Locating new sources of critical minerals begins with understanding where these minerals have been found in the past. However, historical data about mineral occurrences is often locked in disparate, unstructured, and inconsistent formats, ranging from government databases to mining reports and journal articles. To address this challenge, we have developed a set of scalable technologies that extract, normalize, and semantically integrate information from these sources into a unified knowledge graph. Our approach combines ontology-driven modeling, large-language models for information extraction and classification, and tools for linking and validating data across sources. The result is a semantically enriched, queryable knowledge graph that supports reproducible analysis, expert validation, and geoscientific applications such as deposit classification and prospectivity modeling. Through this work, we have successfully integrated information from hundreds of thousands of records across multiple historical sources to build one of the world’s largest repositories of structured data on critical minerals.