基于条件随机场的虚假评论识别研究  

Research on Review Spam Recognition Based on Conditional Random Fields

作　　者：陈颖[1,2] 侯惠敏[1,2] 李援南 Chen Ying;Hou Huimin;Li Yuannan(Beijing Electronic Science and Technology Institute Beijing 100070;Xidian University Xi'an 710071)

机构地区：[1]北京电子科技学院,北京100070 [2]西安电子科技大学计算机学院,西安710071

出　　处：《北京电子科技学院学报》2017年第2期47-50,65,共5页Journal of Beijing Electronic Science And Technology Institute

摘　　要：在Web2.0技术日益发达的今天,人们越来越多地依赖网络评论做出交易决策,由此,虚假评论的识别已经变为一种迫切的需求。本文通过分析研究虚假评论与真实评论的区别与联系,提出了基于条件随机场的虚假评论识别算法。该算法不同于已有的基于已有的真假评论集和语料库的虚假评论识别方法,而是通过对评论文本进行特征序列标注,再利用CRF(Conditional Random Fields, CRF)训练识别模型对虚假评论进行识别。实验结果表明,本文所提的算法在虚假评论识别效果上有着不错的表现。AS Web2.0 technology,people rely on online comments to make trade decisions,the identification of false comments has become an urgent need.In this paper,the differences and between false comments and real comments,a review spam recognition based on conditional random fields is proposed.Our algorithm is different from the existing method based on comment set and corpus.We label the characteristics of sequence of the comment text and then use CRF (Conditional Random Fields,CRF)to train the recognition model.Experimental results show that the proposed algorithm performs well in the effect of review spam Recognition.

关 键 词：虚假评论识别 遗传算法 条件随机场 特征提取 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]