| The proliferation of fake news can cause great impact on society,and fake news detection is the most effective means to curb the spread of fake news.To address the problems that existing fake news detection algorithms lack emotional information and have few types of features,this paper introduces emotional language information and emotional flow information in text features,fuses publisher features and user features in the propagation graph,researches and proposes two fake news detection algorithms,verifies the feasible effectiveness of the algorithms through experiments,and designs and develops a fake news detection system combined with practical application scenarios.The main research work and results include:1.A fake news detection algorithm based on sentiment analysis is proposed.To address the existing content-based fake news detection algorithms that do not consider the sentiment information in the text word vector,a news text word vector containing sentiment linguistic knowledge based on the Senti LARE model is used to extract the news text content features and fuse them with the sentiment flow features of the text content to enrich the text content features.Experiments on the Covid-19 fake news dataset and Fake News Net dataset demonstrate that the algorithm can effectively capture the sentiment information of news text content and has better classification effect in fake news detection.2.A fake news detection algorithm based on attention mechanism and multiple features is proposed.The publisher credibility prediction task and user credibility prediction task are introduced in the fake news detection to capture publisher features and user features in the propagation graph,construct publisher release graphs as well as propagation graphs and extract features using a structural attention mechanism.Meanwhile,in order to further improve the news text content features,the sentiment analysis-based fake news detection algorithm is used to obtain text features containing sentiment information,and the three features are fused to form the final news features,which enriches the feature variety of news and improves the efficiency of fake news detection.Experiments on Twitter 15 and Twitter 16 datasets demonstrate that the algorithm can effectively improve the efficiency of false news detection and perform well in early detection.3.I designed and implemented a fake news detection system using Vue and Django technologies,including the functions of submitting news to be detected,viewing detection tasks,and viewing and exporting detection results.The system uses a computational engine to add the algorithms and comparison algorithms implemented in this paper to the engine library,allowing users to select one or more algorithms for detection according to different needs,and generating fake news detection reports and detection result files.Tests show that the system has certain advantages in terms of efficiency,accuracy and performance of fake news detection. |
