DBpedia domains and URIs

From Mediawiki1
Jump to navigationJump to search

Long-term goals

  • All languages use IRIs, not URIs
Items IRI pattern
Ontology classes and properties http://dbpedia.org/ontology/Xyz
Datatypes http://dbpedia.org/datatype/xyz
Resource IRI for http://xx.wikipedia.org/wiki/Xyz http://xx.dbpedia.org/resource/Xyz
Properties extracted by generic template extractor from http://xx.wikipedia.org/ pages http://xx.dbpedia.org/property/xyz

Deviations from these rules

In the past, we did not follow these rules. For backwards compatibility, we will allow some deviations for a while. Might be a long while. :-)

  • English and French use URIs, not IRIs. (DBpedia Berlin will publish additional datasets with IRIs, but the main datasets will use URIs.)
Items deviating IRI pattern
Resource URI for http://en.wikipedia.org/wiki/Xyz http://dbpedia.org/resource/Xyz
Properties extracted by generic template extractor from http://en.wikipedia.org/ pages http://dbpedia.org/property/xyz

Notes

The main DBpedia release and a DBpedia chapter must use the same syntax for equivalent IRIs.

We used to map page titles from non-English Wikipedias to IRIs using the inter-language link to the English Wikipedia. If there was no such inter-language link, we did not extract any data from the non-English page. That was because we used URIs like http://dbpedia.org/resource/Xyz for all languages and had to 'normalize' the URIs. We will not do that anymore - we will use http://xx.dbpedia.org/resource/Xyz IRIs.

Implementation details

During the extraction, the framework will use http://xx.dbpedia.org/ IRIs for all languages, even English. Different serializers will serialize them differently, according to some rules, for example:

  • convert en.dbpedia.org to dbpedia.org
  • convert IRIs to URIs for some languages (in subjects, predicates and objects)