QueryPlot: Generating Geological Evidence Layers using Natural Language Queries for Mineral Exploration

A research team led by Meng Ye has introduced QueryPlot, a novel AI framework that revolutionizes mineral exploration by allowing geologists to search for deposits using natural language queries instead of manual, knowledge-intensive processes. The system integrates large-scale geological text corpora—including descriptive models for over 120 deposit types—with structured representations of the State Geologic Map Compilation (SGMC) polygons. Using pretrained embedding models, QueryPlot encodes both user queries and region descriptions, computing semantic similarity scores to rank and visualize prospective regions as continuous evidence layers.

Technically, QueryPlot supports compositional querying over multiple deposit characteristics, enabling aggregation of similarity-derived layers for multi-criteria analysis. In a case study on tungsten skarn deposits, the embedding-based retrieval demonstrated high recall of known occurrences and produced prospective regions that closely aligned with expert-defined permissive tracts. The similarity scores can also be incorporated as features in supervised learning pipelines, yielding measurable improvements in classification performance—a significant advancement for predictive modeling in geology.

The framework is implemented as a web-based system supporting interactive querying, visualization, and export of GIS-compatible prospectivity maps. All source code and datasets have been made publicly available to support further research. This represents a major step toward democratizing geological expertise and accelerating mineral discovery through accessible AI tools.