基于循环层叠条件随机场的评价对象识别  

Evaluated object recognition based on recurring cascaded conditional random field

作　　者：霍立平[1] 

机构地区：[1]河北中医学院计算机教研室,河北石家庄050200

出　　处：《兰州理工大学学报》2018年第1期104-108,共5页Journal of Lanzhou University of Technology

基　　金：国家自然科学基金(61100009);河北省社会科学发展研究课题(201603010120);河北省高等学校人文社会科学研究青年基金(SQ161073)

摘　　要：随着微博、论坛等社交平台的兴起,如何针对观点句进行评价对象识别已经成为了自然语言处理领域研究的热点之一.传统的评价对象识别技术在一定程度上取得了不错的效果,但其在模型优化、新词覆盖度和准确率方面仍有很大不足.因此,提出一种全新的循环层叠条件随机场模型,融合基于词性与句法语义的候选评价对象抽取方法提高新词覆盖度,并采用支持向量机对评价对象进行筛选过滤以提高准确率.实验结果表明,基于循环层叠条件随机场的评价对象识别方法可以取得较好的效果.With the rise of social networking platforms such as micro-blog and forums,how to perform evaluated object recognition of opinion sentences has become one of the hot spots in the field of natural language processing,that containing a huge political and commercial value.Traditional evaluated object recognition technique has achieved pretty good result to a certain extent,but it still has a lot of deficiencies in connection with model optimization and coverage degree and accuracy of out-of-vocabulary words.Therefore,a totally novel recurring cascaded conditional random field model is put forward,that integrating the lexical feature and syntactics with semantics for candidate evaluated object recognition to improve the outof-vocabulary words coverage degree,and the support vector machine（SVM）is adopted for candidate evaluated object filtration in order to improve the accuracy rate.The experimental result shows that the method of evaluated object recognition with the model of recurring cascaded conditional random field will achieve a better result.

关 键 词：评价对象 循环层叠条件随机场 句法语义 支持向量机 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]