GNN-based retrieval and recommadation system: A semantic enhenced graph model

As a general method that captures knowledge from relation, graph modeling has been widely used in many fields including recommendation in Internet applications. A typical graph based system must be provided with a nodes and edges. To get information from both nodes and edges efficiently, many re-search use Graph Neural Network (GNN) to process their data. We propose a retrieving and recommendation system based on GNN to be spe-cific for learning the graph data with disordered topology structure, we lead a semantic clustering phase into GNN classification phase which promotes the search accuracy. At training phase, we give a try to go to a deeper layer. In the end, quantitative and qualitative experiments demonstrate the distinctly-improved effectiveness of the proposed approach towards the application of recommendation system.