Using Co-Occurrence Data to Determine a Thesaurus Structure.

Background. In 1992, researchers at the University of Missouri-Columbia began to develop the Thesaurus of Health Informatics. Candidate terms were based on literary warrant and were culled and classified by traditional techniques. Six classes of terms emerged. The researchers identified hierarchical relationships within these classes. Several methods were used to validate the terms and, to some extent, the hierarchical relationships. The researchers made no exhaustive attempt to identify relationships across classes. The resulting thesaurus contained 556 distinct terms.