Re: [Wiki-research-l] [Wikidata] The basis for Wikidata quality

Hoi, You are conflating two things that are not related. ORES is really helpful and there is plenty of room for it to function extremely well on Wikidata. Yes, ORES will do good things for Wikidata but it is separated from the proposed item quality. When an item is created because of the existence on the Kyrgyz Wikipedia, it may not even have a label. This happens a lot and Amir, the developer of ORES has a bot that he regularly runs to add one label. The name of the article. When an article is added the same bot does the same work. We do not have to rate it for a specific "level of quality" when we are to improve our quality, this bot runs every week automatically. When an automated rating system sees this, it can perform the remedy.. When as I understand from Lydia, the ratings are done automagically, in an ORES kinda way, it gains relevance. The point though is that still the quality of all the individual items on their own is of hardly any significance. It becomes relevant when it can accept the result of a query and provide results on the basis of that.. For instance.. all the items with "catalog" "black lunch table" and give me a rating for all the articles with no articles or no articles in Spanish... The problem with rating "item quality" is that on its own, it has no application. I just finished adding award winners [1]. Based on the quality of the English article, I added the award winners using "linked items" and "petscan". I added a few items because as there are no articles. The Turkish article for the award has plenty of red links. It would be a quality improvement when these red links are associated with the items. An article writer immediately finds the English article and the Wikidata statements. That is actionable quality as it provides a way to stabilise both Wikidata, en.wp and tr.wp. For the Turkish language the labels of the red links may be used. When you consider quality, it is like an onion. On the first level all the information is there, on the second level, we may be missing education (that may still be in Freebase).. an employer. whatever makes sense in a context When there is not even a red link but a text like for English, a tool like ORES could recognise the label in the text and accept it as a result that is probably positive. When we are really interested in quality, we need to compare the content of the many projects including Wikidata and find that it is in balance. My key criticism stands of the current quality standard: this is a first step but there are severe doubt of the relevance of several aspects. As a first iteration it will prove what is good and where it needs improvement. But without an interface into query it is useless. Thanks, GerardM [1] https://tools.wmflabs.org/reasonator/?q=Q13582570 On 22 March 2017 at 15:33, Aaron Halfaker <aaron.halfaker(a)gmail.com> wrote: