Wikidata talk:Requests for comment/Help:Basic membership properties

metaclass, has part and quarks

I recommend the following changes:

• Add class - class examples for "instance of" to cover metaclasses. This is a real issue we should have examples for.
• add a "has part" section - and indicate when it should be used (does "has part" necessarily imply the inverse "part of" or only if sometimes?)
• Replace all the "quark" examples with something that actually makes sense - for example "Cecil (Lion)" instance of "lion", "lion" subclass of "Panthera", "lion" instance of "taxon", "griffin" has part "lion"

ArthurPSmith (talk) 17:59, 27 August 2015 (UTC)

@ArthurPSmith: Biology is a sensible issue as Taxonomis prefers "parent taxon", even if we recently could add also subclass of (P279) statements between two taxons which is good. The issue metaclass issue is more discussed on Help:Classification who is more detailed and discussed as an help page on Adopt Help:Classification as an official help page, comments welcome. author  TomT0m / talk page 18:30, 27 August 2015 (UTC)

@TomT0m: well 'parent taxon' is now a sub property of "subclass of" so that relationship is entailed, I don't think it's a problem. If you prefer you could go with something from a completely different domain though, I just think the quark example is unnecessarily confusing because one almost never talks about individual instances of quarks or any elementary particle, in large part all we know about them is from statistics of experiments not individual events - and individual quarks have fractional charge that is in principle unobservable in isolation, further complicating things! Is "top quark" an instance of a quark or a subclass of quarks? Just messy to get into those details - bring in something more common to real life and for which there are concrete examples already in wikidata that illustrate the point. ArthurPSmith (talk) 18:54, 27 August 2015 (UTC)

By the way I've read the Help:Classification (quite good, but it doesn't seem really a help page) and the talk page on Help:Basic membership properties and there definitely seems to be confusion about whether classes can be "instances" of something else. I think putting specific examples of classes that are instances of other classes (metaclasses but we don't have to mention that term on the basic page) on Help:Basic membership properties would help a lot. "instance" has meaning relative to the relationship an entity has with another entity, it says nothing about the characteristics of the entity itself. ArthurPSmith (talk) 18:59, 27 August 2015 (UTC)

It's a subclass of quark as long as we refers to top quark in general an not of a specific top quark. I don't mind changing the example for something understandable by people who don't know anything about particle physics anyway, and there is plenty :)

maybe an example of metaclass would be useful indeed. The "car model" one seems appropriate and common. author  TomT0m / talk page 19:08, 27 August 2015 (UTC)

@TomT0m: I just read the discussion about parent taxon on the chat page, I'd missed that earlier. Ok, it's contentious, probably best to avoid. If you want something science-related for this, how about: "Sun" instance of "G-class star", "G-class star" instance of "stellar classification", "G-class star" subclass of "star", "Sun" "part of" "Solar System" ? I think that's obvious and clear enough an example. ArthurPSmith (talk) 19:32, 27 August 2015 (UTC)

I don't want anything particular, I did not wrote this anyway. I don't like the "instance of classification" style statements as I don't think a metaclass is a classification by itself. The metaclass of all classes in a classification is, but it's not to be confused with the whole classification imho. Labelled "spectral class of star" however this should work, with a note that the criteria used to define that class of star is the kind of light they shine with. author  TomT0m / talk page 20:02, 27 August 2015 (UTC)

@TomT0m: I've made a modified draft of the page available here: User:ArthurPSmith/Test/Help:Basic membership properties that addresses most of my concerns, but I'm not quite finished with it... comments welcome, I am definitely still updating this... ArthurPSmith (talk) 19:29, 28 August 2015 (UTC)

More on has part and part of

I find the discussion of "part of" confusing and I think it would help to indicate what is wanted with regard to the inverse relationship "has part". That is, "griffin" has part "lion" because the mythical creatures supposedly are part-lion. But would you also want to have the claim that lion is "part of" griffin? I don't think that makes sense. Not all instances of lions are parts of instances of griffins, obviously. Similarly with a lot of "part of" relationships I've seen mentioned for classes - is every instance of the subject class part of an instance of the target class, is that what "part of" should mean? I think this help page needs to clarify that somehow, and examples where the inverse is good and when it's not would be useful there I think (if that is indeed the intent). ArthurPSmith (talk) 19:41, 27 August 2015 (UTC)

We try not to mix fictional entities like griffins with real world entities - so the Griffin example should not occur in wikidata (as is) : fictional entities have their own class hierarchy not to be mixes with real one. Fictional entities that maps to real one are connected through the "fictional analog of" property.

"is every instance of the subject class part of an instance of the target class, is that what "part of" should mean" I don't think so this is really practical, a wheel can definitely have a life outside of a car, although that's usually where they are supposed to be :) I don't think the relationship is that strong, we can leave with "usually an instance of the subject is a part of the object" and "use common sense" (avoiding a confusion beeteen instance of and part of is already a good step for community, I proposed a property for this

   Under discussion

Description	MISSING
Data type	MISSING
Example 1	MISSING
Example 2	MISSING
Example 3	MISSING

who needs support :) . There is a lot to say about part of, and a lot that have been written. I'm not sure we totally closed the subject and the last discussion about that I can remember, Refining "part of" was kind of unconclusive. MKrotzch had something I think but it does not seem the rest of the participants did actually conclude :/ We should propose its subproperties. author  TomT0m / talk page 20:19, 27 August 2015 (UTC)

So these last few days I've been doing a lot of reading of discussions from the past 2 years on classification, part of, etc - a lot of background that's somewhat confusing for me as somebody new to it all! Anyway, it did look to me like the Refining "part of" discussion ended up basically recommending leaving "part of" as it is but adding subproperties for more specific cases. Where I think there's a need for distinct subproperties is in cases where there are several different kinds of "part of" relationships to a given item or class so it would be helpful to group those distinct parts by type. One example on the lines I was looking at earlier was for example the "parts" of a university, for example in the template en:Template:University_of_Illinois_at_Urbana–Champaign_campus you have "academic" parts, "athletic" parts, "buildings" that are part of the main campus, other campuses, etc. and it makes sense to group the different kinds of "parts" together. Simply defining suitable subproperties of "part of" to encapsulate this seems like a logical approach. They are not exactly at the abstract level that MKrotzch mentioned but I think there are a practical manifestation of the sort of distinctions being discussed in that RfC.

ArthurPSmith: A Griffin has the body of a lion and the head and wings of an eagle. The body is a part of the Griffin but the body of a Griffin is not a lion; it is the torso of a lion. I'm not sure you can use 'part of' or 'has part' to describe a Griffin. If A is 'part of' B then all of A is in B. That doesn't describe the relationship between a lion and a Griffin. Start again and rewrite the description of the Griffin - it has a body like a lion and a head and wings like an eagle.

Or maybe describe it like this: Griffin <has part:head>(of:eagle). <has part:body>(of:lion)

or alternatively Griffin <similar to:lion>(applies to part:torso). <similar to:eagle> (applies to part:head). I think that is my favourite. Joe Filceolaire (talk) 02:37, 2 September 2015 (UTC)

I like the similar to property, although we have to check its relationship to depicts or fictional analog of (which seems really close if not a duplicate). Although I don't like applies to part but it's another story. I would prefer a direct reusable item Lion's head (<part of> : <Lion>, <subclass of> : <Head>) used without qualifier with similar to

I really don't want to create items for Lion's head, Lion's torso, Lions tail etc. etc. etc. Besides the Griffin doesn't have a Lions body. It has a Griffins body which happens to look like a Lions body.Just as the first descriptions of the unicorn described it as having the feet of an elephant and the tail of a lion. This is because it was actually a description of a Rhino which does have feet like an elephant and a tail like a lion but I don't see anyone proposing to add statements to that effect to rhinoceros (Q34718). Joe Filceolaire (talk) 10:17, 2 September 2015 (UTC)

Of course, it was intended to be used in similar to statements. I don't really understand why creating an item is such a big deal to you. author  TomT0m / talk page 10:23, 2 September 2015 (UTC)

Joe Filceolaire: I don't particularly care about the lion/griffin example, it only came up because that was an actual instance I ran across in the current iteration of wikidata. The main point I was trying to get at was the question of whether "part of" and "has part" should be considered exact inverses (X "part of" Y if and only if Y "has part" X) or only inverses in the sense of meaning. Does "house" "has part" "door" imply "door" "part of" "house? In the case of non-classes (individual instances) the exact inverse should naturally hold because that specific X is indeed part of that specific Y, which has that part, etc. But for the case where X and Y are classes, I would interpret the relations to mean that it holds for (all or most) instances of the subject class. All or most houses have doors. But the majority of doors are likely not part of houses. So "has part" makes sense, "part of" perhaps not. Is that clearer? ArthurPSmith (talk) 15:40, 2 September 2015 (UTC)

Arthur:I think you need to be careful about using "has part" and "part of" with classes rather than instances. "My house" <has part:door> but generic "House" is a class of lots of different houses so it shouldn't have a <has part:door> claim because that applies to instances of house. Where you have a class which is bunch of identical mass produced instances then you may be able to use 'part of' or 'has part' but those classes are usually an <instance of:model> or <instance of:design>. For example:

"iPhone 5" <has part:Sony Exmor R> where "Sony Exmor R" <instance of:digital camera>

"Sony Exmor R" <part of:iPhone 5> is probably appropriate too.

Does that make sense? Joe Filceolaire (talk) 01:43, 3 September 2015 (UTC)

@Filceolaire: models can be considered metaclasses per Help:Classification. And arthur is right, part of can be used with classes, for example human head is a part of human body as well as my own head is a part of my own body. This works if either both subject or objects of the part of claims are classes, or are pure instances. We should however never make statements like my own head part of human body as it mixes the class level with the pure instance level. author  TomT0m / talk page 14:57, 3 September 2015 (UTC)

TomT0m, Arthur: What if every member of a class is part of an instance - as all of the instances of the class of 'college of the University of Cambridge (Q1055028)' are <part of':'Cambridge University'>? This seems similar to the "instance <part of:instance>" relation but different from the "class <part of:class>" relation.

"Body <has part:head>" is true but I think "head <part of:body>" is only mostly true. I think we may need another property I will call "typically part of" for the class <typically part of:class> relation. Joe Filceolaire (talk) 18:08, 3 September 2015 (UTC)

A body without head can also exists, so I'm not sure it's a good example ...

The class should not be a part of the University, this violates the rule. Each of its instances are although ... But you are right this might be a problem for the rule of inferences I proposed in WikiProject Reasoning/Use cases, I'll mention it. author  TomT0m / talk page 18:19, 3 September 2015 (UTC)

I gave another example like the university colleges one in my edited version of Help:BMP at User:ArthurPSmith/Test/Help:Basic membership properties - <inner planets> are <part of> the <Solar System> even though the subject is a class and the value of the property is not. ArthurPSmith (talk) 18:23, 3 September 2015 (UTC)

Revised page

@TomT0m: @Filceolaire: I've finished what I wanted to do with User:ArthurPSmith/Test/Help:Basic membership properties - all the examples I'd like to see are in there now (thanks for suggestions in above comments and elsewhere). It should probably wait until the new Glossary page is done so the definitions at the top can just refer to that. Also if it were to replace the actual page it needs some translation work (I'm not sure how that is done in practice though I can guess a bit from what I've seen on that page so far). There are two things that also probably should be discussed a bit on there before it goes anywhere else: distinguishing between "non-class" and "instance" (which I thought was confused on the old help page), and the "DISALLOWED" lines. Are these changes correct, are they helpful?