Smart city multilingual exhibition navigation with graph neural networks and cross-lingual knowledge graphs

With the rapid advancement of computational technologies and their application in intelligent infrastructure and smart cities, exhibition navigation systems have become essential for delivering personalized and efficient guidance in multilingual and dynamic environments. As a key component of smart city initiatives, these systems aim to integrate cultural accessibility with advanced technological solutions. These systems are integral to enhancing user experiences in cultural spaces and supporting global tourism initiatives. This study proposes an innovative framework that combines Graph Neural Networks (GNNs) with a Cross-Lingual Knowledge Graph (CLKG) to optimize navigation paths and enhance multilingual adaptability. Exhibition spaces are modeled as weighted graphs, where GNNs dynamically optimize navigation by integrating spatial relationships and user preferences, while the CLKG enables seamless semantic reasoning across languages through entity embedding techniques. Extensive experiments on a multilingual dataset with over 10,000 nodes and 26,000 edges demonstrate the framework's effectiveness. The system achieves an average path accuracy of 87.4% and a multilingual adaptability score exceeding 0.92 across five major languages, significantly outperforming baseline methods. Additionally, the response time remains efficient even in large-scale graph scenarios, showcasing the scalability of the approach. This work provides a robust and language-agnostic solution for improving exhibition navigation in complex, multilingual environments.