摘要
随着移动互联网等信息技术的发展与普及,社交网络中的谣言和虚假新闻的传播也更迅速、影响也更广泛,造成了不良的社会影响与经济损失,同时对监管部门的响应速度提出更高要求。提出一种面向社交网络的虚假新闻检测方法,该方法分析社交网络中新闻文本的语言特征,并结合基于BERT全词覆盖模型抽取中文字向量来联合构建新闻文本特征。为结合这两类特征,提出了一种改进的TextCNN模型。在实验中该方法取得了更高的精度,证明了该方法的有效性。
With the development and popularization of information technologies such as the mobile Internet,rumors and fake news in social networks are spreading more rapidly with wider impacts,causing adverse social impacts and economic losses,which requires faster response speed for regulatory authorities.This paper proposes a method for detecting fake news in social networks.This method analyzes the linguistic features of news text in social networks and combines Chinese text vectors based on the BERT Whole-word Mask model to construct features of news text.In order to combine the above two features,this paper proposes an improved TextCNN model.In experiments,the method achieves higher accuracy,and verifies the effectiveness of itself.
作者
楼靓
LOU Jing(Zhejiang Institute of Communications,Hangzhou 311112,China)
出处
《浙江交通职业技术学院学报》
CAS
2020年第2期106-110,共5页
Journal of Zhejiang Institute of Communications
