While existing representations of the biomedical domain may be sufficient for information retrieval purposes, the organization of knowledge in these representations is generally not suitable for reasoning by computers. Reasoning by computers requires the principled, consistent organization usually provided by ontologies.

Medical Ontology Research develops methods whereby ontologies may be acquired from existing resources, as well as validated against other knowledge sources. Specific research targets semantic spaces in order to use data in coordination with other projects or applications developed at the U.S. National Library of Medicine (NLM). Towards this goal, the following aspects of semantic spaces in the biomedical domain are currently being investigated:

• Definition - Semantic spaces can be defined from the semantic information that is provided by existing terminologies, knowledge bases, expert systems, or extracted from the medical literature.
• Organization - The relationships among biomedical concepts must be consistently labeled in order to create robust hierarchical and associative structures.
• Visualization - Once defined and organized, semantic spaces can be visualized and presented to users in order to provide them with a representation of a subdomain. This representation will help users navigate to the information they seek. Issues such as granularity, redundancy, and consistency between sources must be addressed before designing applications for the visualization and navigation of semantic spaces.
• Utilization - By specifying relationships between concepts, semantic spaces provide the basic knowledge used in applications such as concept-based indexing, concept-based retrieval and terminology servers.

Using the UMLS as the primary knowledge source along with SNOMED CT, the Gene Ontology, GALEN, MEDLINE citations, medical encyclopedias, and medical corpora, the Medical Ontology Research project will verify the formal properties of concepts and relationships for consistency and accuracy.