The coherence beetween the modelling principles we adopt and common language definitions is a strong argument to me. This means we're not that wrong, and that any person can understand. It's just essential that the two are consistent, as we're trying in WIkidata to define topics the best we can, just as Wikipedia articles does. This implies that we also will get some kind of international language with statements that reflects the subtelety of the definitions of the different languages.

I also strongly values the "1 item = 1 definition" principle, even at the cost of item disconnections if definitions of some close notions in different language.

2015-10-19 13:09 GMT+02:00 Gerard Meijssen <>:
Semantics prove what? Your excercise with pizza has me wonder what you call food. The definition of food as in the French Wikipedia is not an argument for me. My understanding of French is not sufficient and, arguably any language has its own approach. 

On 18 October 2015 at 19:34, Thomas Douillard <> wrote:
Yes, in the pizza example Emw showed, the definition of "food" is important. If what I ate this morning is a food, then pizza is a subclass of food. This is consistent with the first sentence of of frwiki. And the fact that "pizza" is an instance of food is a mistake, unfortunately a pretty common one on Wikidata. We should write a query to find all such examples where an item is both an instance and a subclass of the same class.

Now pizza is clearly a type of meal it could be relevant in a food classification and could be very well be an instance of it, as it's a preparation common people used to put whatever they can put on it, similarly to

2015-10-18 18:31 GMT+02:00 Thad Guidry <>:
The main problem is that Instance Of is not being used properly sometimes. In general, wrong classifications across Wikidata lead to weird assumptions.

Better documentation, and even helper rules to help prevent wrong classifications is what is needed and its forthcoming.

Lydia has mentioned that these kinds of problems will eventually become less and less as the Roadmap features eventually land into production.

I am looking forward to next year, and the year after, to see the quality improve.

Wikidata mailing list

Wikidata mailing list

Wikidata mailing list