Research On Fake News Detection Based On Neural Network

With the development of the Internet,people are now increasingly used to access information through a variety of social platforms,and at the same time,a large amount of user-generated data containing text,images and other information is generated.As China's largest social platform,thousands of users on Sina Weibo have formed a new social network,which makes a huge amount of information generated and disseminated every day.With the development of social networks,users can obtain the latest information conveniently.But it has also been accompanied by a large number of fake news.The proliferation of fake news on social media continues to worsen the network environment,bringing great negative effects to individuals and even society.The main task of fake news detection is to analyze the content of a piece of news to determine whether it is real news or fake news.In the existing research,most of the work is based on news text,and there are few researches on multi-modal fake news detection combined with picture information.However,in real life,the dissemination of news on social platforms such as Weibo often includes pictures,because pictures can convey information more intuitively.Therefore,this article aims to use news text and image features to build a multi-modal neural network model for false news detection research.To solve this problem,this paper adds the matching degree between the extracted image text and the news text as an input feature to the proposed multi-modal neural network model that combines news text and image information for the first time.First,the news is analyzed from multiple angles,and five features that can be used for fake news detection are excavated.Among them,optical character recognition is used to extract the text in the picture and compare it with the news text to get the matching degree between picture text and news text.Then these features are input into the model together with news text and pictures,and finally the neural network model is used for classification,and then the detection task is completed.This paper uses the Weibo news data set to conduct experiments,and obtains better detection results than the benchmark method,which verifies the effectiveness of the proposed model.And this model is not only suitable for the detection of false news with pictures,but also for the detection of plain text news,which has good universality.