Monday, January 25, 2016

Lecture and Lab Course Switched this Week!

Dear students,
Due to travel and project obligations, we have to switch the lecture and the lab course this week (16./27.01.2016).

  • Lecture (originally on Tue, 26.1.2016, 1.30pm) 
    • -> rescheduled to Wed, 27.1.2016, 9.15 am
  • Lab Course (originally on Wed, 27.1.2016, 9.15am) 
    • -> rescheduled to TUE, 28.1.2016, 1.30 pm
Thanks a lot for your consideration.

Wednesday, January 13, 2016

Ontology Evaluation

It is essential to ensure the quality of the ontology that you are going to develop. Simply for reasons of efficiency with respect to further refinement or additions. When working and discussing the examples with OntoClean in the lecture and the lab course, you have realized that it is important to have clear definitions of your concepts/classes. In our examples, we have treated the discussed ontologies always as rather general upper ontologies with the aim to represent the 'real' world in its entirety...and as you might have noticed, this sometimes is more or less futile since we were not able to reach an agreement upon the interpretation of the represented classes.

Therefore, in real life applications of ontology engineering, always keep in mind the purpose of your ontology, the application for which the ontology is intended to be used. Keep your knowledge representations restricted to the intended purpose. If your application demands countries to be treated simply as geographical entities, you don't have to consider the administrative or legal structure of a country for your application. Thus, it should also not be necessarily considered in your knowledge representation -- or also in your evaluation.

The OntoClean methodology focusses on the taxonomic structure of an ontology, where the suborder relation is considered being a formal subsumption. Strictly speaking, if an individual is member of a class, it is necessarily also a member of the according superclass. To check the semantical consistency of your class taxonomy, all the classes are tagged with metaproperties (Rigidity, Unity, Identity, and Dependency). It is crucial that your metaproperties clearly reflect the intended meaning of the classes. If in doubt, always give an explanation/justification, why you have chosen a specific metaproperty. Then, you look for violations of the given OntoClean constraints for these metaproperties. Whenever you find a constraint violation, this is an indication either to change your ontology or to reconsider the definition (the meaning) of your ontology classes accordingly.

One comment on the exercise in the lab course today:
We should label the owl:Thing (the top concept in the DBpedia Ontology) with -U, since the instances of owl:Thing do not have a common unity criterium. Possibly there are also some classes subsumed by owl:Thing that contain entities which are not necessarily considered a whole, i.e. which must have the meta-property ~U (anti-Unity), but this cannot be assumed for all the individuals of owl:Thing alike (and also for entity in the example from the slides, lecture 5.8). Instead, we have to choose the more general -U (no common unity criterium) for owl:Thing here. 

Thanks a lot for pointing this out to me at the end of the lab course!!!