Abstract: For a long time, geological workers in our country have conducted extensive investigations and research on the three types of rare mineral, accumulating a vast amount of geological exploration data. These data contain rich knowledge about the geological characteristics, genesis, geological environment, and tectonic background of mineral deposits. However, the central management of these data and its formalized expression in a way that supports computational reasoning in geosciences has become a hot topic in the field of geoscience artificial intelligence. This paper introduces knowledge extraction and knowledge representation techniques from knowledge engineering to build a knowledge graph based on unstructured data related to the three types of rare mineral. Firstly, the concept, entities, and relationships of the three types of rare mineral are clarified, and a knowledge system of the three types of rare mineral is established, forming the ontology of the three types of rare mineral. Secondly, knowledge is extracted using deep learning based on the established ontology of the three types of rare mineral, and a knowledge graph of the three types of rare mineral is constructed. Lastly, Neo4j graph database is used to store triplets and visualize the relationships between entities and the attribute sets of the mineral deposit ontology. The research results of this paper can provide functions such as scientific research, management, knowledge acquisition, analysis, and visualization of the three types of rare mineral, serving as an application example for the construction of a knowledge graph in the field of the three types of rare mineral. It also provides support for subsequent applications in geoscience knowledge reasoning, identification and management of mineral resources, and ore deposit prediction based on knowledge graph.