A knowledge-graph-guided interactive system with controllable generation for enhancing the interpretation of Song Dynasty ceramics

Abstract Ancient Chinese artifacts embody rich cultural knowledge and aesthetic value, yet such information is often fragmented and expressed through specialized terminology, creating barriers for public understanding. This study proposes an interactive museum experience system that integrates knowledge graphs with generative artificial intelligence to support semantic-driven interaction with Song Dynasty ceramics. A knowledge graph was constructed to structure information on ceramic types, patterns, and decorative areas, providing semantic guidance for generation and interaction process. A two-stage workflow combining Stable Diffusion with LoRA fine-tuning and the control capability of FLUX.1 Kontext was implemented to enable controllable ceramic image generation. Based on this framework, an interactive process allows users to explore ceramic features and generate images while learning contextual knowledge. Experimental results and user studies suggest that the system produces stylistically coherent ceramic images and shows potential for enhancing visitor engagement and cultural understanding.