Constructing regional mineral prospecting knowledge graph from GIS maps

This is the authors' abstract. We don't add key points for this paper.

Geographic Information System (GIS) layers contain both spatial precision and domain knowledge, making them valuable for mineral prospectivity analysis. This study proposes a task-oriented methodology to construct a mineral prospecting knowledge graph directly from GIS maps. The framework integrates ontology construction, spatiotemporal semantic embedding, and triple confidence evaluation. Ontologies are built from GIS layers through terminology extraction and alignment with existing standards, while spatial and temporal semantics are encoded using GeoSPARQL and the Geological Time Ontology. Graph Convolutional Networks (GCN) combined with the TransE embedding model are then applied to assess triple plausibility. A case study in the Eastern Tianshan region of Xinjiang verifies the effectiveness of the proposed method through semantic evaluation and graph-theoretic analysis. Guided by GIS, ontology construction significantly enhances the semantic fidelity and structural robustness of the prospecting knowledge graph, providing relatively reliable support for subsequent reasoning and predictive studies. • A task-oriented methodology is proposed for constructing mineral prospecting knowledge graphs directly from GIS layers. • The approach integrates ontology construction, spatial–temporal semantic embedding, and triple confidence evaluation. • GIS layers are reused as both semantic carriers and sources of spatial constraints, ensuring semantic consistency and spatial validity. • Experimental evaluation combines semantic analysis with graph-theoretic measures.