Assessing named entity recognition by using geoscience domain schemas: the case of mineral systems

Named Entity Recognition (NER) is crucial for accurately extracting and classifying specialized domain terms from textual data. This study introduces the Schema for Mineral Systems (SMS), designed through domain characterization, word disambiguation, taxonomy development, and expert input to refine NER approaches in geosciences. SMS, featuring nine geological and five general entity classes, enhances the precision of term identification in mineral system texts. Utilizing domain-specific dictionaries and schema-linked annotations, the schema facilitates the distinct recognition of unique terms, underscored by iterative expert validation to refine NER accuracy. Applied to iron and lithium deposit corpora in Western Australia, SMS highlights the challenges and effectiveness of context-specific schemas in specialized knowledge extraction and accurate entity recognition within complex domains.