| With the continuous progress of science and technology,a large number of papers in various fields will be published every year,and how to select papers that meet the interests of scientific research users from the massive paper data to reduce the search cost of users is an urgent problem to be solved.Under the wave of artificial intelligence technology,a recommendation system for academic papers has gradually become an effective way to solve the above problems.Most of the existing paper recommendation systems are based on citation networks,co-authorship relationships or information content,but for academic papers with multiple types of characteristic information,the recommendation results in this way alone are often unsatisfactory.Based on the complex feature information of academic papers,this paper will construct the knowledge graph of academic papers and introduce it into the recommendation system.However,due to the complexity of the data structure of the knowledge graph,the general recommendation system,whether it is based on embedding methods or source paths,can effectively use the information in the knowledge graph.To this end,this paper proposes a paper recommendation model based on academic paper knowledge graph and graph neural network-PKGAT,the main work can be summarized into the following two parts:1.This paper constructs the knowledge graph of academic papers,and designs the paper recommendation algorithm based on knowledge graph and graph neural network.In order to effectively use user interaction information,a new graph model is proposed,which combines user interaction information and paper knowledge graph into a collaborative knowledge graph,and then improves the graph neural network structure,introduces a two-layer attention mechanism,and experimental results show that the new network can effectively use the user-paper interaction information and the information in the academic paper knowledge graph,and can achieve good recommendation effect in the actual paper recommendation scenario.2.According to the actual application scenario requirements,this paper designs and implements the relevant functions of the crowdintelligence open innovation platform.The Landinn Platform applies the PKGAT model to the Web,which can realize the recommendation and retrieval of academic papers,and combine the recommendation results to empower various innovative project competitions.The results of the operation of this website verify the correctness and effectiveness of the research and design work of this paper. |
