5.1 Bio-Ontologies 91
of interrelated biomedical concepts that provide metadata, relationships, and
semantic information for each concept. However, META is more than a sim-
ple concordance of terms. Its developers strive to provide a concept-oriented
organization in which synonymous terms from disparate source vocabular-
ies map to the same concepts. The UMLS contains semantic information
about terms from various sources, and each concept can be understood and
located by its relationships to other concepts. This is a result of the organiz-
ing principle of semantic “locality” (Bodenreider et al. 1998). For example,
interconcept relationships can be either inherited from the structure of the
source vocabularies or generated specifically by the META editors. Relation-
ships can be hierarchical or associative. Statistical relations between concepts
from the MeSH vocabulary are also present, derived from the co-occurrence
of MeSH indexing terms in Medline citations. Finally, each META concept is
broadly categorized by means of semantic types in the SN component of the
UMLS. META has been constructed through lexical matching techniques and
human review (Tuttle et al. 1989) to minimize inconsistencies of parent-child
relationships and to minimize redundancies of multiple META concepts.
The SN is a classification system for the concepts in the META compo-
nent. As an ontology, the UMLS is an ontology with a class hierarchy con-
taining over 1 million classes, represented by the concepts in META and the
semantic types in SN. In this class hierarchy, the semantic types form the top
of the hierarchy. The SN serves the additional function of defining part of
the property hierarchy of the ontology. However, UMLS concepts can have
many other attributes (such as International Classification of Diseases [ICD-
9] codes) that implicitly define many other properties. The semantics of the
UMLS has yet to be defined precisely, and it has not yet been completely
specified using any of the ontology languages.
The SPECIALIST lexicon includes lexical information about a selected core
group of biomedical terms, including their parts of speech, inflectional forms,
common acronyms, and abbreviations.
In addition to data, the UMLS includes tools such as MetamorphoSys for
customizing the META,lvgfor generating lexical variants of concept names,
and MetaMap for extracting UMLS concepts from text. The UMLS know-
ledge sources are updated quarterly (Bodenreider 2004). MetaMapii.nlm.
nih.gov/MTI/mmi.shtmlis one of the foundations of NLM’s Indexing
Initiative System which is being applied to both semiautomatic and fully au-
tomatic indexing of the biomedical literature at the library (Aronson 2001). It
has been used for mapping text to the UMLS META. For example, MetaMap
can be applied to free texts like the title and abstract fields of Medline cita-