The field of recommender systems is witnessing a significant shift towards leveraging graph structures to improve recommendation accuracy. Researchers are increasingly focusing on harnessing the power of graph neural networks (GNNs) to incorporate semantic information from knowledge graphs and property graphs into recommendation models. This has led to the development of novel methods that can capture complex relationships between entities and improve the robustness of recommendation systems, particularly in sparse data scenarios. Notable advancements include the integration of rule-based approaches, context-adaptive attention mechanisms, and simplified GNN architectures. These innovations have demonstrated substantial improvements in recommendation performance and have the potential to enable more effective and efficient recommender systems. Noteworthy papers include: Rule-Assisted Attribute Embedding, which proposes a novel method for attribute embedding in property graphs. Bridging RDF Knowledge Graphs with Graph Neural Networks, which presents a comprehensive integration of RDF knowledge graphs with GNNs. Context-Adaptive Graph Neural Networks, which introduces a context-adaptive attention mechanism for next POI recommendation. LightKG, which proposes a simplified GNN-based knowledge-aware recommendation approach that addresses sparsity issues.