Revision as of 17:26, 14 September 2018 by ArthurPSmith(talk | contribs)(new label looks good; discussion seems to have settled down, marking as ready)
In Wikidata, subclass of (P279) is widely used. Some statements in the child (sub)class are exactly the same as those in the parent (super)class. However, we have no way of explicitly representing that the statement in the child class is inherited from its parent class. Using this property clarifies the relationship between a child and parent classes, and enhances robustness of the data structures. --Okkn (talk) 15:12, 30 July 2018 (UTC)[reply]
As far as I understand our datamodel, the subclass relationship means that there's an implicit inheritance. I don't see the point in this explicit statement. ChristianKl ❪✉❫ 16:09, 23 August 2018 (UTC)[reply]
@ChristianKl: Thank you for your comment. I think that a subclass does not always simply inherit the value in the superclass, and often updates it. This qualifier will be useful especially when the subclass multiply inherit more than one superclasses. For example, stomach cancer (Q189588) is a subclass of both gastrointestinal system cancer (Q5526839) and stomach disease (Q175827), but the anatomical location (P927) value stomach (Q1029907) is only inherited from stomach disease (Q175827). Persons or computers who don't know well about the superclasses cannot understand where the value "stomach" comes from, without additional queries. Explicit statements clarify how the subclass and its superclass are similar or are different, and how the superclasses of a class are different. In addition, even if the value in the superclass mistakenly changed or removed, we can detect that. Regards, --Okkn (talk) 17:27, 23 August 2018 (UTC)[reply]
Comment My concern is that every item in every domain could potentially have dozens of these statements, inheriting or restricting the properties of all of its parent classes. Do we really want to have statements on every type of tree that it <has parts of the class> "trunk", "crown", "leaf", etc.? - PKM (talk) 19:48, 23 August 2018 (UTC)[reply]
@PKM: These qualifiers are similar to reference. Not all statements should have these qualifiers, but if we use them, the relations will become robust. In addition, indeed, statements we can apply these qualifiers are limited, because most of the items in Wikidata are instances (not class) and most of the statements are identifiers or peculiar values (which are not inherited by subclasses). As for the tree example, it is not needed that each types of trees have such <has parts of the class> statements in the first place, is it? --Okkn (talk) 21:00, 23 August 2018 (UTC)[reply]
@Pasleim: This kind of data may not be a source of information, but I don't mind if we make a rule that these properties should be used as source properties, instead of qualifiers. --Okkn (talk) 07:54, 12 September 2018 (UTC)[reply]
It's not a source in the classical sense but the property would indicate from where a value is taken, so I would use it in source section. Basically, the proposed property is a special case of inferred from (P3452) --Pasleim (talk) 10:01, 12 September 2018 (UTC)[reply]
I agree with using it as a reference rather than a qualifier. For example, if you merge two items where one has a statement with this property and the other has the same statement without this property, the statements should be merged. That will happen with references but not with qualifiers. - Nikki (talk) 09:17, 13 September 2018 (UTC)[reply]
Comment If this is added, I think some people are likely to confuse it with inferred from (P3452). I wonder if calling it "item inherits value from" would be better in English? (i.e. phrasing it more actively, to emphasise that it continues inheriting the value) - Nikki (talk) 09:17, 13 September 2018 (UTC)[reply]
Notified participants of WikiProject Medicine: This qualifier is especially useful to describe medical concepts. --Okkn (talk) 15:27, 20 August 2018 (UTC)
Support. --Csisc (talk) 19:30, 20 August 2018 (UTC)
Comment My concern is that every item in every domain could potentially have dozens of these statements, inheriting or restricting the properties of all of its parent classes. Do we really want to have statements on every type of tree that it <has parts of the class> "trunk", "crown", "leaf", etc.? - PKM (talk) 19:48, 23 August 2018 (UTC)
Question Shouldn't that be a source property rather than a qualifier? --Pasleim (talk) 16:18, 11 September 2018 (UTC)
