Fake News Detection Based On News-comment Correlation Analysis

With the penetration of social networks in today's society,social networks are free from geographical constraints to make the dissemination of information easier,and more and more users use social networks to communicate emotionally and emotionally with each other.The development of social networks has led to the use of them for the dissemination of false news,which uses information asymmetry and the untimeliness of information disclosure to create anxiety and create social panic,seriously affecting the normal productive life of the masses and public social order.Timely detection of fake news to limit its spread and rapid disclosure of true information is the most effective way to control fake news,especially since the identification of fake news is the most important part of the process.In order to make different types of news content well support the task of news authenticity,a new false news identification method is proposed,mainly based on the social network containing comment information as the research background,the aim is to combine the rich user behavior information and related deep learning algorithm in social network,from the social network structure and the text information of posting comments,to improve the inadequacy of the existing false news identification algorithm,in order to obtain a more accurate false news identification method.The specific work consists of four parts:(1)using word embedding models for news content and commentary content to convert information into computable vectors.(2)Design a news text information processing module to encode news content by contextualizing it.(3)Use recursive neural networks to adequately combine comment text and contextual information to obtain representational representations.(4)Correlation analysis between the news sentences and comments,judging the authenticity of the news by the sentences with greater correlation between the text and the comment.Experiments on the real social network dataset FakeNewsNet show that,compared with the existing fake news identification methods,the fake news identification method based on news-comment correlation analysis has better performance and can obtain better prediction results for both political and entertainment type news.