Open main menu

Wikidata:Property proposal/Model item for

model item forEdit

Originally proposed at Wikidata:Property proposal/Generic

   Not done
Descriptiondefines which item is a best practice example of modelling a subject, which is described by the value of this property
Data typeItem
Allowed valuesItems which are classes
Example 1Douglas Adams (Q42)human (Q5)
Example 2Nelson Mandela (Q8023)politician (Q82955)
Example 3Zootopia (Q15270647)animated film (Q202866)
Example 4Mona Lisa (Q12418) -> painting (Q3305213)
Planned useRecord model items for all of the most used classes on Wikidata, then gradually expand over time to include more granular concepts
Expected completenessalways incomplete (Q21873886)
See alsomodel item (P5869)

MotivationEdit

Defines which item is a best practice example of modelling a subject, which is described by the value of this property. Providing best practice examples linked from the subject will make it much easier for people to understand how to model types of items, especially new people. The main hope is that this will lead to more consistent modelling of items rather than different editors making up their own structure. It will also allow a central place to decide how to model certain kinds of items.

By modelling this within the structure of Wikidata rather than in Wikiprojects the structure is multilingual. Also looking at where the model item refers to the concept we can run a standard query that will work for most cases to find all the items that the model applies to, this is not limited to ‘instance of’. This could be used to check which items do and do not have high importance statements, e.g authors without a date of birth.

This could later be complemented with another property which defines the schema for a class of item (e.g author or politician)

This property requires that the inverse property Model item also be accepted to be most useful.

Thanks

John Cummings (talk) 20:15, 11 September 2018 (UTC)

DiscussionEdit

  •   Support - NavinoEvans (talk) 21:04, 11 September 2018 (UTC)
  •   Support Jane023 (talk) 06:30, 12 September 2018 (UTC)
  •   Support David (talk) 08:20, 12 September 2018 (UTC)
  •   Support very good idea, but   Comment what is the intended extent? should all classes have a model or just the upper classes? I see animated film (Q202866) in the example but wouldn't film (Q11424) be better? (both way are fine, but too many model would probably end up as confusing as no model at all) Cdlt, VIGNERON (talk) 08:33, 12 September 2018 (UTC)
Thanks very much @VIGNERON:, I don't think these items will define the schema for that class, just act as good practice examples for people to copy. I'm currently working on an idea for schema modelling using a new property. My guess is having model items for more granular items like 'author' or 'animated film' will be helpful and that you'd want for people to be specific, modelling authors from the model item for 'human' would probably mean people miss out a lot of the author specific details. Same for animated film, e.g animators, software used, rendering engine, voice actors (by language version) etc are all specific to animated films (and some specific to computer modelled animated films even). I don't know how granular you would want to go though, do we want a model item for '16th Century Welsh poets'? I think that's a community decision. --John Cummings (talk) 09:34, 12 September 2018 (UTC)
@Jura1:, I suggest that both are needed to help people navigate, they're not exactly inverse properties of each other as a model item could be the model for several subjects e.g Douglas Adams could be the model for 20th Century British authors and 20th Century British Playwrights. Having both will be more valuable than only having one of them. --John Cummings (talk) 12:19, 12 September 2018 (UTC)
I rather debate on the talk page of "20th Century British authors" about the model status of Q42 than the inverse. --- Jura 12:37, 12 September 2018 (UTC)
I think there is some confusion, do you mean you want to have a discussion on the schema for '20th Century British authors' on the item for '20th Century British Authors'? If so then I assume there is some way of setting up a bot that adds a note to the talk page for model items (e.g Douglas Adams) that directs you to the correct location (e.g item for 20th century British authors). I also think this this unlikely to be an issue once we've implemented some way of describing schemas within the items. --John Cummings (talk) 12:51, 12 September 2018 (UTC)
We already have. So even more so. You can always query it. --- Jura 12:57, 12 September 2018 (UTC)
Hi @Pasleim:, when you say enough, for which use cases do you mean? --John Cummings (talk) 22:08, 12 September 2018 (UTC)
It is enough because all the statements can be inferred from Wikidata:Property proposal/Model item. For new users, for whom this property is mainly, it will make most sense to create a list of all model items. To create such a list you only need one of the two properties. If you have both properties you don't gain anything but your maintenance work will be doubled. --Pasleim (talk) 10:08, 17 September 2018 (UTC)
  •   Oppose. Better to point in the other direction. --Yair rand (talk) 22:33, 12 September 2018 (UTC)
@Yair rand:, can you explain why only one direction is better? Surely having two directions will help people to find useful items? These properties are not exactly inverse of each other because an item can be a model item for more than one class. --John Cummings (talk) 22:54, 12 September 2018 (UTC)
  • This is actually not relevant information about Douglas Adam. --- Jura 08:54, 13 September 2018 (UTC)
  • @John Cummings: I don't see how being a model item for more than one class would make it not an exact inverse. If there's a statement going one way, there would be a statement going the other way. (That is, if X is a model item for both Y and Z, this property would point from X to both, and the inverse would point both to X from both Y and Z.) With both directions, we have two versions of the data, which are not automatically in sync. If someone changes one statement but does not change the other statement, one version will be incomplete. Requiring two actions for each change will cause the data to be less reliable. --Yair rand (talk) 19:36, 16 September 2018 (UTC)
@Yair rand:, thanks. I think this is an issue that needs resolving, that inverse properties are very useful for querying and finding items but are problematic for staying in sync, do you know if anyone has proposed a bot or any other solution to keep them in sync? Not having inverse properties feels like throwing the baby out with the bath water... I will do some looking around at possible solutions John Cummings (talk) 09:31, 17 September 2018 (UTC)
@Germartin1:, could you explain why you think we shouldn't have this property? Thanks, --John Cummings (talk) 09:27, 13 September 2018 (UTC)
First of all, it's an inverse of your other proposal, hence redundant according to Wikidata principles. Personally, I don't see how it adds any kind of value to an item. Germartin1 (talk) 09:37, 13 September 2018 (UTC)
@Germartin1: I don't fully understand why this is against Wikidata principles? Is there a guideline or info page you can point? We have around 100 inverse property pairs at the moment. For example, template's main topic (P1423) and topic's main template (P1424) ... or ... is a list of (P360) and has list (P2354). NavinoEvans (talk) 10:43, 13 September 2018 (UTC)
  •   Support Useful, particularly with inverse property. See no harm in using both in supporting developing best practice and some kind of consistency. Stinglehammer (talk) 10:31, 13 September 2018 (UTC)
  •   Support Yeah, can see this being helpful. Lirazelf (talk) 11:06, 13 September 2018 (UTC)
  •   Support Very useful Jason.nlw (talk) 13:21, 13 September 2018 (UTC)
  •   Oppose Nepalicoi (talk) 12:21, 17 September 2018 (UTC)
Hi @Nepalicoi:, can you explain why you oppose the proposal? Thanks, --John Cummings (talk) 14:37, 17 September 2018 (UTC)
  • Given that Douglas Adams (Q42) and Nelson Mandela (Q8023) are item that's violate our rules, I don't think it does a reasonable job to present them as a model item and pretend they are best practice. If we have model items they shouldn't conform to our policies. The fact that you have a 50% rate of examples that violate our rules which might be a lot higher then the average rate of items that violate our rules, make me question the concept a bit. ChristianKl❫ 17:32, 25 November 2018 (UTC)
  •   Oppose I don't see a reason to create an inverse property for model item (P5869). − Pintoch (talk) 16:26, 3 February 2019 (UTC)
  •   Oppose No need of inverse property. Snipre (talk) 06:57, 26 February 2019 (UTC)
  Oppose As others said above, model item (P5869) is sufficient. --Tinker Bell 19:40, 15 April 2019 (UTC)

  Not done per discussion above, the inverse is better and appears to be working well. Bidirectional linking is not necessary here, because if you are looking for a model, you will always be starting from the class. If you're navigating starting from the model item, and really want to know what class it is attempting to model, just follow the link to "instance of". --99of9 (talk) 07:44, 17 April 2019 (UTC)