摘要
针对负面新闻判定问题,现有的构建分类器的统计方法和抽取情感特征的语义分析方法,以其各自存在的局限性,不适用于通用的文本倾向性识别.据此,提出一种融合依存语法和简化的格语法框架理论,结合情感词典对关键句群进行主题相关的语义倾向性分析,进而判定负面新闻的方法.该方法通过依存句法分析识别句中词对间的依赖关系,借用其分析结果,辅助填充基于格语法定义的模板框架槽,可解决单纯使用格语法因标注词典困难而难以实用化问题.实验数据表明,将本方法应用于识别特定主题的负面新闻时,处理速度快、准确性高,具有很好的实用性.
For the problem of negative news judgment, there are two typical methods, one is statistical method based on classifiers, and the other is semantic analysis method to extract the affective characteristics. Due to the limitations of the both algorithms, it cannot be used to the application of the general text orientation recognition. Accordingly, an algorithm combines the dependency grammar and the simplified case grammar was proposed, which is used to the negative news judgment by specific semantic analysis. This algorithm analyzes the interdependence of words through dependency parsing, and then fills in the template framework based on case grammar with the result. Experiments show that this method has better performance in accuracy and efficiency.
出处
《小型微型计算机系统》
CSCD
北大核心
2015年第5期1047-1051,共5页
Journal of Chinese Computer Systems
关键词
依存关系链
语义格
负面新闻
chain of dependencies
semantic case
negative news
