identification based on dependency parsing is presented.Through new words and phrases discovery,pruning and merger on dependency tree,the opinion words are expanded.A key advantage of the proposed method is that it can expand words in direct dependency and indirect dependency.Experiments verify the feasibility of this method.

关 键 词：依存分析 情感词扩展 倾向分析 舆情 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]