Construction and Application of a Knowledge Graph for Iron Deposits Using Text Mining Analytics and a Deep Learning Algorithm

Reports on mineral exploration provide several insights regarding the geological settings of mineral deposits. However, because the reports are presented as unstructured text, it can be difficult for geologists to extract meaningful geological information without manually reading through a sizable number of reports. In this work, geological data from such underutilized exploration reports relevant to mineralization and ore-forming conditions were automatically retrieved. It is shown that a knowledge graph (KG) of ore-forming circumstances can harmonize heterogeneous data to enable effective and efficient data-driven discovery. This is accomplished by creating KGs that define geological entities and their relationships in exploration reports. Based on the sequence features extracted by the bidirectional long short-term memory network, the syntactic structural information from the graph convolutional neural network coding dependency analysis results are used to construct an end-to-end entity relationship joint extraction model by using the improved entity annotation strategy. In this research, six dominant entities and 24 relation types are considered. The generation of high-quality KGs from geological reports of iron ore deposits illustrates the effectiveness of our methodology. The results indicate that the structured information contained in the KGs accurately represents the contents of the source reports and corresponds to domain knowledge. The suggested approaches are capable of quickly and reliably converting text data into structured form, and indicate that KG procedures can help in the knowledge discovery of the metallogenesis and spatiotemporal evolution in mineral exploration. Our study tackles the scarcity of machine-readable KGs for ore-forming conditions and will aid in the integration of geological data from diverse sources in data-intensive research.