摘要
评价对象抽取是情感分析的重要组成部分,针对在线商品中文评论非正规化、网络化的特点,本文提出一种基于句法分析和条件随机场的评价对象的抽取方法,通过实验分析不同模板与不同特征组合对评价对象提取的F值的影响.在系统实现上,主要利用哈工大语言技术平台(LTP)的开放接口和CRFs开源工具对评论数据集进行训练和测试.最终使两类数据集的评价对象抽取的F值达分别达到到82.98%和83.50%.
Extracting object of comment is an important part of emotional analysis. In view of the irregularity of language and the characteristics of network in Chinese online comment, this paper presents a method of extracting objects based on the syntactic analysis and conditional random field. It analyzes experimentally the effect of different templates and different combinations of features on the F value. In the implementation of the system, this paper uses Harbin Institute of Technology language platform open interface and CRFs open source tools to train and test on comment data sets. Finally,the F values of the two types of data sets have reached 82.98% and 83.50% respectively.
出处
《计算机系统应用》
2017年第11期254-259,共6页
Computer Systems & Applications
关键词
评价对象
情感分析
句法分析
条件随机场
特征组合
comment object
sentiment analysis
syntactic analysis
conditional random field
combinations of features
