Modeling Sparse and Heterogeneous Geochemistry Data
Modeling Sparse and Heterogeneous Geochemistry Data to Accelerate Critical Mineral Discovery (STTR Phase-II)
This project develops AI-driven methods to automatically extract and interpret complex geochemistry data from scientific literature, enabling the creation of a global database of trace and byproduct mineral concentrations. The system integrates semantic table understanding, knowledge graph construction, and human-in-the-loop curation to support mineral prospectivity mapping and critical mineral assessments.
Funding Source(s)
The project is funded under the DARPA STTR (Small Business Technology Transfer) program, with collaboration from:
- USC Information Sciences Institute (ISI)
- InferLink Corporation
- Jataware Corporation
It builds upon prior work from the DARPA CriticalMAAS program and is being transitioned to operational use by the U.S. Geological Survey (USGS).