Evidence-based fake news detection is a challenging task that requires retrieving multiple pieces of evidence from the Internet to verify the authenticity of the news.Although the current methods achieve fairly good performances,there are still some problems.For example,they fail to consider the negative impacts of irrelevant evidence retrieved from the Internet on the model,and the processing of noise information in evidence text needs improving.Moreover,they ignore the impact of emotional polarity of news on the authenticity of news.To address these problems,this paper proposes an emotion-aware enhanced multi-granularity filter fake news detection,called EMGFND.First,the text information of news and evidence is aggregated through the graph structure modeling.Fine evidence information is then obtained through multi-granularity filtering.Finally,the news and evidence are interacted through the news emotion-aware attention mechanism.Several experiments are conducted on two public datasets(Snopes and PolitiFact).Our experimental results show the proposed model performs better than the baseline model.
国家科技期刊平台
NSTL
万方数据
