Abstract
Objectives: Medical decision support and other intelligent
applications in the life sciences depend on increasing amounts
of digital information. Knowledge bases as well as formal
ontologies are being used to organize biomedical knowledge and
data. However, these two kinds of artefacts are not always clearly
distinguished. Whereas the popular RDF(S) standard provides
an intuitive triple-based representation, it is semantically weak.
Description logics based ontology languages like OWL-DL carry a
clear-cut semantics, but they are computationally expensive, and
they are often misinterpreted to encode all kinds of statements,
including those which are not ontological.
Method: We distinguish four kinds of statements needed to
comprehensively represent domain knowledge: universal statements,
terminological statements, statements about particulars
and contingent statements. We argue that the task of formal
ontologies is solely to represent universal statements, while the
non-ontological kinds of statements can nevertheless be connected
with ontological representations. To illustrate these four types
of representations, we use a running example from parasitology.
Results: We finally formulate recommendations for semantically
adequate ontologies that can efficiently be used as a stable
framework for more context-dependent biomedical knowledge
representation and reasoning applications like clinical decision
support systems.