Ontology Evaluation

For the evaluation of ontologies we have to consider qualitative, quantitative and organizational aspect. For the quantitative aspects we have defined cardinalities for the knowledge graph of an ontology. Organizational aspects like availability and recoverabilitywill have to be handled by the provider of the ontology. They are not necessarly related to the quality of an ontology. The qualitative aspects of an ontology will fall mainly into the responsibility of the developers and modelers of an ontology.

An important criterion for the evaluation is productivity. On the one hand, productivity refers to how effectively and quickly ontologies are developed. On the other hand, it refers to the benefit achieved through the application of the ontologies.

The productivity of developing an ontology essentially depends on how quickly the schema of the ontologies can be developed and maintained. In this context, the number of instances of the particulars layer is of secondary importance because it is a data acquisition problem analogous to database systems. The number of triples processed per minute is used as a measure of the productivity of schema development. Let m be the time in minutes it took to develop the schema and n be the number of triples that were modeled or modified in that time.

Exact knowledge is to be understood here as knowledge that can be understood as absolute and correct, no matter from which point of view you look at it. These certainly include natural laws and natural constants or theorems, such as the Pythagorean theorem, i.e. everything that can be proven by a truth maker and does not contradict other knowledge.

Vague knowledge can be understood as the amount of knowledge that is subject to a certain degree of uncertainty. This type of knowledge was dealt with in the field of fuzzy logic, among others. However, vague knowledge can also be the kind of knowledge to which a certainty factor can be assigned. For example, in medicine, abdominal pain and headaches in combination are both indicators of gastroenteritis and meningitis. It therefore depends on further information as to which clinical picture must be assumed to be the more probable.

Knowledge that is considered correct from one observer's perspective may be incorrect from another observer's perspective. This is how it was, for example, with the Copernican world view in relation to that of Ptolemy. The latter assumed that the sun orbits the earth, while the former was able to show that the opposite is the case. We show under Kopernikus Belief how this knowledge can be represented in such a way that it does not have to be stored as a reliable fact in the ontology. With the Thematic Roles introduced by John Sowa, arbitrarily nested state of affairs such as Tom thinks Mary wishes to marry a sailor model. We  summarize all types of knowledge that refer to modal verbs and the associated concepts such as belief, desire, ability and will under the term relative knowledge.