On 26 January 2016 at 11:24, Magnus Manske magnusmanske@googlemail.com wrote:
On Tue, Jan 26, 2016 at 7:33 AM Pete Forsyth peteforsyth@gmail.com wrote:
(Note: I'm creating a new thread which references several old ones; in
the
most recent, "Profile of Magnus Manske," the conversation has drifted
back
to Wikidata, so that subject line is no longer applicable.)
Andreas Kolbe has argued in multiple threads that Wikidata is
fundamentally
problematic, on the basis that it does not require citations. (Please correct me if I am mistaken about this core premise.)
Every statement on Wikidata /should/ be referenced, unless the statement itself points to a reference (e.g. VIAF, images). However, at the moment, this is not a requirement, as Wikidata is still in a steep growth phase. Over the last few years, many statements were added by bots, which can process e.g. Wikipedia, but would be hard pressed to find the original reference for a statement.
To extend Magnus' point... This is also the case on Wikipedia. Every Wikipedia sentence /should/ be verified to a reliable source, and those without footnotes can be removed. But, it is not a /requirement/ that every statement be verified. In short - 'verifiable not verified' is the minimum standard for inclusion of a sentence in Wikipedia. The ratio of footnotes-to-sentences in Wikipedia articles is on average probably much lower than the ratio of references-to-statements in Wikidata. It's just that we have more easily available /quantitative/ statistics for Wikidata that we do for Wikipedia, which makes it easy for Wikidata-critics to point to the number of un-referenced statements in Wikidata as a simple measure of quality, even though many of them DO meet the "verifiable, even if not yet verified" minimum standard that we accept for "stubs" on Wikipedia.
For example: even in a Feature Article Wikipedia biography, I've never seen a footnote /specifically/ for the fact that the subject is "a human". That reference is implied by other footnotes - citing for the birthdate, or occupation for example. By comparison, in Wikidata, some people seem to be a feeling that statements like "instance of -> human", "gender-> male" need to be given a specific reference before they can be considered reliable. This is even when there are other statements in the same Wikidata item that reference biography-authority control numbers (e.g. VIAF).
Yes, ideally, every statement could be given a reference in Wikidata, but ideally so should every sentence in Wikipedia. In reality we do accept "stub" Wikipedia articles that have 5 sentences and 1 Reliable Source footnote. Furthermore, we also do also have Wikidata properties that are, in effect, "self verifying": like the "VIAF identifier" property - which links to that authority control database, or the "image" property - which links directly to a file on Commons. So, simply counting the number of statements vs. the number of references in those statements on Wikidata and concluding that Wikidata is therefore inherently unreliable is both simplistic and quite misleading.
-Liam
wittylama.com Peace, love & metadata