Monday, February 09, 2015

#Wikidata - automated #descriptions are GIGO

The great thing about automated descriptions of Wikidata items is that they reflect what is there. The consequence is that when it is not good, it will show.

I discussed automated descriptions with my friend Amir and he pointed me to that "East Jerusalem" is not a "world heritage site". He is absolutely right; world heritage site is a qualification of whatever it is that has been recognised in this way.

Amir also pointed out that whenever labels are missing, automated description will give you something in another language. This is not a bug, it is a feature. All it takes is for someone to add the missing label and wherever it is used, the label will show in the "current" language in future. One alternative is NOT showing statements when they have no labels. Another alternative is to make it configurable if you want as much as possible or information only in the current language.

The point about automated descriptions stands; adding one statement may impact descriptions in all languages. Adding one label will impact all items that include the associated statement for that language.

Obviously an automated description is superior to no description. Arguably automated descriptions are superior to manual description because they reflect the item in any and all language.
Post a Comment