What does Ontology mean?

An ontology is a formal description of the terms in a subject area and the relationships between them. It defines which classes exist, what properties they have, and how they relate to each other as parent and child classes. It differs from a keyword list in that a program can evaluate the stored relationships and draw its own conclusions from them.

An ontology is built from classes, individual objects, properties, and relationships. A taxonomy arranges the classes hierarchically, for example treating a building permit application as a subclass of an application in general. This is described using the standard languages RDF and OWL, and queried through SPARQL. A reasoner derives knowledge from this that is nowhere explicitly stored: if a rule applies to all applications, it therefore also applies to building permit applications.

An ontology pays off wherever several areas use the same terms and want to bring their data together. Examples include product catalogs in retail, heritage and property registers, and medical classifications such as the ICD. For a single dataset with a fixed schema and one department as its user, a plain field description is enough.

The advantage over a list of keywords lies in inheritance. A new rule applies to all subclasses without every individual record having to be touched. A search for timber-frame buildings will therefore also find objects recorded in the system as post-and-beam construction, because the class hierarchy knows this relationship.

An ontology is ongoing maintenance work, not a one-off project deliverable. Without a named subject-matter owner, it becomes outdated faster than the data it is meant to describe. It also only delivers value once several processes actually use the same classes and relationships.