In T244341: Stop using blank nodes for encoding SomeValue and OWL constraints in WDQS, the question arose whether we should use “some value” or “unknown value” terminology. The UI and the data model primer use “unknown value”, but the full data model uses “some value”, and @Jheald has argued that this matches the current use of that snak type in Wikidata better (see T244341#5862287). We’d like to get some number on this, but it doesn’t seem to be possible on the public query service – even for a single property, a query like the following times out:
SELECT ?property (COUNT(*) AS ?count) WHERE { ?statement ps:P20 ?unknown. FILTER(ISBLANK(?unknown)) ?statement ?pq ?qualifier. ?property wikibase:qualifier ?pq. } GROUP BY ?property ORDER BY DESC(?count)
More generally, we would be interested in the number of qualifiers of statements with “unknown value” in the main snak, aggregated by main statement and qualifier property. In SPARQL, that would be more or less:
SELECT ?mainProperty ?qualifierProperty (COUNT(*) AS ?count) WHERE { ?mainProperty wikibase:claim ?p; wikibase:statementProperty ?ps. ?qualifierProperty wikibase:qualifier ?pq. ?subject ?p ?statement. ?statement ?ps ?unknown. FILTER(isBlank(?unknown)) ?statement ?pq ?qualifier. } GROUP BY ?mainProperty ?qualifierProperty
(This doesn’t count statements without any qualifiers, but we could probably get those numbers separately for the top properties we’re interested in.) This may be more efficient to query with a different technology, though (Hadoop?).