Hello everyone,
I was playing around with a recent wikidata dump and extracted the items
that "looked" like classes based on the definition here,
https://www.wikidata.org/wiki/Wikidata:WikiProject_Ontology/Classes
Specifically, an item is a class-item if any of the following are true,
* the item is the value of a P31 ("instance of") statement
* the item has a P279 ("subclass of") statement (subclass)
* the item is the value of a P279 ("subclass of") statement (superclass)
Once I extracted all items that met these criteria (2,399,621 items
from wikidata-20190603-all.json.bz2) I started examining the results. One
of the things I found slightly surprising is that there are about 23k
badminton events that are classes b/c they have "subclass of
https://www.wikidata.org/wiki/Q13357858" statements. SPARQL query below.
https://query.wikidata.org/#SELECT%20%3Fitem%20%3FitemLabel%20%0AWHERE%20%0…
It also looks like there is a badminton project page,
https://www.wikidata.org/wiki/Category:WikiProject_Badminton
https://www.wikidata.org/wiki/Wikidata:WikiProject_Badminton/Subclass
I'd like to remove these statements as it seems that a particular instance
of a badminton tournament
https://www.wikidata.org/wiki/Q121940
is not a class.
It seems that this pattern is also in place for about 1,000,000 items which
are instance of gene (e.g.
https://www.wikidata.org/wiki/Q40108).
I had a couple questions for the mailing list,
1) do folks know if there is an active group working on wikidata ontology
2) i've read a few messages about shape expressions. would it be
worthwhile to setup a shape expression that prevents most items from having
both "instance of" and "subclass of" statements?
3) if these entries are generated by bots, what is the best way to get in
touch with the owner, their user talk page?
I am probably missing a lot of information about what has been done so far
in the community, but I'm happy to read anything someone points me towards.
best,
-Gabriel