Comparing anatomical data across species is unreliable due to
the use of species-specific anatomical ontologies. Existing
approaches include the generation of inter-ontology mappings
using text matching, but these can be error prone and difficult
We have developed an uber-anatomy ontology called Uberon which
consists of classes generic enough to subsume species-specific
classes in existing ontologies. Uberon was seeded using
standard text matching methods, but was iteratively refined
through a combination of manual curation and automated
ontology population and computational reasoning. The resulting
ontology includes groupings that would not have been found
using text matching methods alone, and excludes common
mistakes found in mappings.
Text-based mappings are unreliable, and manual curation can be
time consuming. However if NLP methods are combined with
curation, then we can have the benefits of rapid ontology
construction with expert oversight. Using logic-based
automation techniques the results can be enhanced further.
Uberon has been applied in different domains, including
phenotype comparison and the Gene Ontology.