Community Wishlist Survey 2019/Wikidata/Automatically add inverse properties

Random proposal ►◄ Wikidata  The survey has concluded. Here are the results!

Automatically add inverse properties

• Problem: It takes a long time to add inverse statements, father/mother -> child; owner of-> owned by; Spouse/Partner (including qualifiers start time, place of marriage, etc)
• Who would benefit: Everyone editing
• Proposed solution: Possibility to automatically add inverse statements, including qualifiers.
• More comments:
• Phabricator tickets:
• Proposer: Germartin1 (talk) 16:43, 9 November 2018 (UTC)

Discussion

• At the moment we have d:User:Frettie/consistency check add.js, which allows to add inverse properties manually very quickly; of course doing it automatically would be great. --Epìdosis 14:47, 11 November 2018 (UTC)
• I actually wonder why inverse properties exist in the first place. It seems like they exist so that the inverse can be referenced on other wikis. Perhaps it would be better to make it easy to reference "the inverse" in a sidebar on Wikipedia instead of having to actually duplicate the data in the database. For instance, if there was a way to say that I want the items, where the "father" is this page, it would give me a list of the page's children. That seems like it would be way better than trying to ensure that the inverse statements are correct everywhere. U+1F360 (talk) 23:22, 14 November 2018 (UTC)
• @Germartin1: How would this handle qualifiers and references, and statements with the same value but different qualifiers (for e.g. adjacent stations, offices held)? I like this proposal (and added my support vote) but I think these are things that would need to be addressed in the implementation. Jc86035 (1) (talk) 08:11, 20 November 2018 (UTC)
• I found task T209559 which seems to cover what I mentioned above. --U+1F360 (talk) 04:22, 23 November 2018 (UTC)
• @Germartin1: I use Property:P1545 as qualifier in Property:P26. And i don't want to copy the Property:P1545 value! Manu1400 (talk) 04:15, 25 November 2018 (UTC)

Voting

•  Support --Tohaomg (talk) 18:22, 16 November 2018 (UTC)
•  Support As I added above, I think the reasons why we have inverse properties should be investigated first (as their might be a better technical solution that removes the need for them). But if removing them is impossible, then we should make it faster/easier to add them. U+1F360 (talk) 21:53, 16 November 2018 (UTC)
•  Support Consulnico (talk) 00:07, 17 November 2018 (UTC)
•  Support Meisam (talk) 00:07, 17 November 2018 (UTC)
•  Support Moebeus (talk) 01:06, 17 November 2018 (UTC)
•  Support Inverse properties are trivial case of properties which can be derived from other properties. See https://phabricator.wikimedia.org/T167521 for more complicated cases. We should have some mechanism to define such properties. Maybe as "read-only" properties automatically generated if some condition is met. One danger of inverse properties is thay have to be removed in 2 places. One thing we do not want is a property added by bot, so if a person deleted one of them bots re-adds them based on the other one. Jarekt (talk) 05:47, 17 November 2018 (UTC)
•  Support Hadrianus (talk) 10:39, 17 November 2018 (UTC)
•  Support Frettie (talk) 10:57, 17 November 2018 (UTC)
•  Support Libcub (talk) 11:35, 17 November 2018 (UTC)
•  Support ديفيد عادل وهبة خليل 2 (talk) 11:44, 17 November 2018 (UTC)
•  Support Walter Klosse (talk) 11:56, 17 November 2018 (UTC)
•  Support Shisma (talk) 12:29, 17 November 2018 (UTC)
•  Support Naseweis520 (talk) 15:19, 17 November 2018 (UTC)
•  Support Blue Rasberry (talk) 15:34, 17 November 2018 (UTC)
•  Support Make this optional, and only for logged in users. Micru (talk) 16:19, 17 November 2018 (UTC)
•  Support Kette~cawiki (talk) 17:40, 17 November 2018 (UTC)
•  Support Crazy1880 (talk) 17:55, 17 November 2018 (UTC)
•  Support Theklan (talk) 18:20, 17 November 2018 (UTC)
•  Support Pichpich (talk) 19:01, 17 November 2018 (UTC)
•  Support Shev123 (talk) 19:51, 17 November 2018 (UTC)
•  Support also "has part" - "part of" and similar. Perhaps the properties could be "weak ones" - autogenerated, unless someone overrides them? Andree.sk (talk) 19:58, 17 November 2018 (UTC)
•  Support Property P190 (twinned administrative body) would be anaother good example Gelli1742 (talk) 20:23, 17 November 2018 (UTC)
•  Support JAn Dudík (talk) 20:34, 17 November 2018 (UTC)
•  Support Nk (talk) 20:39, 17 November 2018 (UTC)
•  Support Metrónomo-Goldwyn-Mayer 21:11, 17 November 2018 (UTC)
•  Support Epìdosis 22:52, 17 November 2018 (UTC)
•  Support model and implement 1:1, 1:n, n:m relationships navigable in both directions as single entities and all operations (like add & remove) as atomic, not as two inverse properties (which exposes an implementation detail). The names of the relationship's ends still should make sense, like ownes and owned by.. Herzi Pinki (talk) 22:52, 17 November 2018 (UTC)
•  Support Nickw25 (talk) 23:18, 17 November 2018 (UTC)
•  Support Liuxinyu970226 (talk) 01:27, 18 November 2018 (UTC)
•  Support NMaia (talk) 11:01, 18 November 2018 (UTC)
•  Support As is is already checked if the reverse is missing, it would "just" need a confirm botton. --Eingangskontrolle (talk) 11:28, 18 November 2018 (UTC)
•  Support VIGNERON * ^discut. 12:38, 18 November 2018 (UTC)
•  Support Juandev (talk) 13:07, 18 November 2018 (UTC)
•  Support Jeb (talk) 15:51, 18 November 2018 (UTC)
•  Support —Beleg Tâl (talk) 16:17, 18 November 2018 (UTC)
•  Support Beat Estermann (talk) 16:42, 18 November 2018 (UTC)
•  Support — Draceane ^talk_contrib. 17:26, 18 November 2018 (UTC)
•  Support Schwede66 (talk) 17:49, 18 November 2018 (UTC)
•  Support Gerd Fahrenhorst (talk) 20:26, 18 November 2018 (UTC)
•  Support Hyperik (talk) 20:32, 18 November 2018 (UTC)
•  Support Wesalius (talk) 22:04, 18 November 2018 (UTC)
•  Support Viswaprabha (talk) 23:33, 18 November 2018 (UTC)
•  Support Zeromonk (talk) 09:31, 19 November 2018 (UTC)
•  Support katpatuka (talk) 10:45, 19 November 2018 (UTC)
•  Support Geraki ^T_L 14:04, 19 November 2018 (UTC)
•  Support really this would be super helpful, not having inverse properties makes writing queries sometimes very confusing John Cummings (talk) 15:13, 19 November 2018 (UTC)
•  Support Hibm98 (talk) 16:43, 19 November 2018 (UTC)
•  Support Jmmuguerza (talk) 23:35, 19 November 2018 (UTC)
•  Support Jc86035 (1) (talk) 07:42, 20 November 2018 (UTC)
•  Support 朝彦 | Asahiko (talk) 07:54, 20 November 2018 (UTC)
•  Support Gareth (talk) 13:13, 20 November 2018 (UTC)
•  Support Vadimzer (talk) 17:26, 20 November 2018 (UTC)
•  Support Rachel Helps (BYU) (talk) 19:00, 20 November 2018 (UTC)
•  Support Joalpe (talk) 22:19, 20 November 2018 (UTC)
•  Support CAPTAIN RAJU^(T) 22:58, 20 November 2018 (UTC)
•  Support Omotecho (talk) 03:35, 21 November 2018 (UTC)
•  Support Vulphere 05:13, 21 November 2018 (UTC)
•  Support At least checking and easily creating the other end for symmetrical properties (spouse, etc.) would be great. Note we'd also need to be careful with qualifiers. References should be carried over too. Laboramus (talk) 07:14, 21 November 2018 (UTC)
•  Support Acer11 (talk) 08:01, 21 November 2018 (UTC)
•  Support Ayack (talk) 09:15, 21 November 2018 (UTC)
•  Support β₁₆ - (talk) 10:12, 21 November 2018 (UTC)
•  Support Geogast (talk) 13:36, 21 November 2018 (UTC)
•  Support Novak Watchmen (talk) 15:10, 21 November 2018 (UTC)
•  Support Hienafant (talk) 19:47, 22 November 2018 (UTC)
•  Support TomT0m (talk) 19:53, 22 November 2018 (UTC) Although something more ambitious like inferred statement would be really cool. I guess it’s a step towards that goal :) @Markus Krötzsch: TomT0m (talk) 19:53, 22 November 2018 (UTC)
•  Support I agree Poslovitch (talk) 20:34, 22 November 2018 (UTC)
•  Support Thibdx (talk) 20:47, 22 November 2018 (UTC)
•  Support Wostr (talk) 20:50, 22 November 2018 (UTC)
•  Support. Useful. Nomen ad hoc (talk) 21:51, 22 November 2018 (UTC).
• Strong oppose I'm sorry but this is NOT a good idea for several reasons: (1) We have several properties that are stated as "inverses" but are not really - for example "part of" vs "has part"; for concrete entities they are pretty close to exact inverses, but in more abstract cases not so much, and there are things like "has parts of the class" that are better ways to handle the relation. (2) Do we want every entity in the universe listed under the "universe" item, since everything is "part of" the universe? This is a very general issue: many relations that are invertible are "one to many" type, where there would be only one statement on thousands of items on one side, with thousands of statements on the one item on the other side. Sometimes for many-to-many relations they are close to "one-to-many" in some contexts and "many-to-one" in others - that is, sometimes one side of the relation and sometimes the other side would be unreasonable to reify as statements on items. (3) We would need to be very careful with vandalism of properties: if a vandal stated that some other property was an "inverse of" a widely used property like for example P31 (instance of), and the system automatically created millions of such statements, we would have a terrible mess to clean up. (4) There is already a javascript gadget to do this at least as far as the UI goes - 'User:Pasleim/derivedstatements.js' - maybe that needs some improvement or cleanup, but an approach like that is the only way I could see this being useful, i.e. actually generating the inverse statements in real time based on current data, rather than inserting inverse statements into items in a more permanent manner. ArthurPSmith (talk) 16:39, 23 November 2018 (UTC)
•  Support Sahaquiel9102 (talk) 17:34, 23 November 2018 (UTC)
•  Support Nvrandow (talk) 20:04, 23 November 2018 (UTC)
•  Oppose (with the understanding that oppose votes are not counted here) per Arthur in full. I think this is useful for properties where the likelihood of introducing unnecessary bloat to items is very low, but I am against using this functionality with properties that have a strong likelihood for bloat (such as "part of", "has part", and most recently "owner of" and "owned by"). Mahir256 (talk) 21:07, 23 November 2018 (UTC)
• Strong oppose per Mahir256 and Arthur.Winged Blades of Godric (talk) 06:21, 24 November 2018 (UTC)
• Neutral Manu1400 (talk) 04:12, 25 November 2018 (UTC)
•  Support Ranjithsiji (talk) 22:27, 25 November 2018 (UTC)
•  Support — AfroThundr ^{(u · t · c)} 02:59, 26 November 2018 (UTC)
•  Support Sebastian.Dietrich (talk) 18:47, 26 November 2018 (UTC)
•  Support YFdyh000 (talk) 20:00, 27 November 2018 (UTC)
•  Support Mhmrodrigues (talk) 20:17, 28 November 2018 (UTC)
•  Support Only as optional, not as mandatory. It can be either per property (e.g. father/mother -> child is automatic, part of is not) or as an option (Preferences, pop-up etc.) — NickK (talk) 15:32, 30 November 2018 (UTC)