We present MinMod, one of the largest knowledge graphs of historical mining data. MinMod is built using scalable machine learning methods to extract hundreds of thousands of records from mining reports, databases, and tabular data in articles, and to normalize and integrate the results into a unified knowledge graph. It also provides tools that enable end-users to explore, curate, and leverage the data to support predictions of new sources of critical minerals. In this demo, we walk through the process of extracting and integrating data into MinMod, demonstrate data exploration and curation, and showcase tools such as the Grade & Tonnage model that assist scientists in mineral assessments.
Given the rising global demand for critical minerals in key industries such as microelectronics and electric cars, it is crucial to the assessment process to provide actionable information to decision-makers in a timely and reproducible manner. However, the well-established methodology used by the U.S. Geological Survey (USGS) for mineral resource assessment is too time-consuming because it relies on a slow manual process of data gathering, preparation, and spatial analysis.
To address this problem, we created MinMod, a knowledge graph of mineral data extracted from
mining reports, databases, and tables from articles. MinMod also infers the deposit types of the mineral
sites and consolidates duplicated mineral sites to produce de-duplicated normalized data for mineral
prospectivity mapping [
In this demo, we present MinMod, demonstrating how users can quickly add new tables of mineral
data, explore, and edit mineral site data for a given commodity, and produce grade and tonnage models.
This demo covers the work described in the ISWC 2025 In-Use Paper on MinMod [
MinMod is a comprehensive system comprising multiple components for extracting data from mining reports, tables in articles, and databases, as well as predicting deposit types and linking duplicate mineral sites. Since one of the primary user groups of MinMod is geologists conducting critical mineral assessments, MinMod also includes a web-based user interface that allows users to interactively explore and curate the extracted data, ensuring the highest possible data quality.
In the first part of the demo, we will present an overview of MinMod using the MinMod dashboard (Figure 1) and the interactive map view of mineral sites (Figure 2). When users visit https://minmod. isi.edu, they are greeted with a dashboard summarizing the number of extracted mineral sites, their inventories (e.g., estimated ore amounts, grades, or historical production), and the number of data sources integrated into MinMod. The dashboard also provides a breakdown by commodity, allowing users to hover over a commodity to quickly see the number of associated mineral sites.
Users can explore mineral sites using the interactive map by clicking the Map View tab on the navigation bar. Figure 2 shows the map of copper mines currently available in MinMod across the globe. Users can zoom in and out, toggle the satellite layer to examine surrounding areas, and click on a site to open a webpage containing all data about that site extracted from multiple sources by MinMod.
In this part of the demo, we demonstrate how users can import mineral data stored in tables from
articles into MinMod. For this task, we use SAND [
In particular, we showcase a typical workflow for mapping a table of zinc deposits, which includes deposit names, locations, deposit types, and their grades and tonnages. Figure 3 shows the table after it has been imported into SAND. By clicking the Predict button, the user can automatically generate a semantic description of the table, which annotates the column types and relationships between them. The user can then manually correct any errors in the description. Figure 4 illustrates the corrected semantic description after user edits. In this demo, the only required update is changing the property of Column 10 from tonnage to grade. Finally, the user can extract the data by clicking the More button and selecting the Export function to publish data into MinMod.
One of the essential tools used by USGS scientists for critical mineral assessments is plotting the grade, tonnage, and deposit types of mineral sites in a chart. This plot, commonly referred to as the Grade & Tonnage Model (GTM), can be generated by clicking the Grade-Tonnage Model tab on the navigation bar and selecting a commodity to visualize.
In this part of the demo, we show how to generate a GTM for rare-earth elements (REEs) (Figure 5). Users can select “Rare Earth Elements” from the Select Commodity dropdown menu and click Generate. Since REE is a group of 17 metallic elements, the interface automatically expands it into individual elements and plots all of them (except Promethium, which has a very short half-life and is extremely rare in nature) in a single chart. In the plot, the x-axis represents tonnage (in millions of metric tons), while the y-axis represents grade (in percentage). The right-hand panel displays a list of deposit types in which the selected elements have been found or associated, sorted in descending order by the number of mineral sites. Users can click on any of the deposit types to filter and focus on the types of interest.
The diagonal lines in the chart represent sites with the same order of magnitude of contained metal, calculated as grade multiplied by tonnage. These lines help USGS scientists quickly identify which deposit types are likely to be the most productive when searching for a specific commodity or set of commodities in the case of REEs.
To ensure the highest quality data in MinMod for training machine learning models for mineral
prospectivity mapping [
Users can also refine the grouping of same sites by selecting the checkboxes at the beginning of rows, allowing them to merge entries referring to the same site or split those that are incorrectly grouped. Finally, the data can be downloaded as a CSV file by clicking the Download button at the top right of the page.
In this paper, we demonstrate the MinMod knowledge graph of mineral resources, along with the related tools such as Grade and Tonnage models that enable scientists to explore and analyze data in MinMod for mineral assessments. We also show how tables of mineral data can be added into MinMod using SAND.
MinMod has been deployed internally at the U.S. Geological Survey (USGS), where feedback has been overwhelmingly positive. It has significantly reduced the time required to manually extract and collect data from diverse sources such as NI 43-101 reports, scientific articles, and databases.
In this demo, we did not showcase other tools involved in building MinMod, such as text data
extraction from reports, deposit type prediction, or linking duplicated mineral sites. Readers are
encouraged to refer to the full paper [
This material is based on work supported by the Defense Advanced Research Projects Agency (DARPA) under Agreement No. HR00112390132, and Contract No. 140D0423C0093. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the Defense Advanced Research Projects Agency (DARPA), or its Contracting Agent, the U.S. Department of the Interior.
The authors thank Graham Lederer, Garth Graham and Jane Hammarstrom of the US Geological Survey and Simon Jowitt of the Nevada Bureau of Mines and Geology for their advice and feedback on the technology and tools.
Declaration on Generative AI The content of this paper was written by the authors. Generative AI tools were used solely for grammar checking, sentence polishing, and word choice refinement. The authors carefully reviewed and verified all AI-assisted outputs to ensure that the intent and originality of the work were fully preserved.