User talk:ChristianKl/Draft:New Ranks

Thoughts

Interesting piece. The one area where I think an additional rank would definitely be useful would be for suggestions from automated processes (eg automated image analysis in the case of images), where the suggestion from the automated process is worth recording, but the process didn't regard its probability as beyond question. I would prefer a label of "statement requires assessment" for this rank level. It may also be that the community would prefer such statements to be held in an arms-length separate staging area, rather than written to Wikidata itself (compare eg how Mix'n'match stores its unconfirmed automated matches off-wikidata, rather than writing them to Wikidata). But I suspect the volume and usefulness of automated image analyses may push the balance towards inclusion albeit with modified rank, at least for image pages on Commons Structured Data. A workaround for the present would be to give these statements deprecated rank with a special reason for deprecation = "automated assessment, needs confirmation". But a specific rank for this would I think ultimately be preferable.

As for false statements, I think the existing 'deprecated' rank, with an appropriate 'reason for deprecation' handles them adequately. In terms of retrieval, it doesn't seem to me so much easier to write a query that tests for a specific set of ranks, compared to writing a query that allows a specific set of reasons for deprecation -- not enough (IMO) to justify the complexity increase of adding another rank for this. Jheald (talk) 11:54, 7 February 2020 (UTC)

It's also worth noting we have statement supported by (P3680) to indicate which out of a set of discussed possibilities a reference actually prefers. Jheald (talk) 12:26, 7 February 2020 (UTC)

I don't see how statement supported by (P3680) solves the problem. If I look at the property proposal it's not about having an statement where multiple sources have different views. Do you have an example where the property gets used as you think it should? ChristianKl ❪✉❫ 17:40, 12 February 2020 (UTC)

Use "reason for deprecation"

These both seem to be refinements on the current "deprecated" rank (as neither is intended to appear in "truthy" statements). I think better would be to use the existing qualifier mechanism to clarify - in particular we have reason for deprecated rank (P2241) which can have new items as values to handle these cases, if existing items aren't sufficient. ArthurPSmith (talk) 14:20, 7 February 2020 (UTC)

In the discussion on the project chat there are people who see "uncertain" more of a subclass of "normal" then as a subclass of "deprecated". I think most users see deprecated as a judgement that a value isn't right. Using deprecated in a way where that's not meant (and a reason for deprecated rank (P2241) value is used to indicate that it isn't meant) is confusing for users. ChristianKl ❪✉❫ 10:27, 13 February 2020 (UTC)