Wikidata:Property proposal/star classification

star classification

   Not done

Description	Rating system name and star rating (see voy:en:Rating systems, hotel rating (Q2976556))
Data type	Item
Template parameter	estrellas in voy:es:Plantilla:Listado at Template:Listing (Q14330485)
Domain	WikiVoyage listings
Allowed values	Text: name of rating system and 1-5 star rating level in that system. For restaurants and hotels; only useful if this specifies which rating system is in use.
Example	⟨ Leonardo Hotel Berlin City West ⟩ star classification Search ⟨ Hotelsterne ★★★★☆ Garni ⟩

•  Oppose Needs a structured data model, with separate items for each rating system, and items for each possible rating. See, for example, how we do film (movie) ratings. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 13:30, 16 May 2016 (UTC)
•  Oppose Per Andy. --Srittau (talk) 17:49, 16 May 2016 (UTC)
  • @Srittau: I don't understand your argument, can you explain what you are looking for? How this isn't being met by K7L's proposal?
    --- Jura 17:52, 16 May 2016 (UTC)
    • We would need structured data for this. For example
      ⟨ Super Duper Hotel ⟩ rating (P4271) ⟨ Michelin 2 stars ⟩
      , where Super Duper Hotel and Michelin 2 stars are items. I would  Support such a proposal. --Srittau (talk) 17:56, 16 May 2016 (UTC)
      • I suppose, you could actually do that with this proposal, if I understand it correctly.
        --- Jura 17:59, 16 May 2016 (UTC)
      • My proposal would have
        ⟨ Super Duper Hotel ⟩ rating (P4271) ⟨ Michelin ⟩
        , qualified with value=2 stars. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 18:16, 16 May 2016 (UTC)
        • While I would also support that proposal, the way I described is more flexible, in my opinion, because it can be used for non-numeric ratings and those ratings can have properties of their own. This is how the ratings properties like MPA film rating (P1657), FSK film rating (P1981), etc. work. --Srittau (talk) 18:22, 16 May 2016 (UTC)
•  Support for the proposal with the fixed example. --Srittau (talk) 18:34, 16 May 2016 (UTC)
•  Support, whether the taget items are the ones of the classification (qualified with a value) or the ones for specific number of stars in that classification. Thierry Caro (talk) 11:15, 20 May 2016 (UTC)
•  Oppose I think the rating system and the value need to be stored separately, and the film (movie) rating analogy seems a good one. Something like
  ⟨ Super Duper Hotel ⟩ rated by Search ⟨ Michelin ⟩
  rating (P4271) ⟨ 2 <star> ⟩
  ("star" being an item used as a unit) or have separate properties for each rating system, e.g.
  ⟨ Super Duper Hotel ⟩ Michelin rating Search ⟨  2 <star> ⟩
  . Thryduulf (talk: local | en.wp | en.wikt) 10:57, 21 May 2016 (UTC)
  • Actual movie reviews use review score (P444). This may be closer to those included by Wikivoyage.
    --- Jura 11:25, 21 May 2016 (UTC)
    • review score (P444) does seem like it would work well here. Thryduulf (talk: local | en.wp | en.wikt) 12:33, 21 May 2016 (UTC)
      • However, review score (P444) uses string datatype, which is different from the proposed. Lymantria (talk) 05:41, 23 May 2016 (UTC)
        • Yes, actually review score (P444) is an example of how things should not be done. --Srittau (talk) 06:48, 23 May 2016 (UTC)
          • The "better" way would be "score X qualifier out of Y". Video games and other works also use review score (P444). This is because review scores come and go and change format on a regular basis that isn't always translatable to item-version easily. How else would you deal with "37/40" or "91%"? --Izno (talk) 09:12, 24 May 2016 (UTC)
            • By storing them all as percentages; that would make it easier to aggregate, average or compare them. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 13:48, 24 May 2016 (UTC)
              • There is consensus at least at en.WP that "36 of 40" is not a percentage and cannot be reduced to its percent equivalent at Wikipedia. I suspect other projects have the same concerns with such an approach. Similar for 4.5 of 5 stars or any other rating mechanism (not all of which will be numerical in nature, of course, but that's a side point). --Izno (talk) 16:38, 24 May 2016 (UTC)
                • That's about as meaningful as a consensus that the Earth is flat. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 19:58, 25 May 2016 (UTC)
                  • That's a personal attack on a whole lot of active, productive, users. Please refrain from such. You can disagree, and that's your prerogative, but casting aspersions on the value of their contributions is not permitted. --Izno (talk) 23:27, 25 May 2016 (UTC)
                    • Like bollocks is it a personal attack. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 20:10, 28 May 2016 (UTC)
•  Comment seems to be ready for creation. Given the support of 3-4 users, including its proposer (KL7). I would created it myself if I wasn't opposed to properties for ratings in general.
  --- Jura 08:03, 11 June 2016 (UTC)
  • This does indeed have some support. It also has opposition - and this is not a vote. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 14:15, 29 June 2016 (UTC)
• I'm going to formally  Oppose this per my comment above. Review scores need to be reviewed (heh) in their implementation on Wikidata. --Izno (talk) 18:06, 27 July 2016 (UTC)
  • @Izno: I agree with you that review scores are a hard thing to model, but star classification is not a review system. Each classification system has a fairly limited number of possible classifications, which means that the suggested modelling would work well for that, in my opinion. --Srittau (talk) 19:04, 7 August 2016 (UTC)
    • @Srittau: I've seen star systems out of 10, out of 5, out of 3, with and without decimals, which sometimes are the easy ".5" and sometimes not. And essentially, these are review systems which are reassessed on some periodic basis. That this integer case out of 5 presents itself doesn't mean we shouldn't use a more general scheme to deal with this, and that more general system is review score. Better still would be, as I suggested above, "review score" with an "out of" qualifier. --Izno (talk) 11:34, 8 August 2016 (UTC)
      • I like your proposal for "review score", but I do not think it would work in this case. If you look at the example, you see four different things encoded in the value: the star system used ("Hotelsterne"), the number of stars (four), the total number of stars (five, which I consider a property of the system, not of the hotel), and the "Garni" status. This is too much to encode into qualifiers, in my opinion, especially since we are looking at a relatively small number of possible discrete values, many of which would usefully have properties of their own. -Srittau (talk) 12:04, 8 August 2016 (UTC)
      • And to add to the last argument a bit: If you have a look at [1] (in German, but it should be enough to get a general idea), you will see that the stars each have a fixed set of criteria. This is something that differs quite a bit from "traditional" review scores like 7.5/10 or 88/100. These criteria are an example of what could be encoded in the star value items. --Srittau (talk) 12:08, 8 August 2016 (UTC)

        Currently review score (P444) works mostly like your description already (though I'm not sure what "Garni" status is--?). I don't see how this is too much to use qualifiers for--they don't need to be properties of the object in question.

        I don't see how it's valuable to capture other organization's meanings for these stars--but especially, that particular organization's. Even if you do capture that, somehow, you've lost the useful data comparison, no? You can't use constraints to represent that data either. --Izno (talk) 11:39, 15 August 2016 (UTC)

@Pigsonthewing, Jura1, Thryduulf, Lymantria, Izno:  Not done, no consensus (sadly) --Srittau (talk) 19:09, 15 August 2016 (UTC)