Abstract:

Knowledge graph construction is an effective means of acquiring and representing knowledge in data-driven research, however, existing knowledge graphs have many problems and limitations in mineral resource prediction. Firstly, relevant studies are few while existing knowledge graphs lack spatiotemporal semantics, which limits the effective representation and analysis of the spatiotemporal characteristics of mineral resources. Secondly, existing graph construction methods emphasize text extraction at the data level, but lack ontology construction involving complex logical relationships and lack effective association between ontology and data layers. As a result, existing knowledge graphs lack in-depth and sufficient semantic information to meet the requirement of mineral resource prediction in expressing complex geoscience concepts and relationships. To address this issue, this study takes an ontology-guided approach to construct a knowledge graph suitable for mineral prediction tasks. We first construct the initial domain ontology on the basis of in-depth understanding of mineral prediction theories and methods; we then integrate the domain ontology with selected mature geological time ontology and geographical space ontology to expand the initial ontology—by embedding spatiotemporal semantics we can effectively express the spatiotemporal characteristics of mineral resources. We also pay attention to the association between ontology and data layers—by establishing rich semantic relationships we can achieve effective inter-node connection and information sharing in the knowledge graph. Experimental results show that the knowledge graph outperformed other existing graphs in terms of knowledge richness and confidence. This study provides a methodology for multi-ontology based knowledge graph construction for mineral prediction, thereby promoting further development of this field.

Key words: mineral resources, knowledge graph, ontology engineering, mineral prediction theory