SSAR-GNN: Self-Supervised Artist Recommendation from spatio-temporal perspectives in art history with Graph Neural Networks

Artists and the works of artists are an important part of art history. Existing studies on artists mainly focus on the formation and transformation factors of artistic styles of individual artists, especially well-known individuals, and there is little research on the similarity relationship between artist groups and artists. This paper is the first to study the problem of artist recommendation based on artist relationship. This study can provide more accurate recommendations for artist similarity analysis, avoid invalid searches, and provide a comprehensive understanding of the relationship among artists. In this paper, we propose a dataset of artists to analyze the similarity relationship among artists in art history in terms of space and time. Specifically, based on this dataset, we propose a self-supervised learning method to build historical knowledge graphs of artists. To integrate the learned knowledge graph into existing models, we propose a new recommender network named Self-Supervised Artist Recommendation with Graph Neural Networks (SSAR-GNN). SSAR-GNN applies a simplified graph convolutional network (GCN) on the historical knowledge graphs to enrich the representation of each artist. Experimental results on this dataset demonstrate the effectiveness of our proposed method in terms of accuracy.(c) 2023 Elsevier B.V. All rights reserved.