Research And Implementation Of False Information Detection Tech-Nology Based On Knowledge Graph

The rapid development of the Internet has greatly increased the convenience of posting information and accessing information on social media.However,the quality and credibility of information have decreased,making social media a hotbed for the generation and dissemination of false information,which has a negative impact on social stability.Traditional false information detection methods use the model of manual reporting and manual verification,but this model cannot achieve the purpose of real-time and effective detection,so it is crucial to study the automated false information detection methods based on artificial intelligence and social media analysis technology.The existing deep learning-based false information detection methods have made some progress,but false information is multi-domain,multi-modal,and dynamically evolving,presenting a high degree of confusion and deception,and it is difficult to effectively detect the information by relying only on its own content,and other external auxiliary information usually needs to be considered.For example,knowledge graphs can be introduced as clues or evidence to discover richer modal associations in information,which can be used to enhance the perception and representation capabilities of models.However,the complex association relationships among information-related text,images,communication patterns and other multimodal data impose high requirements on the association discovery ability of false information detection models,especially after the introduction of knowledge graph information,capturing the association relationships between knowledge and each modal information is a major challenge for model design.In addition,information-related multimodal data and knowledge graph data are diverse in terms of modality and structure,which are difficult to be directly correlated and fused.Therefore,the main challenge of this study can be summarized as how to construct a unified representation of multimodal complex structural data and model the association between heterogeneous data to obtain knowledge enhanced information feature representation.To address the above challenges,this paper proposes two knowledge graph-based false information detection methods to fully explore the associations and clues among multimodal data,thus improving the performance of false information detection algorithms,and develops an integrated prototype validation system.The main work and contributions of the paper are as follows.Firstly,a knowledge graph-based multimodal anti-common knowledge false information detection method is proposed for the correlation relationship between text information and image information on social media.The method fully considers the association between text-image modal data and between knowledge and modal information,defines two kinds of inconsistencies of false information:cross-modal inconsistency and content knowledge inconsistency,and designs a dual inconsistency neural network based on knowledge graph to capture the two inconsistency features and perform information authenticity discrimination.Experimental results on two real datasets demonstrate the effectiveness of the model.Secondly,in order to capture and model the complex semantic structure relationships of information posts,user comments and the propagation patterns among them on social media,a false information detection method based on knowledge graphs and dynamic propagation structures is proposed.The method integrates the spatial structure,temporal structure and background knowledge of information propagation,fully considers the association between text,each modal data of propagation structure and between knowledge and knowledge,knowledge and each modal information,and innovatively designs a dual dynamic graph structure:dynamic propagation graph and dynamic knowledge graph according to the temporal and spatial structure of information propagation to capture key information and comment node feature representation,so as to perform false information detection.The model proves its effectiveness by conducting experiments on two real data.Finally,integrating the above two detection methods,a false information detection system based on the knowledge graph is designed and developed.The system has a user-friendly visual operation interface,which can interactively complete operations such as processing data,training models,and displaying predictions,and provides model application interfaces.In addition,the user can adjust the model parameters according to the task requirements.