Wikidata Graph Split

Technical details on the Split

See: https://m.wikidata.org/wiki/Wikidata_talk:SPARQL_query_service/WDQS_graph_split

Scholarly graph:
   1. Find subjects whose P31 is a scholarly article (Q13442814) and find all quads whose context matches those subjects.
   2. Find the references for the elements in 1.
   3. Get the values for the elements in 1 and 2.
   4. Add those together to produce the scholarly graph.
   Main ("non-scholarly") graph:
   5. From the full graph, subtract the items identified in 1.
   6. Then remove from 5 the references and values that are only attached to the scholarly graph, but keep any other references or values - 

• Entity classes that are intended to be included in the split: https://docs.google.com/spreadsheets/d/1eKX_2Z1rXj1s_zOapQvn_0uD6MVhc-qyqqxbn5loIvk/edit?gid=0#gid=0
• Additional information: https://m.wikidata.org/wiki/Wikidata:SPARQL_query_service/WDQS_graph_split/WDQS_Split_Refinement

Endpoints:

wikidata-main

https://query-main-experimental.wikidata.org/

wikidata-scholarly

https://query-scholarly-experimental.wikidata.org/

Which entites are were?

• queryable location of an entity was discussed (https://m.wikidata.org/wiki/Wikidata:SPARQL_query_service/WDQS_graph_split/WDQS_Split_Refinement) but is currently not queryable in the demo endpoints.

• The current endpoints only include entities of type scholarly article
  • the proposed additional subtypes/additional types are not included
• If a entity has two entity classes e.g. scholarly article and an additional one the entity is still included in main
  • I do not know if this behavior is intended

Example

SELECT DISTINCT ?valueLabel (count(?valueLabel) as ?count)
WHERE 
{  
  ?entity wdt:P5008 wd:Q112895606; wdt:P31/wdt:P279* wd:Q1266946 .  
       ?entity p:P921 ?prop . OPTIONAL { ?prop ps:P921 ?value }  
       SERVICE wikibase:label { bd:serviceParam wikibase:language "[AUTO_LANGUAGE],en". }
      }
GROUP BY ?valueLabel
ORDER BY DESC(?count)

• ?entity are of type thesis, which is of the candidate list of split entities. Thus, the query is currently not affected but it can be in the future
• ?prop is of type subject and is in the main graph, thus a federated query needs to be used