I would be surprised if that theory held true. I expect that both very abstract (fruit) and extremely specific (golden delicious) items would have a lower sitelink count than the "golden layer of most useful terms" (apple) in the hierarchy (I am reminded of the theory of word length and term frequency in linguistics).

But I would assume that indeed in the subclass hierarchy that Wikidata will eventually exhibit would have such a "golden layer" (and that these terms are not randomly distributed over the hierarchy).

Would be fun to examine :)


2013/9/24 Klein,Max <kleinm@oclc.org>
Hello All,

It struck me that one interesting way to see if subclasses are useful was to test this hypothesis.

Let QID_a and QID_b be two Wikidata items.

Conjecture: if QID_b is subclass of QID_a,
then count_stelinks(QID_b) <= count_sitelinks(QID_a).

Has anyone investigated this problem, or can think of an efficient way to test it? Or can tell me why it ought not to be true?

Maximilian Klein
Wikipedian in Residence, OCLC

Wikidata-l mailing list

Project director Wikidata
Wikimedia Deutschland e.V. | Obentrautstr. 72 | 10963 Berlin
Tel. +49-30-219 158 26-0 | http://wikimedia.de

Wikimedia Deutschland - Gesellschaft zur Förderung Freien Wissens e.V. Eingetragen im Vereinsregister des Amtsgerichts Berlin-Charlottenburg unter der Nummer 23855 B. Als gemeinnützig anerkannt durch das Finanzamt für Körperschaften I Berlin, Steuernummer 27/681/51985.