Wikimedia Bugzilla is closed!

I've been adding a lot of "instance of" (P31) properties to Wikidata objects recently. Specifically, I've been adding a lot of instance of properties that point to "railway station" (Q55488).

The autocomplete results are less than ideal. Once you've typed "railway s", this is the list that comes up: http://cl.ly/image/3d1V3w2F0E3G

The index used to generate the autocomplete results could be based on the use of those objects in the database. If I've selected P31, why not have it so that the results that are returned are ordered by the frequency of use of that object in P31-propertied statements? If I've typed in "railway s", then "railway station" (Q55488) should appear higher than "Railway stations in Cromer".

If doing the autocomplete indexing on a per-property basis is too computationally intensive or storage intensive, it might be worth considering doing it for all properties. Because of categorisation properties like P31, it'll be likely that there'll be a power law distribution of objects that are used as the object-term in statements.

This will make it easier for humans to consistently add properties to Wikidata objects.