GoodOD

GoodOD

This section is based on the "Guideline on Developing Good Ontologies in the Biomedical Domain with Description Logics, Version 1.0" [ScSe2012].

Why does the document restricted itself to usage in the Biomedical domain? In our opinion, these guidelines are applicable to the development of ontologies in any domain as well.

Pitfalls for Class Selection

[ScSe2012], page 49: Modellers have thus to avoid the following:

1. "Introducing multiple classes that in fact refer to the same type of entity in reality."
2. "Introducing only a single class that encompasses unrelated types."
3. "Introducing classes that do not have corresponding entitties in reality."

We agree only partially or not at all with these recommendations:

• Ad 1: The example for SyphilisInfection describes an application where the realization that two classes describe the same concept was only gained over time. However, this is normal in the modeling process and can be corrected later by equating the classes using the object properties <>EQ or <>same_as.
• Ad 2: The example of NileCrocodile shows the opposite case, where it was only realized over time that the species should be divided into two.
• Ad 3: In our opinion, there is no reason not to introduce classes for which there are no real entities in reality. Just as one would not be allowed to model unicorns, one would then not be allowed to model fictional characters like Tarzan or James Bond.

Catch-all Classes

Page 51: "catch-all" classes like FractureOfLegNotOtherwiseSpecified shoud be avoided. The simple reason for this is that a patient's specific leg fractures could easily be managed as an instance of the class FractureOfLeg. If necessary, one could later assign this instance to a better characterizing subclass.

Page 57: "Since ontologies are not supposed to contain individuals ..." . We strongly disagree with this position. From this one could deduce that any knowledge base containing particulars could not be an ontology in the eyes of the authors.

Page 57 footnote: "Again, it is important not to confuse the name Paul, which can, albeit ambigously, apply to multiple persons, with the person Paul, wo is a single human being" . We strongly recommend to follow our naming conventions, then the individual Paul would be identified by >Paul or even better >NPS-Paul, and the class Paul would be identified by ^Paul. Then we could have (>Paul, <>iof, ^Paul) or as a more meaningful example (>ATW-Guernica, <>iof, ^Picasso). The latter can then be interpreted as "The artwork Guernica is a Picasso".