On 19.08.2014 22:23, David Cuenca wrote:
...
Actually I have one last question :) At the moment Gerard is using "is
a list of:<value>" on category item pages which has the effect of being
the inverse of "instance of". And then he adds further conditions as
qualifiers, see:
https://www.wikidata.org/wiki/Q6562
While this method of works for simple categories, more complex ones
would be hard to model using this method, like
https://www.wikidata.org/wiki/Q8380098
I was thinking of modelling it like:
<Category:Discoverers of extrasolar planets> is a list of <human>
<Category:Discoverers of extrasolar planets> has items used as value of
<discoverer>
Of course it would require to have a link between the item "discoverer"
and the property "discoverer", but would that make sense?
Well, it depends on what the intended use of "is a list of" is. First
note that it is not the inverse of "instance of" (the inverse of a
relation R holds between all pairs where R holds, just in the opposite
direction; this is not what happens here). Rather, "is a list of"
describes some class that all of the elements of a list are instances of.
I don't think that it one should try to capture *exactly* what the items
on the list are. Many lists are based on complicated criteria and it
would be very hard to express them in a good way using statements. What
you suggest above would be an ad hoc solution (a.k.a. hack) for a few
cases; many other cases would need different features. Even if one would
have a way to capture some lists exactly, one would need to document
this very carefully in order for the information to be useful to others.
In essence, one would specify a query language there. Since we already
are working on queries for Wikidata, the better way to solve this in the
future would be to refer to actual queries (as soon as they are
expressive enough).
Anyway, as I understand it, Gerard is adding these statements to help
with the organisation of lists (and to give some more relevant
statements to list items, e.g., to assist the auto description). Since
we want to support automated list generation in the future (using query
results on Wikipedia pages), it might be handy to have some overview of
the lists (how many, about which topics, etc.). But I am just guessing
here -- maybe Gerard has other reasons too.
Cheers,
Markus