Detection of New Crown Epidemic Rumors Based on Knowledge Graph

Since the outbreak of the new crown epidemic, a large amount of epidemic-related information has emerged on the Internet, including many false information and rumors, which may mislead the public and pose a potential threat to public health and social security. Therefore, in the context of the new crown epidemic, how to automate the rumor detection and identification of rumor information has become an important issue that cannot be ignored. In this context, this paper integrates and improves several mainstream deep learning models to achieve rumor identification by introducing the entity representation of external knowledge, and the main work contents are: new crown rumor detection model based on knowledge graph, new crown epidemic rumor detection system, and construction of new crown rumor dataset. In this paper, a directed heterogeneous document graph is constructed for the text to be tested, subject and entity information are introduced, structured and unstructured feature extraction is performed for external information, entity features based on external knowledge are obtained, and entity comparison is carried out through the entity comparison network, and finally the rumor detection results are obtained. Based on the above new crown rumor detection model, this paper designs and implements a new crown epidemic rumor detection system to assist users in identifying the authenticity of new crown epidemic news. In addition, this paper constructs the latest COVID-19 Rumor Dataset (CRD), including 24915 rumor data and 9995 non-rumor data, including a total of 34910 data, which has the advantages of large data volume, high timeliness and rich themes.