The DBPedia Ontology

Download the Protégé ontology editor. Note: do not use WebProtégé; the ontology you will have to open is quite large, it is more efficient to manipulate it in a local instance of Protégé. Run Protégé and click on the menu item File > Open, to open the DBPedia ontology (DBO). In the “Active ontology” tab, in the “Ontology metrics” panel, you should see statistics about the DBO (~1200 classes, ~1200 object properties and ~1800 datatype properties, among others). As shown in this panel, OWL definitions are called axioms. Axioms can apply to classes or properties. In OWL, a difference is made between object properties and data properties. Object properties are properties with a URI as value; data properties are properties with a literal as value. For instance, in the following two triples, op is an object property and dp is a data property: <s> <op> <o> ; <dp> "o" .

Exploring the DBPedia Ontology

  1. Move to the “Entities” tab. The left panel shows the class hierarchy defined in DBO. What is the upmost class in the hierarchy? Is it part of DBO?
  2. In the class hierarchy, you should also see many QIDs (such as Q146). Are these classes part of DBO?
  3. Q146 is declared to be equivalent with dbo:Cat. What OWL triples are defined for dbo:Cat (i.e. triples that use a term defined under the owl: namespace)? Use content negotiation to retrieve a Turtle representation of the class. Explain in natural language what the OWL triples mean.
  4. What are all superclasses of dbo:Cat, up to the class dbo:Species?
  5. What is the difference between dbo:Cat and dbr:Cat? What triple could relate dbr:Cat to dbo:Species?
  6. Assume the following triple: dbr:Cat dbr:genus dbr:Felis . was defined in DBpedia. List all triples that can be inferred from that triple and DBO axioms and explain from which axioms these triples could be inferred. To do so, look at the definition of dbo:genus and all of its superproperties. To make sure the inference is correct, you can use one of the reasoners embedded in Protégé: move to the “Individuals by class” tab, create the individuals dbr:Cat and dbr:Felis and add an object property assertion (lower right panel). Use autocompletion to find the right property.

Expanding the DBPedia Ontology

  1. Ash Ketchum (dbr:Ash_Ketchum) is defined in DBPedia as an instance of dbo:FictionalCharacter. What subclass(es) of dbo:FictionalCharacter would be a better fit?
  2. dbr:Ash_Ketchum is the same entity as the Wikidata entity wd:Q308950, which has a finer-grained definition. What OWL triple could you add so that it would be automatically inferred from DBpedia and Wikidata statements that Ash Ketchum is an instance of dbo:AnimangaCharacter? Your triple should be of the form ?someClass owl:subClassOf ?someOtherClass.
  3. Some object or data properties do not have any domain or range, although there are obvious candidates. Consider for instance the property dbo:notableWork. According to the comment defined for this property, what class could be defined as domain of this property? In the “Description” panel (lower right panel), add the class expression as domain. Can any triple be inferred from that OWL definition?
  4. dbo:parent and dbo:child are not related in the DBPedia ontology. Among all options in the two panels on the right (“Characteristics” and “Description”), which one(s) is (are) relevant to add a relation between the two? Same question for dbo:parent and dbo:father. You can read the OWL 2 Primer document to learn about property characteristics.
  5. Give an example of object property that can be declared as irreflexive. To make sure that this new definition does not lead to contradictions on DBPedia, verify with a SPARQL query that no DBPedia resource violate this characteristic. You can use the official SPARQL endpoint of DBPedia to test your query.
  6. Read Section 5.2 of the OWL 2 Primer document (on property restrictions). OWL restrictions are useful to define qualified domain and range constraints. For instance, it is possible to specify that if some artist is of type dbo:Poet, the range of dbo:notableWork must be dbo:Poem. What kind of restriction would you have to specify to correctly formalize this qualified range constraint? Add it in the “Description” panel of Protégé.