The DBpedia Ontology


The DBpedia Ontology is a shallow, cross-domain ontology, which has been manually created based on the most commonly used infoboxes within Wikipedia. The ontology currently covers over 205 classes which form a subsumption hierarchy and has 1,210 properties.


With the DBpedia 3.2 release, we introduced a new infobox extraction method which is based on hand-generated mappings of Wikipedia infoboxes to the DBpedia ontology. The mappings define fine-granular rules on how to parse infobox values. The mappings also address weaknesses in the Wikipedia infobox system, like having different infoboxes for the same class (currently 685 Wikipedia templates are mapped to 205 ontology classes), using different property names for the same property (currently 2,820 template properties are mapped to 1,210 ontology properties), and not having clearly defined datatypes for property values (the ontology used 55 different datatypes). Therefore, the instance data within the infobox ontology is much cleaner and better structured than the infobox data within the DBpedia infobox dataset which is generated using the old infobox extraction code.


Overview of the class hierarchy of the DBpedia Ontology.


The DBpedia Ontology currently contains about 1,173,000 instances. The table below lists the number of instances for several classes within the ontology:


Class Instances
Resource (overall) 1,173,000
Place 339,000
Person 282,000
Work 234,000
Species 130,000
Organisation 119,000
Building 30,000

The DBpedia Ontology is provided for download in four parts:

  1. DBpedia Ontology T-BOX (Schema)
  2. DBpedia Ontology RDF type statments (Instance Data)
  3. DBpedia Ontology other A-Box properties (Instance Data, strict extraction)
  4. DBpedia Ontology other A-Box properties (Instance Data, loose extraction)

The DBpedia Ontology can also be queried via the DBpedia SPARQL endpoint and can be explored via the DBpedia Linked Data interface. Examples: Class Place, property elevation.


Until now, we have only written mappings for a subset of all Wikipedia infoboxes. Therefore, the DBpedia Ontology currently does not cover the whole range of infoboxes and infobox properties within Wikipedia. For the next DBpedia release, we plan to provide a public, wiki-style interface for writing infobox mappings. This will allow external contributors to define mappings for the infoboxes they are interested in and to extend the existing DBpedia ontology with additional classes and properties.


 
Files [Hide files/form]
There is no comment on this page. [Display comments/form]

Information

Last Modification: 2009-11-11 10:27:24 by Chris Bizer