User:Paperoastro/sandbox

This table is awesome! I'm glad that folks are starting to discuss how to use the W3C-recommended "subclass of" property to build taxonomies of concepts.

The table above is constructing what's called a "subsumption hierarchy" of astronomical objects. In other words, the item in each row is a subclass of astronomical object, either directly or indirectly. Thus, subclass of (P279) is being used to define what type of astronomical object the item in each row is, which makes "type of astronomical object" (P60) property -- P60 (P60) -- redundant with subclass of (P279). The P279 value seems like it would duplicate the P60 value in every case.

I'd also point out that none of the items classified as "subclass of planetary system" are actually types of (subclasses of) planetary system, at least to my understanding. They can be part of (P361) a planetary system, but they aren't types of planetary system. en:Planetary_system#Types_and_attributes lists "Multiplanet systems", "Main-sequence system", "Red dwarf system" etc. as types of planetary system, for example, not "planet", "moon", "comet" etc. Cheers, Emw (talk) 05:03, 10 July 2013 (UTC)

Thank you very much for your comments, Emw! I have two questions:

1. I cannot understand why P60 (P60) is redundant: for me Jupiter (Q319) <P60 (P60)> planet (Q634). Do you suggest something as Jupiter (Q319) <subclass of (P279)> planet (Q634)? Please, clarifies this point.
2. You have right: items that I considered "subclass of planetary system" are better as part of (P361). I'll make the change, but I have a doubt. Can this create confusion in the hierarchy?

--Paperoastro (talk) 10:18, 10 July 2013 (UTC)

Actually for subclass of planetary system I'm not shocked : the question is can we consider a binary star without the rest of its planetary system ? The question is the binary star the main object, in which case we should have a type binary star with several instances, say <AB star> instance of <binary star>, or is the whole planetary system the same object, in which case the type-item should be binary star solar system, or both, in which case we would have the two item types, <binary star> part of <binary star solar system>, and for instances <AB star> instance of <binary star> ; <AB star> part-of <AB star solar system>; <AB star solar system> instance of <binary star solar system> and <binary star> part of <binary star solar system>. TomT0m (talk) 11:24, 10 July 2013 (UTC)

Paperoastro, regarding question #1: more broadly, P60 (P60) is redundant with instance of (P31) and subclass of (P279). Jupiter isn't a subclass of planet, but rather an instance of (P31) planet (Q634). The difference is that P60 is being used to apply to both classes (e.g. 'comet') and instances (e.g. 'Jupiter'), and P31 and P279 distinguish between the two. Even though one approach distinguishes between classes and instances and another doesn't, they're still entirely redundant. Emw (talk) 12:30, 10 July 2013 (UTC)

@User:TomT0m: en:planetary system is defined as a set of gravitationally bound non-stellar objects in orbit around a star or star system (with two references); en:star system as a small number of stars which orbit each other (with one reference). Furthermore we have articles (and items) concerning only the star systems and others concerning the single planets. For this reason, I decided to put planetary system (Q206717) and star system (Q595871) at the same level of the hierarchy. To connect planets and their star systems, I'm using parent astronomical body (P397) and child astronomical body (P398). Probably we will need something to comprehend stars and their planetary systems, but at now I did not found the right "entity".

@User:Emw: Thanks for the clarification! Using only the property instance of (P31) or more specific "type properties" was discussed various times. In this discussion it is not clear where is the consensus. Besides, constraint violations of P59 and P60 properties work very well, so I prefer to use the specific property P60 instead of P31. --Paperoastro (talk) 21:54, 10 July 2013 (UTC)

"Part of" relates objects but "sub class of" relates classes so<AB star> "part-of" <AB star system> "part of" <CD constellation> but <binary star> "sub class of" <star> "sub class of" <astronomical object> and <binary star system> "sub class of" <star system> "sub class of" <astronomical object>. Note that <binary star> isn't a star; it's a class of stars and it isn't a "sub class of" <binary star system> because a <binary star> is not a type of <binary star system>.

So "part of" creates a hierarchy of objects (with <the universe> at the top), and "sub class of" creates a hierarchy of classes (with a generic class such as <entity> at the top).

Note that while it is true that "type of astronomical object" is a subproperty of "instance of" and could be replaced by "instance of" this would deprive us of certain advantages that "type of astronomical object" gives us compared to a more general property.

• An item may have more than one "instance of" statement. Using the "type of astronomical object" property instead makes it easier for scripts to import the correct statement.
• Using the "type of astronomical object" property means we can have a bot check for 'type violations' since this property should only be used to refer to items which are a sub class of (or a sub class of a sub class of) <astronomical object>.

Does that make sense? Filceolaire (talk) 22:23, 10 July 2013 (UTC)

The first seems irrelevant to me. In what use case does it not serve us to have made both statements?

You can do the same with instance of and subclass of, and I haven't seen any reason why not. Also, as has been argued ad nauseum at the RFC on the primary sorting property, of which this is a subset, we don't need any sort of "type of" property which sits outside of the "instance of"/"subclass of" tree. This particular property is really just a subclass of a sorting property, and thus inherits its flaws. --Izno (talk) 23:11, 10 July 2013 (UTC)

The second is like the first makes it easyier to check whether or not it is an astronomical object, but it is redundant as well : to check if it is an astronomical object, it is engouh to retrieve it's parent classes and check if astronomical object is in them. Or add a little redundancy and add a redundant direct statement subclass of <astronomical object> as it is an important type for the project, as I pointed in one of the discussions pointed by Izno. TomT0m (talk) 08:53, 11 July 2013 (UTC)

My comments:

1. Concerning "type of astronomical object": Filceolaire expressed better than me the advantages to use a specific property. However Emw expressed [1] an interesting feature using instance of (P31) and subclass of (P279). If it is technically possible in Wikidata, both for queries and for constraints, for me became irrelevant the use of "type of astronomical object" or not. For now, I suggest to postpone the discussion of using "type of astronomical object" property when the first queries became available and we will make tests (or we can discuss it in a better page, as noted by Izno and TomT0m).
2. Thanks, Filceolaire, to clarify me the differences between "part of" and "sub-class of". We are lucky: the situation is more simple than you described. For astronomers, <binary star> is not a class of <star>, but of <star system> (see en:binary star). So <AB star> "part-of" <AB star system> with <AB star system> "is" <binary star> and <AB star> "is" <star>. <binary star> and <star> "sub-class of" <star system> that is "sub-class" of <astronomical object> (we can considered <star> as a particular case of <star system> with one star...). Is it sound you correct? --Paperoastro (talk) 13:27, 11 July 2013 (UTC) P.S.:Your observation of a hierarchy using "part of" is very interesting.

Yes. That does sound correct. Filceolaire (talk) 15:05, 11 July 2013 (UTC)

Thank you very much for your comments. My idea is, for now, use this hierarchy as is. When inheritance will be available, I will be happy to delete P60 (P60) for instance of (P31) and subclass of (P279) (if it will be technically possible). Part of this structure could be used to make a hierarchy with part of (P361) (as suggested by Filceolaire). For me there is no problem to overlap the two classifications, to take advantage, in the future, of the inheritance.

I will move the table and this discussion in P60. A last question: exists a tool to view automatically hierarchies? --Paperoastro (talk) 11:26, 4 August 2013 (UTC)

I'm not aware of one. Such a tool would be immensely useful for Wikidata. Protege (http://protege.stanford.edu/) is a standard tool for this, but won't be applicable until at least Wikidata's RDF/XML export feature is available. (And even then, we would need to convert Wikidata's RDF/XML to substitute rdf:type and rdfs:subClassOf for instance of (P31) and subclass of (P279) statements, respectively.) Emw (talk) 15:19, 4 August 2013 (UTC)

Just to let you know that triple star system (Q2088753) also exists, though it has no associated enwiki article. - LaddΩ chat ;) 23:00, 10 July 2013 (UTC)

Thanks to suggest me this item. It can be used for triple star systems instead of <multiple star system> class, but I do not know if it can be useful for Wikipedia infoboxes. I'll have to ask to astronomical projects of some Wikipedias if they are interested on it. --Paperoastro (talk) 13:34, 11 July 2013 (UTC)

Other items that can be added are extrasolar moon (Q679411) and rogue planet (Q167910). I have just one question: excluding rogue planet (Q167910), a planetary system (Q206717) needs a star (Q523) to exist. Why planetary system (Q206717) is not subclass of (P279) of star system (Q595871)? In my opinion these item should be in a higher rank than the previous one and not in the same. Thanks! - Sarilho1 (talk) 23:41, 4 August 2013 (UTC)

Good question! In principle you are right, but the classification is complicated by brown dwarf systems! (see Planetary system). To consider planetary system (Q206717) as subclass of star system (Q595871), we should distinguish planetary systems around stars and around brown dwarfs. --Paperoastro (talk) 12:31, 7 August 2013 (UTC) P.S.: I inserted extrasolar moon (Q679411), thanks. Have you suggestions for rogue planet (Q167910)?

I'm not sure. In the current system, it could be consider subclass of (P279) astronomical object (Q6999). In my propose it would be subclass of (P279) galaxy (Q318). (I forgot referring that star system (Q595871), star cluster (Q168845)... subclass of (P279) galaxy (Q318). galaxy (Q318) subclass of (P279) galaxy cluster (Q204107)... Essentially, I'm proposing a classification based not in what we see from Earth, but in the way the Universe is organized. The problem is that this arrangement is almost the same as the property parent astronomical body (P397). The different with both properties is that one designs terms and the others designs celestial bodies, so I don't know if would be ambiguous. Another problem is that all celestial bodies would be subclass of (P279) Universe (Q1), so, or Universe (Q1) subclass of (P279) astronomical object (Q6999) or I don't know how to include astronomical object (Q6999) in the hierarchic tree.) If you don't understand my proposal I may try to create a table like yours with it. Thanks! - Sarilho1 (talk) 12:49, 7 August 2013 (UTC)

I'm sorry to not answer you here. Your proposal is very interesting and I answered there my comments. --Paperoastro (talk) 12:06, 20 August 2013 (UTC)

I'm seeing a rather large problem after looking at some definitions. The definition of "planetary system" according to en.Wikipedia is "[the] set of gravitationally bound non-stellar objects in orbit around a star or star system". Planets, particularly in the exo-planet case for example, are not necessarily planetary systems then. In that case, it's wrong to define a "planet" as anything but an astronomical object (directly), rather than as a subclass of "planetary system". This is true of several of the items above. This is in fact what the en.Wikipedia article does: "an astronomical object orbiting a star or stellar remnant that is massive enough to be rounded by its own gravity, is not massive enough to cause thermonuclear fusion, and has cleared its neighbouring region of planetesimals". --Izno (talk) 02:10, 20 August 2013 (UTC)

I hope to understand correctly your comment. Probably the problem is solved considering exoplanet (Q44559) as planets around stars (or brown dwarf), and add in this hierarchy the item rogue planet (Q167910) for the planets that do not "belong" to any star. It could be a subclass of "astronomical object" directly or subclass of substellar object (Q3132741) as suggested by Sarilho1. --Paperoastro (talk) 12:21, 20 August 2013 (UTC)

I found myself creating items for star classification, like G-type star (Q14745295) and the other letter-class items. They are subclasses of star and superclasses of other kind of stars. TomT0m (talk) 13:03, 29 August 2013 (UTC)

Good! :-) Pay attention that in Wikidata already exist at least these items concerning star classification for main sequence stars: O-type main sequence star (Q911809), B-type main sequence star (Q767432), A-type main sequence star (Q471805), F-type main-sequence star (Q1353952), G-type main-sequence star (Q5864),K-type main-sequence star (Q863936), red dwarf (Q5893). --Paperoastro (talk) 14:10, 29 August 2013 (UTC)

A curiosity: how would you want to use these items? --Paperoastro (talk) 14:21, 29 August 2013 (UTC)

I don't know the domain really well, but is not main sequence [see https://en.wikipedia.org/wiki/Main_sequence] an other axis to classify stars ? For example I have set F-type main-sequence star (Q1353952) as a subclass of F-type star. I mean I though there exists A class star out of main sequence, that's why I created those items. I had no particular thing in mind, I just fill some label & descriptions, sometimes and I did this just because I can :). I found a F-type main-sequence star and wondered what it was is all. TomT0m (talk) 15:26, 29 August 2013 (UTC)

I gave a look to your classification: you made a great work. F-type main-sequence star (Q1353952) as subclass of F-type is correct: there is very hot stars in the en:Horizontal branch with the same spectral type (stars of type III). We can distinguish different evolution phases in the classification using the suffix I, II, III, IV, V. --Paperoastro (talk) 20:43, 30 August 2013 (UTC)

1. ↑ for Wikipedia infoboxes it is not necessary to distinguish stars and compact stars at this level of classification