302 12 Building Bioinformatics Ontologies
how the measurement was performed (e.g., orally, rectally, etc.). But one
might not stop there. Body temperature normally fluctuates with a circadian
rhythm, so the time of day should also be considered. One could continue
this elaboration forever.
As the example suggests, there is no limit to the degree of detail for any
concept. Aside from the additional development cost and effort that results
from scope creep, larger ontologies are harder to understand and to use. In
addition, overelaboration can result in overlapping scope with other ontolo-
gies. This is not a problem in itself, but it can become a problem when the
designs of the overlapping concepts are significantly different and it is nec-
essary to make use of both ontologies.
All ontological commitments should be documented with a rationale for
why the commitment was made. Documenting such commitments is much
harder than it seems. The problem is that one may not be aware of the as-
sumptions that are being made. Realizing that one is making implicit as-
sumptions can be a difficult exercise. The best way to discover such assump-
tions is to have a well-stated purpose and scope for the ontology. Ontological
commitments most commonly occur at the “boundaries” of the project scope.
It is best to keep the ontology as simple as possible and to elaborate all con-
cepts only as required. Staying within the scope not only limits the amount
of work required, it also furnishes a good rationale for ontological commit-
ments.12.6.4 Strict Taxonomies
In many taxonomies, such as the taxonomy of living beings, there is an im-
plicit assumption that subclassifications are nonoverlapping. The mathemat-
ical term for this situation is that the subclasses aredisjoint. In the medical
chart ontology, the four subclasses of the Event class are disjoint. For ex-
ample, a test cannot also be an admission event. Specifying that these two
classes are disjoint will help detect errors that would otherwise be missed.
Another way to look at this distinction is whether an instance can belong
to two different subclasses of another class. When this happens it may be
an indication that the taxonomy is inaccurate. For example, in the taxonomy
of living beings originated by Linnaeus there were just two kingdoms: Plant
and Animal. When microscopic living beings were discovered, it soon be-
came apparent that there were living beings that could be regarded as being
both plants and animals. Rather than allow the kingdoms to overlap, a new
kingdom Protista, was added to deal with these new kinds of living being.