社交媒体的普及和信息传播的便捷化使得虚假新闻的传播和影响力不断增加,给社会带来了严重的负面影响.为了应对虚假新闻的干扰,提出了一种基于伪孪生网络的虚假新闻检测方法(Fake news detection method based on Pseudo-Siamese network,FNPS).受计算机视觉领域任务的启发,将虚假新闻的检测视为多模态语义匹配问题,采用双向长短时记忆(Bi-directional long short-term memory,BiLSTM)网络和50层残差网络(50-layer residual nets,ResNet50)分别提取新闻数据的文本特征和图像特征,并将它们从原始空间映射至新的目标空间来衡量文本与图像的语义匹配程度.通过测试微博数据集,FNPS模型可以有效检测跨领域虚假新闻,并优于其他的多模态虚假新闻检测模型.
FNPS:fake news detection method based on Pseudo-Siamese network
The popularity of social media and the convenience of information dissemination have led to a continuous increase in the spread and influence of fake news,which has brought serious negative impacts to society.In order to tackle the interference of fake news,the Fake news detection method based on Pseudo-Siamese networks(FNPS)is proposed.Inspired by tasks in the computer vision field,the detection of fake news is regarded as a multi-modal semantic matching problem.The FNPS adopts the feature extractor,and uses Bi-directional long short-term memory(BiLSTM)and 50-layer residual nets(ResNet50)networks to extract text features and image features from news data respectively.This mothod maps the multi-modal features from the original space to a new target space,and further measures the semantic matching degree between text and images in the target space.By testing on the WeiBo dataset,FNPS model detects cross-domain fake news effectively and outperforms other multi-modal fake news detection models.
NETL
NSTL
万方数据
