Research On News Recommendation System Based On Knowledge Graph

In recent years,providing information that customers like has become an important goal of Internet news platforms.However,the traditional recommendation method is not ideal for the feature extraction of news information,resulting in users can not accurately find the news they are interested in.In this regard,researchers propose to use knowledge graph as auxiliary information of recommendation algorithm to improve the accuracy of news recommendation,which has become a hot topic in the current recommendation research field.This thesis proposes News Recommendation based on Knowledge Graph,which uses Knowledge Graph Embedding Based on Neighborhood Information Augmentation and Attention-Title Body Event to extract news information,so as to improve the accuracy of news recommendation.The main research contents are as follows.(1)This thesis proposes Knowledge Graph Embedding Based on Neighborhood Information Augmentation(NA-KGE),The model constructs the interaction between entities,not only considers the semantic relation and influence weight between entities and neighbor entities,but also extracts the structural features of the knowledge graph,effectively excavates the rich information in the knowledge graph,and improves the coding ability of the model.The experiment verifies the validity of this knowledge graph embedding method.(2)This thesis proposes News Recommendation based on Knowledge Graph(KGNR).This model combines the news feature information with the semantic information of the knowledge graph to build a recommendation model.News feature information is extracted through Attention-Title Body Event(ATBE),ATBE combines the news title,body and event information through the attention mechanism,and the effectiveness of KGNR recommended method is verified by experiments.(3)A financial news recommendation system based on KGNR is designed and implemented.According to the user needs to achieve the personal homepage module,website management module,registration and login module,news homepage module,collection scoring module,expounds the technical and functional design of each module.The system architecture adopts B/S and is divided into presentation layer,business layer,policy layer and data layer.KGNR generates recommendation results at the policy level,bringing users instant and accurate news recommendation services.