Breaking the Semantic Silo: LLM-GIS Fusion for Context-aware Cultural Heritage Documentation

Cultural heritage documents contain rich historical, social, and spatial information, yet their unstructured nature presents significant challenges for effective integration with Geographic Information Systems . This paper proposes an innovative framework that leverages the deep semantic understanding and contextual reasoning capabilities of Large Language Models to achieve intelligent parsing of cultural heritage documents, context-aware information extraction, and precise fusion with GIS spatial data. By constructing a spatiotemporal knowledge graph, designing context-aware information extraction strategies based on prompt engineering, and developing a dynamic LLM-GIS interaction interface, this framework significantly enhances the depth and precision of spatial representation for cultural heritage. Experimental results demonstrate the system's superior performance in historical toponym disambiguation, spatial relationship reconstruction, and multi-dimensional cultural landscape visualization, providing robust spatiotemporal intelligence capabilities to support cultural heritage research, preservation, and public dissemination.