数据-知识协同驱动的共伴生矿产知识图谱构建方法

The growing disconnect between geological big data and metallogenic knowledge poses a significant challenge to modeling co-associated mineral relationships, underscoring the urgent need for a knowledge-based methodology capable of supporting intelligent analysis. To address this, we propose a data-knowledge synergy-driven approach for constructing knowledge graphs, which integrates domain ontology with the BERT-BiLSTM-CRF model. By leveraging a “knowledge-guided, data-informed” mechanism, the method enables dynamic collaboration between ontology evolution and information extraction, systematically identifying ore deposit features and co-associated relationships from multi-source geological texts and establishing semantic mappings between exploration data and metallogenic knowledge. Experimental results show that entity recognition achieves an F1 score of 83.2%, representing a 15.4 percentage-point improvement over the baseline; entity redundancy is reduced by 5.7 percentage points, markedly enhancing graph consistency. The resulting structured knowledge graph, which comprises 12,000 nodes and 28,000 relations, has been deployed in visualization analysis, intelligent question answering, mineralization prediction, and data platform services. This work realizes deep integration of data and knowledge, offering an interpretable and actionable technical pathway for transforming mineral exploration from an experience-driven paradigm to one driven by data-knowledge synergy.