Applications of Natural Language Processing to Geoscience Text Data and Prospectivity Modeling

• DOI: 10.1007/s11053-023-10216-1
• Publication Date: 2023-06-02T03:31:57Z
• Authors: Christopher J. M....; Michael G. Gadd; Mohammad Parsa; Graham W. Lederer; Garth E. Graham; Arianne Ford
• Keywords: Prospectivity mapping; Mineral exploration

Abstract Geological maps are powerful models for visualizing the complex distribution of rock types through space and time. However, descriptive information that forms basis a preferred map interpretation is typically stored in geological databases as unstructured text data difficult to use practice. Herein we apply natural language processing (NLP) geoscientific from Canada, U.S., Australia address knowledge gap. First, descriptions, ages, lithostratigraphic lithodemic information, other long-form translated numerical vectors, i.e., word embedding, using geoscience model. Network analysis associations, nearest neighbors, principal component then used extract meaningful semantic relationships between types. We further demonstrate simple Naive Bayes classifiers area under receiver operating characteristics plots (AUC) how vectors can be to: (1) predict locations “pegmatitic” (AUC = 0.962) “alkalic” 0.938) rocks; (2) mineral potential Mississippi-Valley-type 0.868) clastic-dominated 0.809) Zn-Pb deposits; (3) search analogues giant Mount Isa deposit cosine similarities vectors. This form promising NLP approach assessing with limited training data. Overall, results highlight new reduce exploration critical raw materials.