Emw

You may want to take a look at Help:Label and Help:Description. Descriptions do not need to be capitalized, and should not begin with an article, and labels should not include extra disambiguating text in brackets. Also, please don't add aliases that do not actually mean the same thing as the item's name. --Yair rand (talk) 22:54, 30 December 2012 (UTC)Reply

Also, Wikipedia copyrights are incompatible with Wikidata's, so Wikipedia article text shouldn't be copied verbatim as descriptions. Thanks. --Yair rand (talk) 23:01, 30 December 2012 (UTC)Reply

Hi there. Just so you know, labels like this one that are copies of Wikipedia articles' first sentences violate the terms of Wikipedia's licensing arrangement, both because you have not provided attribution, and because, since Wikidata mainspace contributions are licensed in the public domain, you are in effect releasing something into the public domain for which you do not actually hold the copyright. I see that Yair rand (talk • contribs • logs) has brought this up already, but I imagine that the different licensing arrangement over here might be a bit confusing, and I know that many long-term Wikipedia contributors aren't particularly aware of en:WP:COPYWITHIN; I hope this has clarified the matter somewhat. — Francophonie&Androphilie(^{Je vous invite à me parler}) 08:05, 4 January 2013 (UTC)Reply

Hi Francophonie&Androphilie. The content in question was not a copy of the first sentence; it was a copy of a fragment of the first sentence, and had different capitalization. Do formulaic 9-word sentence fragments exceed the threshold of originality needed to be copyrightable? Emw (talk) 18:18, 5 January 2013 (UTC)Reply

I agree that there are some cases where the Wikipedia lede is the legitimate most sensible label. But take the label you gave at Moxie - it's all but the first three words of the lede sentence of en:Moxie, and while yes, it's rather unlikely that whoever wrote that sentence would try to claim it as their intellectual property, I'd say it is original enough that it's inappropriate for you to represent it as your own (which you effectively do by agreeing to the CC0 terms), especially when it's not really an appropriate label anyways (since all you really need to say is something like "American carbonated soft drink". — Francophonie&Androphilie(^{Je vous invite à me parler}) 16:15, 6 January 2013 (UTC)Reply

Where is the threshold of originality for this type of short description? If a quotidian 16-word sentence fragment is original enough to be copyrightable, then how about a similarly common 9-word sentence fragment? Is the 12-word description for Universe on Wikidata, which was copied from English Wikipedia then trivially pared down (diff), a violation of copyright? I don't think such copying implies that one is representing the content as their own intellectual property, but rather that they're asserting such content is everyone's property -- i.e., that it is not creative enough to be copyrightable. The article Copyright Protection for Short Phrases from the Stanford Copyright and Fair Use website provides some insight on this matter. Emw (talk) 18:14, 6 January 2013 (UTC)Reply

You switched on Q251 (Java) from Property:P31 (instance of) to Property:P279 (subclass of) Q9143 (programming language). I think this is wrong. --Fomafix (talk) 07:40, 21 March 2013 (UTC)Reply

Why do you think it's wrong? Java the programming language is a concept and has no particular location in space and time, and other foreseeable items like versions of the language could be said to be members of it. Those features suggest that Java is a class and not an instance. Also, the Software Ontology (SWO) project defines Java as a subclass of programming language. Emw (talk) 11:58, 21 March 2013 (UTC)Reply

I think something like Q211496 (high-level programming language) is a subclass of Q9143 (programming language). When you think that Java is not a instance of a programming language, what else are the instances of the class programming language? --Fomafix (talk) 08:05, 25 March 2013 (UTC)Reply

Great question.

Instances are things with a unique location in space and time. An instance of a programming language would be a copy of http://www.python.org/ftp/python/3.3.0/Python-3.3.0.tgz on a specific computer. Clearly, Wikidata is not concerned with particular instances of a programming language, it's interested in the class of thing of which those instances are a type.

In many cases, though, Wikidata actually does concern itself with instances. For example the project has a ton of items about instances of the class 'person' and 'place'. I find the overview at User:Stevenliuyi/Differences among basic membership properties helpful in summing up the difference between instance and class. The example there of "quark" is analogous to this situation with specific types of software. Types of things that won't foreseeably have Wikidata items about their instances -- like software, quarks, etc. -- often trigger a double-take when thinking about we should say they are classes or instances. Emw (talk) 11:39, 30 March 2013 (UTC)Reply

Hi Emw. Thanks to report me the end of the discussion. I moved the property proposal to the "pending" page. Best, --Paperoastro (talk) 16:13, 24 March 2013 (UTC)Reply

Yes, I'd like to add it to Wikidata help page, though I think we should discuss and improve it first, so we can make sure everything is clear. And of course, we should avoid the jargon "ontology" when we make it public. --Stevenliuyi (talk) 15:19, 28 March 2013 (UTC)Reply

Hello Emw. I saw that you argued several times with the statement above. Please consider that even thought you're right, there are reasons to subdivide properties in more specific properties. E.g. "Album type". How would we work with infoboxes if we lump all values in the same property? In this case we can't pick out the specific instance value for the album type and use it for the infobox. I hope you see what I mean. If so, it would be nice if you would withdraw your oppose vote for 'album type' as I was looking to create this property. Thank you. --Nightwish62 (talk) 10:17, 29 March 2013 (UTC)Reply

Hi Nightwish. The reason I make that argument is because it promotes consistency and interoperability in vocabulary, which are important for Wikidata properties. Properties that define an item's type are special kinds of properties. The W3C -- the international standards organization for the Web -- recommends two properties for defining an item's type in the Semantic Web. The first is rdf:type, which corresponds to instance of (P31). The second is rdfs:subClassOf, which corresponds to subclass of (P279). Because these properties match W3C standards, they will enable Wikidata to be more interoperable with other projects in the Semantic Web.

The 'instance of' and 'subclass of' properties map to the name of an infobox, or its 'type' parameters. Having 'type' parameters in infoboxes doesn't really make sense -- these parameters and the corresponding properties that have been proposed are describing what class an item is a member of, which is a special kind of property that should be the name of the infobox and not one of its parameters. Many infoboxes on Wikipedia are problematic in that sense, but the problem is solvable.

To answer your question "how would we work with infoboxes if we lump all values in the same property?", I think a workable approach would be to follow the recommendations of the W3C and conventions in the knowledge representation community and define all class information with the rdf:type and rdfs:subClassOf properties, i.e. 'instance of ' (P31) and 'subclass of' (P279). Best regards, Emw (talk) 00:33, 30 March 2013 (UTC)Reply

Thanks for your extensive explanation. To be honest I didn't understand all I read about 'semantic web'. Half because this stuff is new for me, half because English isn't my native language. I've an average understanding of databases and little about database normalization. However, you're right (I found this quote of you somewhere else) "Wikidata is not merely a database, it is a project to structure all knowledge". Nonetheless in my opinion an own property 'album type' makes sense. I still doesn't see how to retrieve this information otherwise from the 'instance of' property, if there are other values (e.g. is a 'media', 'compact disc', 'debut album'). I think the situation is similar to the question how much normalization of a database is desirable. There are situations to deliberately NOT to make normalizations, e.g. for better readability/administration. The same for the case we're talking about. Even you're right and there are standards about structuring data from the W3C, we need to find a solution which fits for our purposes the best. And I think most of the editor/users/data collectors/template creator wishes for an own property. Perhaps we should start a poll about this question (not only about the 'album type', but in general), but I still hope you agree with me.

Beside this: I don't see any difference between 'album type' and properties like 'gender' or 'country', 'GND type' etc. All this could also be summarized under 'instance of'. --Nightwish62 (talk) 16:10, 30 March 2013 (UTC)Reply

We seem to have been taken up this discussion at Wikidata:Requests_for_deletions#.22Type_of.22_properties. If my replies there don't answer your questions above, then please let me know here or there. Emw (talk) 04:43, 1 April 2013 (UTC)Reply

Protonation subclass of protonation? Sure about that? :) --Izno (talk) 01:50, 29 April 2013 (UTC)Reply

Ah, thanks for catching the typo. Fixed. Emw (talk) 01:52, 29 April 2013 (UTC)Reply

Hello, I agree with most of your explanations here. But what whould you think about this schema (types have a simple frame, tokens have a double frame):

      +------------+                            +---------+
      | individual |------- if part of -------->| species |
      +------------+                            +---------+
            Λ                                         Λ
            |                                         |
       subclass of                               instance of
            |                                         |
     +--------------+                         ++==============++
     | homo sapiens |------ is part of ------>|| Homo Sapiens ||
     | (individual) |                         ||   (species)  ||
     +--------------+                         ++==============++
           Λ
           |
      instance of
           |
++===================++
|| François Hollande ||
++===================++

I think that we should distinguish homo spiens and Homo Sapiens: the last one could have some attributes like date of appearance, population, ... --Gloumouth1 (talk) 12:06, 13 May 2013 (UTC)Reply

Hi Gloumouth, thanks for the note. The schema above seems somewhat problematic to me. The 'instance of' property automatically denotes its subject (here, François Hollande) as an instance and its object (Homo sapiens) as a class, so having two separate items for the object -- one for the item as an instance and one for the item as a class, seems extraneous. Also, distinguishing between a subject as an instance and a subject as a class through capitalization (homo sapiens vs Homo Sapiens) strikes me as kludgy.

An alternative schema to consider:

• François Hollande

instance of: Homo sapiens (human)

• Homo sapiens

subclass of: Homo

taxonomic rank: species

same as: natural person

• species

subclass of: taxonomic rank

• natural person

subclass of: person

That said, I think it's worth exploring how "class properties" can be distinguished from "instance properties". Zolo has brought this up at Help:Basic membership properties. Continuing your example, we know that 'Homo sapiens date of appearance 198,000 BCE' would apply to Homo sapiens as a class, but in some sense only directly applies to one instance of the class -- i.e. the first anatomically modern human, not François Hollande. On the other hand, we also know that 'Homo sapiens has part hand' would apply to all humans (with rare exceptions) -- i.e. the first human as well as François Hollande. How can a machine reasoner distinguish between those two kinds of properties? In other words, how can a machine know that François Hollande has two hands, but did not appear in 198,000 BCE? Emw (talk) 03:43, 16 May 2013 (UTC)Reply

Let's forget the captialization to distinguish them, it is a naming detail. I understand you agree that having 2 different entities is necessary if we want to allow a machine reasoner to get the information. I think that my schema allows this. You asked the good question: which entity has to carry a given attribute? --Gloumouth1 (talk) 10:47, 16 May 2013 (UTC)Reply

I actually disagree that having two entities is necessary: I would not support having two 'Homo sapiens' items in which the only difference was that one described Homo sapiens as a class and the other described it as an instance. I think it would be better to specify the semantics of the properties -- as in the 'has part hand' and 'date of appearance' examples above -- through a qualifier on the property. Emw (talk) 12:13, 21 May 2013 (UTC)Reply

Ok,... So if I understand correctly, the Homo Sapiens entity would carry two kinds of attributes : "inheritable for instances" attributes (e.g. has part hand), and "uninheritable for instances" attributes (e.g. date of appearance). The "inheritability" would be specified thanks to a qualifier associated to each attributes. It is not very "clean", as the Homo Sapiens entity would be "hybrid", i.e. partially a type and partially a token, but it should be sufficient for the machine reasoners, and maybe it would be less counterintuitive for the users. --Gloumouth1 (talk) 13:02, 21 May 2013 (UTC)Reply

Dear Emw, on the 27th of March you responded to my proposal for the property "was a(n) / war ein(e) / était un(e)" (see here). The proposal didn't get sufficient support and the conclusion was to wait for qualifiers. Now that those have been implemented, I have proposed the creation of a qualifier "former". Would you please be so kind to have a look at this proposal on this page. It's header is "former / ehemalige / ancien(ne) / voormalige". Thanks in advance, NormanB (talk) 22:49, 13 May 2013 (UTC)Reply

Hi. I was away from Wikidata for a while so I didn't catch your post with the above title in Project chat in time. However, I think it's an interesting issue so I'll write my opinion here.

There is a fundamental difference between the semantics of the "domain" and "allowed values" fields in the property documentation template and rdfs:domain and rdfs:range. Unless I'm mistaken, the fields in the template are meant to serve as a restriction of allowed values, though at this point unenforceable and thus more-or-less merely informative. rdfs:domain and rdfs:range declarations, however, have the effect of implicit instance/subclass statements. I'm not sure whether your suggestion had the intention to change the semantics of the fields. Personally, I think property domain and range restrictions could be useful, although I haven't given it much thought. However, I'm not sure I see the usefulness of rdfs:domain and rdfs:range--one can always make explicit instance/subclass statements, and implicit ones only create a great potential for confusion and bugs.

Silver hr (talk) 20:02, 24 May 2013 (UTC)Reply

Can you give an example of when an item's class implied by rdfs:domain or rdfs:range would not match the class (or derived class) implied by rdf:type (P31) or rdfs:subClassOf (P279)? My initial impression is that that would represent an inconsistency, an error that would need to be fixed. I think the usefulness of basing "domain" off rdfs:domain and "allowed values" off rdfs:range is more about helping to rigorize the definitions of those special properties and making them interoperable with the rest of the Semantic Web than providing a way to make indirect statements about an item's class. Emw (talk) 12:26, 29 May 2013 (UTC)Reply

For example, the current docs for member of (P463) list the domain as "person". If that were an rdfs:domain, and someone made the statement "Germany member of EU", that would amount to stating "Germany instance of person", but this would be hidden from view (at least in the current system). Also, it would not be an inconsistency in the logical sense unless there were an explicit (or inferred, but I don't think we're there yet) declaration that "country" (or whatever Germany is stated as an instance of) and "person" are disjoint classes, because, although I haven't seen anything in the official Wikidata docs, I assume the Open world assumption applies to Wikidata like it applies to the Semantic Web in general. So as a result, some external reasoner would include Germany in the list of persons. Of course, you could say that the domain of the property should be expanded to countries and other administrative regions, organizations etc. as well as persons, and that would be a valid point. However, we would have no way of doing that with rdfs alone--multiple rdfs:domain statements amount to an intersection, not a union. To state the intended meaning of union, we would ideally need support for a formal class definition language, e.g. OWL, or at the very least the ability to formally specify class unions. Another example is sex or gender (P21). Someone could use it on an animal, and for it to come up as an error upon automatic (non-human) analysis, we would first have to have the relevant class disjointness statements in place, and second we would have to have an actual reasoner performing reasoning on the whole ontology--alternatively, Wikidata itself could catch it if had reasoning capabilities, but we're not there yet, and it's uncertain whether we will be.

Don't get me wrong, I'm generally in favor of interoperability with the Semantic Web, but in this particular case I'm not sure that alone outweighs the downsides. To use rdfs:domain and rdfs:range to detect inconsistencies and ontology errors we'd either have to explicitly reject the OWA entirely (which would not be compatible with the Semantic Web), explicitly state for each class all the classes it's disjoint with (which could be a huge number of statements, O(n²)), or perhaps implicitly consider classes disjoint unless they're declared non-disjoint. I mean, it could work and I'm not outright opposed to it, but if what we want is the restriction of values, it seems easier to me to just directly interpret the "domain" and "allowed values" fields as restrictions.

Silver hr (talk) 18:55, 30 May 2013 (UTC)Reply

The property P596 (P596) that you supported is available now. --Tobias1984 (talk) 08:05, 3 June 2013 (UTC)Reply

Hello, you signed up as a participant for the Molecular biology task force. As promised, we are keeping you up to date on important discussions as they arise. Currently, we are very interested in your thoughts on how orthologs are handled in Wikidata. Please chime in here if you have thoughts. There is also another discussion on whether genes and proteins should be combined in a single wikidata item or kept separate. (Thank you, Emw, for initiating this one...) Both these issues are critical for how we model genetic data, and it would be much better if we established solid community consensus now. Your input is appreciated! Cheers, Andrew Su (talk) 00:16, 23 June 2013 (UTC)Reply

I've replied to your reply here. Also, don't worry about it being delayed--mine mostly are too since I can't dedicate as much time to Wikidata as I'd like. Silver hr (talk) 21:11, 24 June 2013 (UTC)Reply

As you have requested, I have undeleted reelin (Q13561329) because it met notability criterion 2. Sorry for the mix up. Have a great day! TCN7JM 13:54, 29 June 2013 (UTC)Reply

No worries, thank you! Emw (talk) 14:01, 29 June 2013 (UTC)Reply

You recently participated in a deletion discussion for P107 - main type (GND). The discussion has been closed, as it is clear that a resolution won't come from PfD, and an RfC has been opened on the matter at Wikidata:Requests for comment/Primary sorting property. You are invited to participate there. Please note that this is a mass delivered message, and that I will not see any replies you leave on this page.

Yours, Sven Manguard Wha? 18:24, 30 June 2013 (UTC)Reply

Hi Emw (and pointing Boghog here too), Thanks for all your help with the data modeling around genes and proteins. Just wanted to let you know that I'm bunkered in grant writing mode for the next week, so my participation is going to be pretty limited. But I think we see eye-to-eye on most issues, so don't feel compelled to wait around for my input if you want to forge ahead on anything... I'm excited about how great this is going to be! Cheers, Andrew Su (talk) 20:23, 30 June 2013 (UTC)Reply

Sounds good! Exciting times indeed. Regards, Emw (talk) 20:30, 30 June 2013 (UTC)Reply

The property genomic assembly (P659) that you supported is available now. --Tobias1984 (talk) 12:20, 2 July 2013 (UTC)Reply

The property EC enzyme classification (P660) that you proposed is available now. --Tobias1984 (talk) 12:29, 2 July 2013 (UTC)Reply

What would you classify a resource as? The en.wiki definition is a "supply" or "source"... --Izno (talk) 16:40, 4 July 2013 (UTC)Reply

Hmm, I don't know. The classification of resource is a bit interesting because "Resource" is the root of the internal type hierarchy used by RDFS and OWL (see http://www.w3.org/TR/rdf-schema/#ch_resource). Moreso, classifying it is interesting because at least the English Wikipedia page (en:Resource) is very similar to a disambiguation page (en:Resource_(disambiguation)), but just has more prose for each disambiguation pointer. It would probably make sense to examine how external upper ontologies classify "resource". Emw (talk) 11:23, 6 July 2013 (UTC)Reply

Speaking of disambiguation, I actually needed to split all the pages tagged as actual disambiguation pages to resource; the majority of links looking like either one or the other of the en.wiki pages. As for en.wiki's choice of having two pages for it, that's simply en:WP:SUMMARY for you. >_>

I'm not sure it's valuable to associate our notion of a resource ("supply" or "source") with the RDF schema notion of a resource (though, I was not aware that is what the R in RDF means!). --Izno (talk) 15:14, 6 July 2013 (UTC)Reply

The property molecular function (P680) that you commented on is available now. --Tobias1984 (talk) 08:49, 9 July 2013 (UTC)Reply

The property cell component (P681) that you commented on is available now. --Tobias1984 (talk) 08:58, 9 July 2013 (UTC)Reply

The property biological process (P682) that you commented on is available now. --Tobias1984 (talk) 09:04, 9 July 2013 (UTC)Reply

The property ortholog (P684) that you supported is available now. --Tobias1984 (talk) 09:34, 10 July 2013 (UTC)Reply

The property Gene Ontology ID (P686) that you supported is available now. --Tobias1984 (talk) 14:15, 11 July 2013 (UTC)Reply

The property encodes (P688) that you supported is available now. --Tobias1984 (talk) 10:49, 12 July 2013 (UTC)Reply

The property afflicts (P689) that you supported is available now. --Tobias1984 (talk) 07:52, 13 July 2013 (UTC)Reply

The property Gene Atlas image (P692) that you supported is available now. --Tobias1984 (talk) 08:43, 17 July 2013 (UTC)Reply

The property PubMed publication ID (P698) that you supported is available now. --Tobias1984 (talk) 08:58, 18 July 2013 (UTC)Reply

The property Disease Ontology ID (P699) that you supported is available now. --Tobias1984 (talk) 09:11, 18 July 2013 (UTC)Reply

Hi, the property encoded by (P702) that you supported in now available. --Paperoastro (talk) 12:49, 21 July 2013 (UTC)Reply

The property that you supported, separated from (P807), has been finally created. More participation in the property proposal pages would make property creation faster. Go ahead!--Micru (talk) 02:10, 17 August 2013 (UTC)Reply

Somebody merged Tyrosine-protein kinase transforming protein Src (Q13360422) into SRC (Q424737). Not sure if that is right. --Tobias1984 (talk) 17:33, 8 September 2013 (UTC)Reply

Great find, that merge was indeed problematic. It's fixed now. Emw (talk) 17:44, 8 September 2013 (UTC)Reply

I have some concerns. Please, answer on Wikidata:Project chat. Andreasm ^{háblame / just talk to me} 19:14, 9 September 2013 (UTC)Reply

Hi Andreas, I've replied here. Emw (talk) 00:50, 10 September 2013 (UTC)Reply

I, just reading the w3c owl guide. In OWL full, classes can be instances http://www.w3.org/TR/owl-guide/, and in OWL DL it can not. I quote The choice between OWL DL and OWL Full mainly depends on the extent to which users require the meta-modelling facilities of RDF Schema (i.e. defining classes of classes). The main reason it is not allowed in OWL DL is that such features makes harder to reason on the ontology and some questions and the consistency of the ontology might be undecidable. (just to continue an old discussion ;) ). TomT0m (talk) 17:43, 19 September 2013 (UTC)Reply

Reading further, I see The wine ontology as it currently exists would require the ability to treat classes as instances in order to support such an interpretation. Note that OWL Full permits such expressivity, allowing us to treat an instance of a wine variety simultaneously as a class whose instances are bottles of wine. I think Wikidata needs that expressivity. I'll put a word into Help:Basic_membership_properties talk page about that. TomT0m (talk) 18:05, 19 September 2013 (UTC)Reply

[1] I don't understand this diff, in french this item is linked to an article describing any characters in a fiction, not only humans ... Plus we should not mix the fictional and real hierarchy. TomT0m (talk) 08:58, 5 October 2013 (UTC)Reply

Not all persons are human. As noted in the Coco Chanel discussion, the definition of "person" is quite vague (see Personne, Person). That said, the edit you link to was simply me re-adding the 'subclass of person' claim I had removed minutes prior. I replaced the 'subclass of person' claim with 'subclass of fictional entity' at first. I re-added 'subclass of person' minutes later when I consulted the item's Wikipedia article and noted that "person" might apply to non-human fictional characters, since "person" means a being "that has certain capacities or attributes constituting personhood" and I can't think of any fictional characters that does not fit in the very general definition.

If you think the 'subclass of person' claim should be removed, then please go ahead and remove it. I have no strong opinion on the matter. Emw (talk) 15:21, 5 October 2013 (UTC)Reply

Ok, I did not see it was also a subclass of fictional entity. But imho it's not a subclass, it's a fictional equivalent of ... TomT0m (talk) 20:42, 6 October 2013 (UTC)Reply

The notion of "fictional equivalent" needs more specification. Would "fictional equivalent" mean that the subject inherits all the properties of its object? This seems like it might work for a some classes of fiction, but problematic are unusable for other classes of fiction. For example, if we were to say 'Luke Skywalker fictional equivalent of human', then how does that account for the fact that Luke can engage in telekinesis, a property not possessed by any actual humans? Emw (talk) 16:50, 13 October 2013 (UTC)Reply

No it would not, there is no strong constraints in any place of Wikidata anyway, and I don't think there will be at all. It would just be a hint that the fictional stuff is more or less similar to the real stuff for an entity suggester who can suggest the properties he would suggest for the real stuffs, or something that can be used in queries, and something that make sense in general. Of course fictional stuffs are not always totally equivalent to their fictional equivalent, that's sometimes what makes things interesting. TomT0m (talk) 17:08, 13 October 2013 (UTC)Reply

The phrases "equivalent" and "more or less similar" have quite different implications. I'd recommend a less strong, more accurate word; perhaps "fictional analog". Emw (talk) 14:15, 14 October 2013 (UTC)Reply

I noticed Wikidata:Project chat#Many term properties await promotion. Maybe you should become a property creator so you could help out? I can assign the flag and it includes a free {{Sofixit}}! ;-) Multichill (talk) 20:22, 16 October 2013 (UTC)Reply

I'd appreciate it if you granted me property creator permissions. Thanks, Emw (talk) 14:17, 19 October 2013 (UTC)Reply

Hi Emw, I just wonna notify you about this. Best regards, Bene* ^t_a^l_k 08:16, 22 October 2013 (UTC)Reply

Thank you for the note, Bene. I've replied. Emw (talk) 12:27, 22 October 2013 (UTC)Reply

Hi Eric,
The items have been nominated again. I have put the nominations on hold pending your reaction. -Cycn (talk) 13:53, 12 November 2013 (UTC)Reply

Thanks, I have replied again. Emw (talk) 03:10, 13 November 2013 (UTC)Reply

I've raised a Request for Comment for this issue. -Cycn (talk) 13:48, 15 November 2013 (UTC)Reply

you have added many such statements.--GZWDer (talk) 14:00, 15 December 2013 (UTC)Reply

Thanks for the correction in mouse chromosome 17 (Q15305616). I was filling in 'found in taxon' claims for Mus musculus chromosomes, and left a mistaken claim 'mouse' instead of 'house mouse' on that item. If you're aware of more than one such error, please let me know! Emw (talk) 14:11, 15 December 2013 (UTC)Reply

What is the academic discipline or tradition that you are using to give you the language to be able to discuss classification in the way that you do? Do you have any suggestions for what I as a total beginner might be able to read to be able to explain concepts on Wikidata as you do? What is your level of expertise in talking about this kind of classification and how much time commitment did you make to be able to talk in this way? Blue Rasberry (talk) 14:56, 16 December 2013 (UTC)Reply

Lane, when I talk about classification I often use language from ontology, a field of computer science concerned with formal knowledge representation. I tend to use vocabulary from the Web Ontology Language -- OWL. OWL is an ontology language for the Semantic Web recommended by the W3C. It's based on RDF and RDFS.

In addition to the resources linked above, I'd recommend anyone interested in the subject to read the W3C OWL2 Primer. I also recommend The Semantic Web, a 2001 Scientific American article by Tim Berners-Lee and other researchers that laid out a motivating vision. Since you're interested in biology, it's worth pointing out that biologists have been a very active bunch in ontology and Semantic Web practices, coming up with a broad family of ontologies called OBO (Open Biomedical Ontologies) -- see The OBO Foundry: coordinated evolution of ontologies to support biomedical data integration by Barry Smith et al. Check out the Wikidata reading list, too.

I have been enthralled with Wikidata and its potential for structuring knowledge for about a year. I don't consider myself to be an ontology expert, but I have spent a major chunk of this year familiarizing myself with the field and related topics discussed above. I think ontology, the Semantic Web and OWL are all very important things to have in mind as we try to structure all human knowledge. Emw (talk) 13:58, 17 December 2013 (UTC)Reply

The property chromosome (P1057) that you proposed is available now. --Tobias1984 (talk) 23:17, 24 December 2013 (UTC)Reply

This is maybe a dumb question, but I want to understand classes vs. instances better.

What makes something a class? Taking plant (Q756), for example, I see that it is both:

• instance of (P31) → taxon (Q16521) → subclass of (P279) → set (Q36161) → subclass of (P279) → class (Q217594), and
• subclass of (P279) → organism (Q7239).

Would either of these be sufficient for me to infer that plant (Q756) is a class (Q217594)? The latter, because I'd assume that the property subclass of (P279) has a domain constraint to class (Q217594).

What is the actual (inferred) statement that says that plant (Q756) is a class (Q217594)? (As I write this, I realize I'm pretty confused.) Is it:

• plant (Q756) → instance of (P31) → class (Q217594), or
• plant (Q756) → subclass of (P279) → class (Q217594)?

I'd also guess that if there are any statements that specify this item as the object of instance of (P31), then it could be inferred to be a class. Is that right?

For a given (perhaps new) item, is there a canonical way to specify that it is a class?

If you have a minute or two to explain, I'd really appreciate it. Thanks! Klortho (talk) 20:07, 4 January 2014 (UTC)Reply

Klortho, those are great questions. The answers are important, but not at all apparent without diving into the W3C specifications for RDF, RDFS, OWL, etc. I'll try to accurately summarize what they say about your questions.

In a nutshell:.

• Classes are sets, and instances are elements of those sets.
• Being part of an instance of (P31) or subclass of (P279) claim alone is sufficient to infer that an item is a class. As the definitions rdf:type and rdfs:subClassOf say, such items are implicitly instances of rdfs:Class.
• Relatedly, it is invalid to assert in a Wikidata P31 or P279 claim that an item is a class as it concerns those properties (i.e. rdfs:Class). rdfs:Class is reserved vocabulary, and thus fundamentally different than any item on Wikidata, even those labeled 'class'.
• There is no direct way in Wikidata to specify that a given item is a class. P31 and P279 are currently the only ways indicate an item is a class, but they are somewhat indirect.
• An item can be both an instance and a class.

In full:

First, some definitions and background. A class represents a set of individuals, and individuals represent objects from a domain. As an OWL 2 formalism, instances are individuals that fulfill a class expression, but I have never noticed this fine distinction between 'individual' and 'instance' used in practice. The terms 'individual' and 'instance' seem effectively identical. A class is a set of instances.

When the specifications describe a class as a set, they mean it in a rather formal sense. For example, the OWL model theory specification defines subclass of (P279) as SubClassOf(d1, d2) = d1 ⊆ d2 -- 'subset of'. instance of (P31) is effectively ∈ -- 'element of'. The basis of P279 is rdfs:subClassOf and the basis of P31 is rdf:type.

The fact that P31 and P279 are based on rdf:type and rdfs:subClassOf -- special, built-in properties in RDF and RDFS -- entails certain things. First, to answer your questions, if an item is the subject or object of a subclass of (P279) claim, or if it is the object of an instance of (P31) claim, then it is a class. In other words, those conditions are sufficient to infer the item is an instance of a built-in 'class' concept, i.e. rdfs:Class.

I think we should avoid linking rdfs:Class to class (Q217594) or any other item in Wikidata for a few reasons:

1. In Section 5.1: Classes, the OWL 2 syntax specification says "IRIs from the reserved vocabulary other than owl:Thing and owl:Nothing MUST NOT be used to identify classes in an OWL 2 DL ontology." In other words, classes in the ontology -- here, Wikidata -- should not make claims involving resources in the rdf:, rdfs: or owl: namespaces other than owl:Thing and owl:Nothing. (Small digression: 'owl:Thing' exists on Wikidata as entity (Q35120). Q35120 is the top concept, conventionally referred to as ${\displaystyle \top }$  in description logics, the basis of OWL.)
2. class (Q217594) is actually ambiguous. There are different theories of sets, each of which assigns notably different properties to the term 'class' -- which one does rdfs:Class have the semantics of? (ZF, NBG, MK?) rdfs:Class is defined in RDF model theory. Per this technical note in the RDF MT document, rdfs:Class fulfills a notable condition of Zermelo-Fraenkel set theory, but I don't know if rdfs:Class and ZF classes are equivalent. (The second-to-last paragraph here and page 5 here seem to suggest they are not.)
3. When we start talking about upper ontology on Wikidata, the project's assertions enter the domain of metaphysics and philosophical ontology to some degree. (A side note: I think Wikidata claims upper ontologies are useful as a way to practically organize our claims about more specific domains. I do not consider them to be political assertions about the nature of the universe.) While claims about metaphysics and philosophical ontology are in scope for Wikidata, the W3C Semantic Web standards aim to separate the languages themselves from those concerns:

[RDF model theory] tries to be metaphysically and ontologically neutral. It is typically couched in the language of set theory simply because that is the normal language of mathematics - for example, this semantics assumes that names denote things in a set IR called the 'universe' - but the use of set-theoretic language here is not supposed to imply that the things in the universe are set-theoretic in nature.

—Specifying a formal semantics: scope and limitations. RDF Semantics specification

Relating rdfs:Class to class (Q217594) or any other item on Wikidata seems like it would unnecessarily erode the separation between ontology as a philosophy and ontology as field of computer science that the W3C desires. Wikidata is in the business of making content claims (big and small) about the universe based on reliable sources from domain experts, but RDF, RDFS and OWL are designed to provide only the structure of those claims, not their content.

You also ask "is there a canonical way to specify that it is a class?" There is currently no standard way in Wikidata to directly specify that an item is a class. In OWL, this is done with <owl:Class ...> ... </owl:Class> (see e.g. trans.owl), but there is no analog of owl:Class on Wikidata. The most direct way we currently have is to put a P279 claim on the item; the next most direct is to make the item the object of a P31 or P279 claim on another item. These mechanisms seem OK to me for now. They encourage users to connect the items to our growing class hierarchy.

Since you note how plant (Q756) has claims for both instance of (P31) and subclass of (P279), you might be interested in the notion of metamodeling in OWL 2. I wrote about it a bit in Project chat a little while ago: Is Mary a scientist or a profession? Punning in Wikidata with OWL. In brief, a class can be an instance. Think of it this way: a set can be considered a subset or an element of another set. In other words, it is valid to state 'plant (Q756) instance of (P31) taxon (Q16521) and plant (Q756) subclass of (P279) organism (Q7239)'. Note that this is fundamentally different from how plant (Q756) is an instance of rdfs:Class and a class of subclass of (P279); this case mixes the syntax of the ontology (rdfs:Class) with the content of the ontology (organism (Q7239)) and is not what "metamodeling" refers to.

In that Project chat discussion, Filceolaire made a proposal similar to your question. I didn't address the proposal very directly, but my points above about rdf:Class and class (Q217594) do. I think we should avoid adding explicit claims like instance of (P31) class (Q217594) or subclass of (P279) class (Q217594) on items to fulfill a constraint that "Anything using the 'Subclass of' property should be an 'instance of:Class' or some subclass of 'Class'." That constraint is built into the definition of rdf:type and rdfs:subClassOf, and thus 'instance of' and 'subclass of'. Filceolaire's concern is prescient (as usual), it's just that it seems already addressed by default in the linked terms from the RDFS specification, and indicated against by the numbered arguments above.

Please let me know if clarification or more information would help. Emw (talk) 17:21, 5 January 2014 (UTC)Reply

My name was mentioned above so I got a site notification so I'm going to put in my $0.02 worth.
First: Emw this discussion was fascinating and helpful. I hope my comments below add something.
'Plant - instance of - Taxon' tells us what type of class 'Plant' is. It is a taxon, which is a specific type of class, and it is an instance/member of the 'Taxon' class along with every other taxon.
'Plant - subclass of - organism' tells us that every instance/member of the 'Plant' class is also an instance/member of the 'organism' class. Note that none of these instances/members are Taxons; they are plants. None of them are instances/members of the 'Taxon' class. I think that this is the basis on which we should decide whether something is an 'instance of' or a 'subclass of'.
We can get even more meta. 'Plant' is an instance of 'Kingdom' and 'Kingdom' is a subclass of 'Taxon' but 'Kingdom' is also is an instance of 'Taxon rank'. It is a member of the Taxon rank class along with 'Species' and 'Genus'.
At least that is my opinion. Hope that helps. Filceolaire (talk) 22:05, 5 January 2014 (UTC)Reply

This is great, thanks very much to both of you for the discussion!

"The terms 'individual' and 'instance' seem effectively identical." → I guess, not quite: an individual that is not declared to belong to any class is still an "individual", but not an "instance".

"'is there a canonical way to specify that it is a class?' ... The most direct way we currently have is to put a P279 claim on the item; the next most direct is to make the item the object of a P31 or P279 claim on another item. These mechanisms seem OK to me for now." → The problem is that in the absence of any "subclass of" statements, there is no easy way for an editor to see that an item is a class. It may be a class because other items are "instances of" it, but those statements are not visible on the original item's page. I understand your reasons (I think) for not wanting to *require* it, but would making such an item be "instance of" → "class" cause any problems?

I think I now understand the issues that I brought up originally, but one (new?) area of confusion is how classes are used in Wikidata as the objects of properties other than instance of (P31) or subclass of (P279). Three recent examples:

• In your, "Is Mary a scientist or a profession?" post, you wrote about "statements of the form 'X occupation Y' entail that X is an instance of Y". The example was Mary → occupation → scientist, where scientist is a class.
• In the pathogen transmission process property proposal, the objects of that property are all classes.
• Likewise, for the sex or gender (P21), the objects are typically subclass of (P279) "male" or "female"

So, I think these are examples of punning, right? In the statement Chelsea Manning (Q298423) → sex or gender (P21) → trans woman (Q15145782), I think trans woman (Q15145782) is being treated as an individual, not as a class. But on the other hand, since trans woman (Q15145782) → subclass of (P279) → female (Q6581072), I think one should be able to infer the sex or gender (P21) → female (Q6581072) relationship. But maybe there is no general rule like that.

Klortho (talk) 04:32, 6 January 2014 (UTC)Reply

I do not understand why you say that 'trans woman (Q15145782) is being treated as an individual'. I think the P21 statement means that the person (in this case, Chelsea Manning (Q298423)) is an instance of a certain sex or gender (in this case, trans woman (Q15145782)). So the gender or sex is a class. Bever (talk) 05:17, 7 January 2014 (UTC)Reply

@Bever:, that is what I thought at first, but I don't think it is right, and the distinction is crucial, see my last comment under "still to do", here. If Chelsea Manning is an instance of transgender female, then we would be justified in inferring that she is a biological female, which she is not. In the Semantic Web (Emw can correct me if I'm wrong) you're not allowed to infer an "instance of" relationship from some other property, unless that property is declared to be a subproperty of "instance of". And so it still seems to me that the object of sex or gender (P21) should be interpreted as an individual/instance, rather than a class, but that begs the question, what exactly are the subclass relationships there for? Klortho (talk) 02:13, 8 January 2014 (UTC)Reply

Klortho, I do not think it is valid OWL DL to declare a subproperty of P31, i.e. rdf:type. Filceolaire and I have been talking about this intermittently for few a months. My argument has been that I think we should avoid declaring subproperties of rdf:type (P31) because I have never seen such usage in any of the ontologies I've looked at, and it's unclear whether it's even supported by OWL.

I found a note by Ian Horrocks, an editor of the W3C OWL recommendations, that seems to support the notion that it is not valid in OWL DL to have a statement 'X rdfs:subPropertyOf rdf:type':

...the restriction being discussed here, i.e., not being able to create a subPropertyOf rdf:type, is nothing to do with DLs per se, but is required in order to keep the language inside what I think we agreed to call "conventional" FOL. In fact separating the syntax of the language from the domain of discourse is fundamental to most logics.

—Ian Horrocks, Re: Inferring Class Membership w/o OWL Full?

I also found a relevant note in a report about expressing Dublin Core in RDF:

dcterms:type vs rdf:type
There is an issue with declaring a sub-property
of rdf:type. OWL-DL will not accept the assertion that
dcterms:type is rdfs:subPropertyOf rdf:type.
Proposal: This is really an issue for the DCMI Usage Board.
-- Remove statement that dcterms:type is rdfs:subPropertyOf rdf:type.
-- State that in mapping from DCAM to RDF, dcterms:type maps to rdf:type

—Mikael Nilsson, Report from public comment period on Expressing Dublin Core metadata using the Resource Description Framework (RDF)

Regarding sex or gender (P21), I think the objects of such claims are conventionally considered classes, not instances. All classes can be considered instances, but 'man' and 'woman' are canonically treated as the former. See for example http://www.w3.org/TR/owl2-primer/.

You also say "In the Semantic Web (Emw can correct me if I'm wrong) you're not allowed to infer an 'instance of' relationship from some other property, unless that property is declared to be a subproperty of 'instance of'." We can certainly infer 'instance of' relationships from some other property (and, as noted above, we shouldn't do it with a subproperty of P31). This is the point of P279 (rdfs:subClassOf): if Methuselah (Q590039) is a tree, then Methuselah is an organism, because a tree is a subclass of organism. In fact, almost all properties can be used to infer rdf:type relationships, since domain and range entail that the subject or object of a claim with a given property is an instance of the class denoted in the domain or range. (See the definitions of domain and range in those RDFS specification links.) So we see that what classes an item is an instance of can be inferred in many ways. Emw (talk) 05:37, 8 January 2014 (UTC)Reply

The restrictions on declaring subproperties of rdf:type are interesting. And, I hadn't considered subClassOf, domain, and range, when I wrote that 'instance of' relationships can't be inferred. I'm also familiar with a lot of OWL, and I reviewed the primer you referenced. However, the outstanding question here is very simple, and has not been addressed: given the statement "Chelsea Manning (Q298423) → sex or gender (P21) → trans woman (Q15145782)", where trans woman (Q15145782) is a class, can one infer "Chelsea Manning (Q298423) → instance of (P31) → trans woman (Q15145782)"? If so, by what mechanism? By convention? Does the Wikidata RDF export reflect that? If so, how is that different from declaring sex or gender (P21) to be a subproperty of rdf:type?

You pointed me to the OWL primer as a reference to suggest that objects of properties such as sex or gender (P21) are often classes, but I don't see anything in that document that supports that. Yes, they have classes :Man and :Woman in the sample ontology, but I don't think there are any implied properties similar to Wikidata's sex or gender (P21). In other words, :Man and :Woman are only used as classes, in class-type relationships like type, subClassOf, and all of the OWL restriction statements, for example. The more I dig into this question, the more I am convinced that statements of the form "sex or gender (P21) → trans woman (Q15145782)" are just wrong, and should be replaced with "instance of (P31) → trans woman (Q15145782)".

Since I doubt that I'm the only one confused by this, do you think it would be good to move this discussion to Project_chat? Klortho (talk) 14:56, 8 January 2014 (UTC)Reply

In OWL, there is several ways to a class (see http://www.w3.org/TR/owl-ref/#ClassDescription about OWL1 but it should not be so different with the version 2). The one that answer to your question is, in this link property description. With a property restriction, it is possible to define the <female> class (or an anonymous class as all the individual who have a sex statement with the <female> (assuming the fact it is both a class and a value in the same definition is not a problem) value. This way you do not have to explicitely say that an individual belongs to the female class without stating it explicitely, a resoner deduces that. Of course this is more interesting with more complex class definition. TomT0m (talk) 17:12, 8 January 2014 (UTC)Reply

---

Per Klortho, I have copied this conversation to Project chat: Wikidata:Project_chat/Archive/2014/01#Question_about_classes.2C_and_.27instance_of.27_vs_.27subclass.27. Let's talk there. Emw (talk) 00:48, 9 January 2014 (UTC)Reply

 

Hi Emw! I have spent some time reading over the "disjointed" property proposal and I noticed that I don't really understand how statements work on Wikidata. I sometimes add two subclass of statements and think of thems as competing. Other times I add them and think that they both contain one part of the information. How does semanticWeb solve this? --Tobias1984 (talk) 15:48, 7 January 2014 (UTC)Reply

Hi, sorry if I answer instead of EMW but I have this interesting Talk page in my Watchlist /o\. In standard interpretation of ontology in semantic web, statements are all supposed to be true. That means if you add two subclass statements, the class is a subclass of both classes.

If two statements are competing, they therefore should be explicitely noted as competing, with an equivalent of the logical or operator to maintain the consistency of the ontology, or some kind of way we have to find.

Now if, for example, we have a disjoint with statement between two classes, and unfortunately an item is an instance of two classes, that means a reasoner would infer some inconsistency, and that is pretty bad because a reasoner can usually infer anything from an inconsistency per en:explosion principle, which means the ontology is useless for him. Maintaining consistency is an important and almost impossible challenge in an ontology as big as Wikidata. It will be easier to just selects some items an statements to reason with and check if they are consistent before injecting them to a reasoner.

That said Wikidata is by design meant to have inconsistencies are some sources may be competing. We would have, to reason with them, either mark them as inconsistent, just select the consistent claims before reasoning with them, for example the claims that are sourced with sources we know are consistent, or by using some kind of en:modal logic, which reasons in terms of possible world. I think the description logic that are the basis of OWL are close (or even equivalent) to modal logics, but I have to investigate. TomT0m (talk) 16:54, 7 January 2014 (UTC)Reply

Does that mean that "disjoint" is just like a second check on the logic inferred from the subclass of statements? --Tobias1984 (talk) 20:28, 7 January 2014 (UTC)Reply

Because Wikidata works under the open-world assumption (Q851949), a statement which is made is only an affirmative statement, and that a statement which goes unmade is not implicitly an incorrect statement. The semantics of "disjoint with" would make the unmade statement an incorrect statement explicitly, in the case of classes. --Izno (talk) 20:42, 7 January 2014 (UTC)Reply

To be a little more correct, Wikidata does not work under Close/Open world as it does not make any inference at all :) It's OWL that is specified like that, and external tool we will plugs on the datas and understands OWL. Wikidata is not about truth, like Denny said, Wikidata will have a query engine, probably not an inference engine. TomT0m (talk) 21:25, 7 January 2014 (UTC)Reply

We talk about OWL so much that I apparently fail to distinguish between OWL and Wikidata. ^_^ --Izno (talk) 16:50, 8 January 2014 (UTC)Reply

No, an individual can absolutely be an instance of any number of classes, even if they are not subclasses of each other. Instance can even be inferred from a definition : if we define a <women> class beeing all instances of human that are of female sex, if we know it's a human and a female, then the reasoner can infer it belongs to the Woman class without having to state that explicitely. TomT0m (talk) 21:07, 7 January 2014 (UTC)Reply

I would oppose locking this page from edits. (We should also not blow up at occasional editors like that.)

I will not lock the page; I've been too long involved with the discussion to even think about it. Just letting the frustration show, I guess…

As for "occasional", the editor makes it clear that he left, came back, and decided to be unhappy for a second time. I give short shrift to editors who "vow [...] never to get involved again" (his wording), and when they do inevitably return, they tell us what's wrong in a completely unconstructively critical fashion, without showing any evidence that he has ingested the information at the very top of the page about the property. It's my opinion that if someone is going to leave, he should just leave, and that being dramatic about the fact that he is leaving is a really poor decision on his part. Grack. --Izno (talk) 19:17, 10 January 2014 (UTC)Reply

Just created Statistics Denmarks classification of occupation (DISCO-08) (P1069). --Tobias1984 (talk) 16:18, 15 January 2014 (UTC)Reply

To delete an item, you can start an request for deletions on WD:RFD or use the gadget RequestDeletion. --Pasleim (talk) 22:00, 21 January 2014 (UTC)Reply

Thanks Pasleim. I've added the 'RequestDeletion' item to my gadgets via 'Preferences'; I didn't know about it. I requested the item be deleted. Emw (talk) 00:31, 22 January 2014 (UTC)Reply

Just created Mohs' hardness (P1088). --Tobias1984 (talk) 07:37, 31 January 2014 (UTC)Reply

Just created electronegativity (P1108). --Tobias1984 (talk) 17:22, 31 January 2014 (UTC)Reply

Just created national team appearances (P1129) --Tobias1984 (talk) 14:55, 2 February 2014 (UTC)Reply

Just created Kunstindeks Danmark Artist ID (P1138). --Tobias1984 (talk) 18:07, 4 February 2014 (UTC)Reply

Hi, Emw and Multichill. I saw this example of page where users can put information concerning properties. Here I used this type of page to put hierarchy model of the property P196. What do you think? --Paperoastro (talk) 15:22, 22 February 2014 (UTC)Reply

Paperoastro, I think the footer template would be best used as way to make RDF-exportable statements about properties, for example as seen in http://dbpedia.org/ontology/capital. The content of Property_talk:P492/footer seems more appropriate for the top of the Property_talk:P492. Emw (talk) 15:49, 22 February 2014 (UTC)Reply

You are right: I forgot that discussion. I will move the contents of the page and delete it. Thank you for the explanation! --Paperoastro (talk) 16:06, 22 February 2014 (UTC)Reply

Something is wrong with ProteinBoxBot and biological process (P682): G1 (Q5512664) [2]. G1 (Q5512664) is a website. There seems to quite a few erroneous insertions https://www.wikidata.org/wiki/Special:WhatLinksHere/Q5512664 — Finn Årup Nielsen (fnielsen) (talk) 15:39, 20 March 2014 (UTC)Reply

  WikiProject Molecular biology has more than 50 participants and couldn't be pinged. Please post on the WikiProject's talk page instead. --Tobias1984 (talk) 07:23, 21 March 2014 (UTC)Reply

I put a note on the bot talk page. Klortho (talk) 03:56, 30 March 2014 (UTC)Reply

moved to Wikidata:Project_chat#best_pratices per

TomT0m, can we please keep this conversation in one thread, i.e. Wikidata:Project_chat#Reboot? It helps keep things coherent. Thanks, Emw (talk) 17:46, 10 May 2014 (UTC)Reply

I'm really sorry for reverting your edit. https://www.wikidata.org/w/index.php?title=Wikidata:Properties_for_deletion&oldid=135524583&diff=prev it's my fault. It's my mistake. I'm really sorry. --Konggaru (talk) 02:25, 1 June 2014 (UTC) Konggaru, no worries, that's OK! Emw (talk) 02:44, 4 June 2014 (UTC)Reply

Just so you know, this item is appearing at User:Pasleim/notability. It may be deleted if it isn't made to pass WD:N. Cheers. Del_♉sion23 (talk) 19:41, 8 July 2014 (UTC)Reply

Thanks for the heads up! Emw (talk) 01:39, 9 July 2014 (UTC)Reply

Certainly not for all, but what superclass would you use at fruit (Q1364)? Tamawashi (talk) 01:33, 11 July 2014 (UTC)Reply

Tamawashi, agreed. food (Q2095) would fit, I suppose. Emw (talk) 02:09, 11 July 2014 (UTC)Reply

Maybe there are fruits that are not food. Applied natural physical object (Q16686022) for now. Tamawashi (talk) 11:41, 11 July 2014 (UTC)Reply

Tamawashi, good point, there are indeed fruits that are not foods. I think it would be more appropriate to classify fruit as a "biological object", or perhaps even a "botanical object". Emw (talk) 11:50, 11 July 2014 (UTC)Reply

I find it a little bit difficult to find subclasses of a class when I am on the class page. That would have helped with biological object. I saw physical object on the food page and remembered to have seen natural physical object, that's the reason I used that. Any tool that would help here, e.g. gadget that shows subclasses? Tamawashi (talk) 11:55, 11 July 2014 (UTC)Reply

Tamawashi, I'm not aware of such a gadget, but it would be very useful. Perhaps it could use the code Wikidata generic tree does, and limit the displayed subclass tree to the first 3 or 4 layers for speed.

Returning to the content issue, I have reservations about using the vague terms "natural" and "artificial" as criteria for main branches of "physical object". Many fruits only exist because of intense biological selection from millenia of careful cross-breeding by humans. So can all instances of fruit really be instances of "natural physical object"? Those are more terms of informal folksonomy and lead to dubious organization. My initial impression is to remain neutral on such questions and do away with natural physical object (Q16686022) and artificial physical object (Q15222213). Emw (talk) 12:04, 11 July 2014 (UTC)Reply

Agreed. Tamawashi (talk) 13:03, 11 July 2014 (UTC)Reply

@Tamawashi: I created {{Classification}} for this purpose, {{Item documentation}} includes it and is on the easy accessible talk page of the item. It would be really easy to create a gadget : take the Reasonator gadget code as a model or User:TomT0m/Leonore.js for example. TomT0m ^(talk) 14:54, 11 July 2014 (UTC)Reply

@TomT0m: I added {{Item documentation}} to several talk pages, but was too lazy to do it here once again. Thanks a lot for this template, it is really helpful. Tamawashi (talk) 15:03, 11 July 2014 (UTC)Reply

@Tamawashi: True, I just added this future bot in my TODO list :) TomT0m ^(talk) 16:50, 11 July 2014 (UTC)Reply

GUI toolkit or framework (P1414) is done. Thanks for your input. -Tobias1984 (talk) 16:30, 11 July 2014 (UTC)Reply

Related to Wikidata talk:Item classification#Difference between instance and class : I linked there to [3], where you wrote about items that can be both. Maybe you can create a section in Wikidata:Item classification based on that text to clarify it for a wider audience? Tamawashi (talk) 16:53, 12 July 2014 (UTC)Reply

Emw, I am afraid that we are entering a deeply philosophical domain... and that we might not be ready to deal with it... one always approaches the topic with preconceptions, expectations, a cultural baggage, etc... and that might prevent us to see clearly. Knowledge doesn't help here.

OTOH, a user is about to be blocked because we didn't provide clear guidelines of our basic properties... and that hurts me twofold: for not being able to help him understand and share a common vision with all our colleagues, and also because of the negative consequences that it will cause to everyone of losing a valuable contributor. We have been dragging on this issue for so long... and now we seem to be so close. Please, do not only read ontologies, do take some hours to look at the world around you, meditate, and observe how your mind works.

I am learning a lot by reading the papers you recommend me, to express insights in a way that can be understood by people familiar with web ontologies. I hope that you can also learn something by following my recommendation of not just reading papers, but also observe how thoughts emerge in the perfect ontological machine that is the mind, and make your own decision if the way that we have been describing reality lately, matches your own subjective reality.

Thanks for your patience with everyone, and for your dedication. I deeply appreciate it.--Micru (talk) 14:12, 21 July 2014 (UTC)Reply

"...and now we seem to be so close."

Such optimism is helpful. The discussions currently underway are the latest iteration in a series of several long-running, often-dormant, interweaving threads of investigation about the nature of classes and instances. It's rare to see so many of them revamp at the same time, so publicly and with so much energy.

I spent a good while yesterday in thought, per your recommendation. I thought about the nature of things, existence, my subjective experience, how I relate to the universe, and how my thoughts come into being. I didn't have any major revelations but it was a nice time of introspection and meditation.

My approach to inquiry is somewhat different, as you know. The two can be complementary, but I prefer description logic to phenomenology. The Semantic Web -- which Wikidata is a seminal project in -- overwhelmingly uses the language of Horrocks, not Husserl.

Thanks for the suggestion to try something new. It was a good experience. I'm sure we'll be exploring further the intersection of certain schools of philosophy with certain schools of formal knowledge representation, as it should be. Hopefully that discussion will produce some greater clarity on how to apply those basic properties throughout Wikidata. Emw (talk) 12:38, 22 July 2014 (UTC)Reply

There is an emerging consensus in neuroscience that the brain is a metastable system. Besides of Tononi's paper, if you feel reading something new there is the very interesting Phenomenological architecture of a mind and Operational Architectonics of the brain: the unified metastable continuum. As it appears to be, the brain-mind is an system that collapses quantum states either into "phenomenal consciousness", when exposed to phenomena, or into "operational brain functioning", when thinking and using abstraction. What we experience either as "stream of thought", or as "experience continuum" is the activity wavefront collapsing into somethingness-nothingness. But actually this impresion of "some-thing (real)"/"no-thing (abstract)" is because the energy levels present and our unified volition chosing one of them. Sleep is necessary to keep the brain "on the edge", allowing us both to think and to experience without falling into too much stability or instability.

The dichotomy between description logic and phenomenology is a false one, and we should approach it the same way as a wave–particle duality. The Semantic Web uses the language of Horricks because it has been written by "thinkers" for "thinkers". If it had been written by "experiencers" we would see Husserl's vocabulary all over the place. Wikidata is the confluence of both, that is why we had seen so many conflicts, because applying just one of both views is inherently wrong and unrepresentative of the natural world. Only by acknowledging both natures, it would be possible to accommodate both intensional and extensional definitions in a formal knowledge framework closer to the nature of reality. And yes, thanks also to the contributions of TomT0m, and to Zolo we are closer to that than we had ever been before! :)

I'm glad that you enjoyed introspection, it is a powerful tool to bring mental clarity. If you repeat it consider that the main goal is to "not-think" and to "not-experience", thus enabling the third way of knowledge acquisition (the other two being "a priori" and "a posteriori") that is insight (νοῦς or विपश्यना, according to classic traditions). It is normal not to have any insight for some time because, as my own experiences showed me, it is extremely hard to not-think and to not-experience. Besides of all prejudices that one has to set aside, it also appears as a very counter-intuitive and unfruitful path, and that makes it even a higher personal barrier to overcome.--Micru (talk) 13:44, 23 July 2014 (UTC)Reply

Hi, I want to point to you the discussion about number of platform tracks (P1103) on WD:PFD. It seems to highlight some of the points I pushed about the use of the same properties on classes and instances, here part of (P361) and has part(s) (P527), per Zolo's comment. But I also thinks it applies to the causal properties you proposed.

I tend to think for causal model we could have another property for classes of events, for example <this type of illness> triggered by <infection>. TomT0m ^(talk) 14:43, 21 September 2014 (UTC)Reply

Thank you very much for the explanation, especially for the aliases. Is there any chance of renaming that term? UBERON is pretty much the go to ontology for anatomy. There were other e.g. CARO, MA but UBERON has taken over everything. I agree that it would be a good one to use for Wikidata. I think we will wait a bit more and then move forward with a property proposal.

Emitraka, I'm glad that a freer alternative to the rather copyright-encumbered FMA ontology has picked up steam. The fact that Chris Mungall is involved with Uberon is also a good sign. Regarding the name of 'alias', I don't think there's much chance in it being changed -- it's a fairly baked-in term throughout Wikidata. Emw (talk) 03:31, 24 September 2014 (UTC)Reply

Hi Emw, after scanning constraint violations for value types for the targets of property interaction (P517), I figured that gravity (Q11412) was not defined like other interactions, being subclass of fundamental interaction (Q104934) rather than instance... but after fixing it I got from history that you had done the opposite change 2 days ago. Now I don't have strong opinions about subclass vs instance in that case, but I would like things to remain consistent. All 4 fundamental interactions should remain defined the same way, and the Value type constraint on Property talk:P517 should reflect that convention.

Now on instance vs. subclass, I would prefer instance because no other entity will ever use "gravity" as its class - I don't see these interactions as classes, thus I'd keep them as plain instances of iterations. If you perceive that differently, go ahead and change them but just make sure things remain coherent, as much as possible. thks - LaddΩ chat;) 22:56, 21 September 2014 (UTC)Reply

@Laddo: (pardon my irruption) I guess gravity is a class of interaction for sure. But if I look at the definition of fundamental interaction in enwiki : Fundamental interactions, also known as fundamental forces or interactive forces, are modeled in physics as patterns of relations in physical systems I would say fundamental interactions are patterns, so more like class of classes. Gravity interaction is the class of interaction who matches laws of gravity. I'd use instance of (P31).

The other option would be defining the fundamental interaction class as a class of interaction. The definition would then be something like: a fundamental interaction is an interaction beetween physical corpses who match any of the pattern : gravity interaction, weak interaction, ..., and that does not seems very satisfying to me as it does not tell much about what is a fundamental interaction law except an explicit list.

There is a third option : using both way of classifying, and have two items ; one to class the patterns of interaction, or to class the laws of physics themselves equivalently, and one to divide interaction themselves beetween fundamental one and non fundamental one.

That would give something like

• ⟨ gravity ⟩ instance of (P31)   ⟨ fundamental interaction law (or pattern/or type) ⟩
• ⟨ gravity ⟩ subclass of (P279)   ⟨ fundamental interaction ⟩

TomT0m ^(talk) 09:46, 22 September 2014 (UTC)Reply

Hi, you may have read in another post that I am coming to Wikidata as a student in fundamental Physics. In fact I am currently starting to write an ontology for the description of physical sciences and as a commonner, I plan to participate to Wikidata as soon as I have enough skills to be of any help.

By the way, I am quite new to Linked Data and I would like to know if OWL ontologies and the Wikidata ontology are compatible - i.e. in what extend can an external ontology be of any use to the Wikidata one and vice-versa.

Best regards,

Jibe-b (talk) 12:24, 23 September 2014 (UTC)Reply

Hi Jibe-b -- welcome to Wikidata. The Wikidata ontology has only a subset of features in OWL, so while Wikidata can be exported as an OWL ontology, not all OWL ontologies are importable into Wikidata. By and large, though, the Wikidata ontology should be compatible with OWL ontologies, and vice versa. I recommend reading Introducing Wikidata to the Linked Data Web, which describes Wikidata in terms of Semantic Web technologies like RDF, RDFS and OWL. Note that rdf:type and rdfs:subClassOf exist on Wikidata as instance of (P31) and subclass of (P279), respectively. You can find official RDF and OWL exports of Wikidata at http://tools.wmflabs.org/wikidata-exports/rdf/exports/.

What do you have in mind for your physical sciences ontology? Have you taken a look at existing ontologies in that domain, like ChEBI or the Ontology of Physics for Biology (browse, paper)? There are ongoing discussions about how (and how much) to import external ontologies into Wikidata here. Best, Emw (talk) 12:53, 23 September 2014 (UTC)Reply

Thanks you for your answer. I will read the references about the Wikidata ontology and external ontologies and come back if I have any question.

What I have in mind is the development of an ontology for Physics written with the physicist approach. Indeed, I will base part of it on what already exists (one of my problem is the difficulty to find the license of the ontologies I read) and I will have some help from a German researcher who developped the biotop ontology [4]. It is first a project for getting to learn about ontologies and Linked Data but as soon as I reach something valuable, I plan to work further on this ontology.

What should I take care of in writing an OWL ontology in order to make it relatively close to the Wikidata ontology structure? (I may find the answer to this in the papers about but the answer would help me)

Best regards

Jibe-b (talk) 14:33, 23 September 2014 (UTC)Reply

Jibe-b, regarding what to take care of, BioTop is probably a decent model to start with. The Wikidata ontology is immature at the moment, so as long as you construct an ontology in valid OWL 2 DL, you'll be well on your way. I also recommend looking at how subjects in physics are modeled in Wikidata, e.g. the Wikidata 'particle' taxonomy at http://tools.wmflabs.org/wikidata-todo/tree.html?q=Q1621273&rp=279&lang=en. (The full taxonomy is here, but takes very long to load.) See also: Wikidata:List_of_properties/Natural_science, Wikidata:List_of_properties/Generic. Let me know how your ontology building goes! Emw (talk) 03:10, 24 September 2014 (UTC)Reply

https://lists.wikimedia.org/pipermail/wikidata-l/2014-October/004678.html -Tobias1984 (talk) 18:00, 2 October 2014 (UTC)Reply

Just created Orphanet ID (P1550). Thanks for supporting it. -Tobias1984 (talk) 16:52, 6 October 2014 (UTC)Reply

Just created number of casualties (P1590). --Tobias1984 (talk) 17:29, 3 November 2014 (UTC)Reply

and defendant (P1591) -Tobias1984 (talk) 17:35, 3 November 2014 (UTC)Reply

and prosecutor (P1592), defender (P1593), judge (P1594), charge (P1595), penalty (P1596) Tobias1984 (talk) 17:46, 3 November 2014 (UTC)Reply

number of cases (P1603) is ready. --Tobias1984 (talk) 18:12, 7 November 2014 (UTC)Reply

I took the liberty of reproducing a part of your explanations from the project page onto the the property metadata proposal page: Wikidata:Property_proposal/Property_metadata#Read_this_first. I hope you agree, it was an excellent primer indeed. -- LaddΩ chat ;) 23:52, 14 November 2014 (UTC)Reply

Looks great, thanks! Emw (talk) 12:12, 15 November 2014 (UTC)Reply

Dear Eric, I have added an item (Khaldoun Zreik - researcher), and I found in the "History section" that you have improved it. I thank you for that, but I would like to understand, how did you get the information you added? Was it automatic '? (a script that searches the web for structured data about "Khaldoun Zreik - researcher" and adds it to WIkidata automatically) ? or it was You who found and added the data manually?

this is important for me to understand how to better participate in the future? Thanks!

Hi there, I found the information in the first two search results for https://www.google.com/search?q=khaldoun+zreik and added it to Khaldoun Zreik (Q18536959) manually. The item struck my interest when I saw it was full of nonsensical statements (see here) like "Khaldoun Zreik population 1,200,000", presumably entered as part of an experiment. I stumbled upon the item while researching a new user's assertion that we need domain and range constraints because such unsemantic population claims etc. can be made. See that discussion; I don't think domain and range are designed for that kind of thing.

(P.S.: When leaving a post, it's conventional to sign it, which you can do by putting ~~~~ at the end of your comment.) Best, Emw (talk) 13:44, 18 November 2014 (UTC)Reply

BBC Things ID (P1617) is ready. --Tobias1984 (talk) 11:20, 23 November 2014 (UTC)Reply

date of official opening (P1619) is ready. --Tobias1984 (talk) 11:37, 23 November 2014 (UTC)Reply

plaintiff (P1620) is ready. --Tobias1984 (talk) 11:43, 23 November 2014 (UTC)Reply

has melody (P1625) is ready. --Tobias1984 (talk) 12:03, 23 November 2014 (UTC)Reply

TERC municipality code (P1653) is ready. Thank you participating in the discussion. --Tobias1984 (talk) 12:11, 21 December 2014 (UTC)Reply

P1655 (P1655) is ready. --Tobias1984 (talk) 12:17, 21 December 2014 (UTC)Reply

unveiled by (P1656) is ready. --Tobias1984 (talk) 12:23, 21 December 2014 (UTC)Reply

MPA film rating (P1657) is ready. --Tobias1984 (talk) 12:27, 21 December 2014 (UTC)Reply

I've created "has index case", let me know if I erred, this is my first property creation. :P --AmaryllisGardener ^talk 03:55, 28 December 2014 (UTC)Reply

I just saw this diff

We really need to solve the hydrogen definition issue. So will we create two items, one for hydrogen atom and the other substance made only of hydrogen atom ? There is jsut no controversy if we do this, we just need to split this item. TomT0m ^(talk) 07:58, 22 April 2015 (UTC)Reply

Please see the comment above. --- Jura 09:48, 9 May 2015 (UTC)Reply

Do you have a list of acceptable value for this field? --Richard Arthur Norton (1958- ) (talk) 13:49, 16 July 2015 (UTC)Reply

I see them now at w:Casualty (person)

Spiders can't cause an insect bite, because they are Arachnida (Q1358)--Kopiersperre (talk) 08:07, 26 October 2015 (UTC)Reply

What about this line:

Arachnida (Q1358) #Araneae "Only about 200 species, however, have bites that can pose health problems to humans" d1g (talk) 16:04, 16 May 2017 (UTC)Reply

This is a kind reminder that the following properties were created more than six months ago: number probable (P1675), number suspected (P1676), EC enzyme classification (P660), has index case (P1660), index case of (P1677), ploidy (P1349), age of onset (P2841). As of today, these properties are used on less than five items. As the proposer of these properties you probably want to change the unfortunate situation by adding a few statements to items. --Pasleim (talk) 19:11, 17 January 2017 (UTC)Reply

https://www.wikidata.org/w/index.php?title=Special:Contributions/JakobVoss&offset=20170825203248&limit=1000&target=JakobVoss

Hi, I created a new page for collecting sites that could be added to Mix'n'match and I plan to expand it with the ones that already have scrapers by category. Feel free to use, expand. Best, Adam Harangozó (talk) 19:57, 19 October 2019 (UTC)Reply