-
Notifications
You must be signed in to change notification settings - Fork 270
Wikidata
Wikimedia started to move data from the Wikipedias to Wikidata. DBpedia will need to adapt to this one way or other.
Also see issue 30 and issue 35.
Currently (April 2013), almost all inter-language links have been moved from the Wikipedias to Wikidata. Other data probably hasn't changed much.
One important question / distinction: which subject URIs and which domains should DBpedia use to publish Wikidata data?
- Use http://data.dbpedia.org domain and subject URIs and publish new dataset files for download.
- Merge Wikidata data into existing datasets and use the current domains, e.g. http://dbpedia.org and http://fr.dbpedia.org.
These don't contradict each other and we should probably do both.
How should DBpedia construct specific http://data.dbpedia.org URIs? Wikidata item IDs are not human-readable, and not all have English labels. For example, Q5849921 is an item corresponding to [Estático](http://es.wikipedia.org/wiki/Estático_(canción_de_Zurdok\)) on the Spanish Wikipedia, but no corresponding English Wikipedia article. DBpedia URIs for Wikidata items should probably look like http://data.dbpedia.org/resource/Q5849921. We could use English labels where they exist and other languages where they don't, but then the URIs will be prone to change.
In the near term, it's probably enough to extract inter-language links from Wikidata and the few Wikipedia pages where they still exist, merge them and publish them as above.
In the medium term, DBpedia should extract data that will be moved from Wikipedia infoboxes to Wikidata items.
In the long term, almost all Wikipedia data will move to Wikidata. DBpedia should maybe concentrate on extracting and processing natural language text.