Wikidata:Project chat

From Wikidata
Jump to: navigation, search
Shortcut: WD:PC
Wikidata project chat
Place used to discuss any and all aspects of Wikidata: the project itself, policy and proposals, individual data items, technical issues, etc.
Please take a look at the frequently asked questions to see if your question has already been answered.
Requests for deletions and mergers can be made here.
IRC channel: #wikidata connect
Start a new discussion
On this page, old discussions are archived. An overview of all archives can be found at this page's archive index. The current archive is located at July.


Administrators'
noticeboard

Translators'
noticeboard

Bureaucrats'
noticeboard

Requests
for permissions

Requests
for deletions

Properties
for deletion

Requests
for comment

Project
chat

Contents

Javascript devs : an entity selector for Wikidata discussion[edit]

@Tpt, Magnus Manske, Lydia Pintscher (WMDE), YMS: (@...) Hi, I begun (sketched actually) the dev of a javascript gadget on test Wikidata to insert an entity in wikitext but I can't find the time to develop it right now. It's upsetting me because I think it's important, so I through a bottle here, if someone think it's important we can make this effort collaborative :)

I thought I would replicate the search or entity selector in item pages in the wikitext edit gadget. Except validation would insert something like {{subst:Q}} in the Wikitext. One other solution is adding something to the search widget to insert in wikitext instead of going to the entity page (Lydia could this be a good query to add a ticket or a feature of the future UI ?)  – The preceding unsigned comment was added by TomT0m (talk • contribs).

Thematic mapping of geographic statics, US and World[edit]

My interest lies in statistical phenomenology that can be shade mapped by states in the US and countries around the world. Such phenomena as demographics, crime, opinion, health, etc. -- any phenomena that has statistics by area for which the statistic can be listed, statistically summarized, then I want to convert the statistical phenomena into shade maps of the geographic areas they represent and possibly apply bi-variate statistics on the geographic distribution such as center of gravity, etc.

I am new to Wikidata and want to get an understanding of the statistical data you are assembling and if data mapping is already produced or would be amenable to adding such map graphic products to your data series.

I am having trouble getting connected into Wikidata. I contribute financially to Wikipedia.

Please help me get connected and directed appropriately.

Charles Barb cebjrphd@yahoo.com

Proposing item merge and merging[edit]

Hoi,

I've nominated/proposed two Wikipedia articles which describes the same person for merging. I know Wikidata Game can be used for merging. But how do I nominate items for merging? Also merging without the game. Items involved Q17323735 and Q17320640. --Enock4seth (talk) 12:05, 8 July 2014 (UTC)

I would enable the "Merge" gadget at https://www.wikidata.org/wiki/Special:Preferences#mw-prefsection-gadgets in your settings. Then you get a "Merge with..." link/button at the top of the Wikidata website under the "More..." menu. With that one you can easily merge two items. The problem in your case though is that those items have interwiki links that conflict with each other. Both have a link to the English Wikipedia, a Wikidata item can only store one link per Wikipedia (language). So this one also needs to be resolved at the English Wikipedia, those two articles probably should be merged into one single article. BTW: Did you manually insert the time stamp in your signature? Somehow looks wrong ;) --Bthfan (talk) 07:46, 9 July 2014 (UTC)
Bthfan, the problem is the merge on Wikipedia, Wikidata behaves fine.

I was wondering the same thing, and since I do tend to merge items now and then on English Wikipedia after noticing this with two Italian articels I went ahead and merged them (gasp!) the Italian Wikipedia. So far I received no backlash, but the guy was very dead and the info looked 100% covered. This one is going to take some time, because it's a recent death and they are both new articles. The same rules hold as for here: oldest article trumps newer one, so I would say "Castro de Destroyer" is where the item should point to, but it's best to just wait it out a week or two - I see a merge template is already in place. Jane023 (talk) 07:54, 9 July 2014 (UTC)
@Ebraminio: ideally, I think that when the two items to be merged have sitelinks to the same language, MediaWiki:Gadget-Merge.js should display a message asking if the two articles are really about the same topic, and if we click yes, it should send a message or add a Template:Merge (Q6919004) to the articles. --Zolo (talk) 09:54, 9 July 2014 (UTC)
What a really great and creative idea but it would be subject of another gadget IMO. I wrote merge.js when there were no merge API and anyone was doing merge (moving each of claims and sitelinks) manually. IMO it should just do things that would be easy review and revert-able by other users and a cross-wiki request Template:Merge adding is outside of its goal however I will not stop anyone else to implement such ability on merge.js itself. Just IMHO requesting content merge is a prestep of an actual item merge, you see a lot of specially lengthy articles on enwiki that is not merged after years of requesting merge and this essentially could be related on activity of that wiki community so it is not an action that after a request and just a short time would be possible. This can be subject of a bot that automatically request merge for Wikipedia communities and check if community done the content merge then do the Wikidata items merge –ebraminiotalk 15:07, 9 July 2014 (UTC)
@Ebraminio: Maybe we should start by marking the item as a duplicate here on Wikidata. Something like Wikimedia duplicated page (Q17362920) (perhaps using a "duplicate of" property rather than an "of" qualifier, but you see the idea). Would that be difficult to adapt the tool so that it can automatically add it ? --Zolo (talk) 17:24, 15 July 2014 (UTC) + copy-edited: Zolo (talk) 17:54, 15 July 2014 (UTC)
@Zolo: You mean property ? this is a detail, but it's a claim about the item, not about its type. TomT0m (talk) 17:48, 15 July 2014 (UTC)
@tomT0m: Yes, I meant property sorry (fixed above for clarity). -Zolo (talk) 17:54, 15 July 2014 (UTC)
@Zolo: Sorry I didn't get any notification. Hmm, and a bot can track the item actually, I like the idea. –ebraminiotalk 16:20, 17 July 2014 (UTC)

I have encountered duplicate articles in the same language in the past. If I feel like trying, I look for the templates in that language, to propose mergers:

I assume it's the same in most languages as on English Wikipedia:

  • On the article with the better name: {{mergefrom|Some worse name}}
  • On the article with the worse name: {{mergeto|Some better name}}

or if you don't know which is better:

  • {{merge|The other name}} on each article

The help page for each project is Wikipedia:Merging (Q11773605). --Closeapple (talk) 21:08, 10 July 2014 (UTC)

ttwiki duplicate Barack Obama[edit]

The Tatar Wikipedia (ttwiki, татарча/tatarça) has two articles for Barack Obama. First, for Barack Obama (Q76), there is tt:Барак Обама, which I guess is language tt-Cyrl. Then for Barack Obama (Q13202704) there is tt:Barak Obama, which I guess is tt or tt-Latn; it has a template that says "Bu mäqälä Tatar (Latin) Wikipediäseneñ saylangan mäqälälär rätenä kerä." I don't understand that language. Is that allowed on ttwiki? What does that template mean? (@Dobroknig: because that's the only active user with Wikidata:Babel en + tt.) --Closeapple (talk) 20:28, 10 July 2014 (UTC)

For Estonia (Q191) too... I've created duplicate (no label) (Q17363906) as a temporary solution. --Infovarius (talk) 21:53, 15 July 2014 (UTC)
I've left a message at User talk:Dobroknig#2 articles on ttwiki? that points here. I hope that user sees it: I think that is the only user that knows both English and Tatar. It looks like the user appeared here every week or so for the last month. --Closeapple (talk) 15:36, 17 July 2014 (UTC)

Era[edit]

How do I show that Aymestrey burial (Q17351670) belongs to the Early Bronze Age? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 21:15, 13 July 2014 (UTC)

I also miss a property for dating that is not year based. How do I state that Siljan Ring (Q4993174) was created during Devonian (Q65955)? /ℇsquilo 13:11, 16 July 2014 (UTC)
@Pigsonthewing, Esquilo: There is valid in period (P1264). Is that what you are looking for?--Micru (talk) 13:17, 16 July 2014 (UTC)
Not really. Anything about Siljan Ring (Q4993174) that was valid 100 million years ago is still valid. It is still an impact structure, it is still in Sweden etc. /ℇsquilo 17:46, 16 July 2014 (UTC)

Subclass of two different things[edit]

How to we account for items which are a subclass of two distinct things? Example biocide which can be a chemical substance or a micro-organism. But with the way the data is structured it will look like all items below it in the tree will be both a chemical substance and a micro-organism, rather than one of either. Is there a way of fixing this that I'm not aware of? Thanks. Delsion23 (talk) 21:17, 14 July 2014 (UTC)

Split the item? --ValterVB (talk) 21:22, 14 July 2014 (UTC)
I've learned from discussions with Emw to be open minded: Belonging to subclass of (P279) means all and nothing.--Succu (talk) 21:38, 14 July 2014 (UTC)
Jabs aside, the statement "x subclass of (P279) y" means "all instances of x are also instances of y", per rdfs:subClassOf.
The question Delusion23 asks is more interesting. What do the statements below mean?
subclass of microorganism
subclass of chemical substance
It means that any instance of biocide is an instance of microorganism and chemical substance. In other words, it means that any biocide is both a microorganism and a chemical substance. Having multiple subclass of statements means that the subject is a Boolean AND of the object of each subclass of claim. It is the intersection, not the union -- see page 14 here. Thus those subclass of statements lead to the unintuitive and incorrect statement that something can be both a chemical substance and a microorganism. Chemical substance and microorganism are disjoint; anything that is an instance of both classes implies an inconsistency (a falsehood) among our statements.
This example from Delusion23 indicates that we need a way to express union -- i.e. Boolean OR. We should do this in a way that is easily compliant with W3C standards and the fact that subclass of (P279) has the semantics of rdfs:subClassOf. Emw (talk) 01:23, 15 July 2014 (UTC)
Indeed, handling of and/or claims is something that often seems lacking. Also, the ability to indicate weather something is mandatory or not, i.e. a bottle can be made of glass, plastic, but neither are mandatory. But, it is mandatory for a gold watch to be made in part from gold. Danrok (talk) 02:06, 15 July 2014 (UTC)
Mandatory and optional property restrictions can be achieved with expressions like "all gold watches has part (P527) gold" and "some bottles has part glass". This requires universal quantification (∀) and existential quantification (∃), which are supported in OWL as owl:allValuesFrom and owl:someValuesFrom, respectively.
These complex statements would be awkward to capture in the current UI. They require class expressions. The human-readable, W3C-standard way to make such statements is with Manchester Syntax -- a good overview and comparison is here. We would likely need some UI innovation to suitably enable class expressions, which would allow us to make these more complex -- but very interesting and important -- statements. Emw (talk) 02:35, 15 July 2014 (UTC)
What about something like volleyball (Q1734)? It's listed as a subclass of both team sport (Q216048) and ball game (Q877517). We should probably only have one, but which? Both of those are in turn listed as a subclass of sport (Q349), so we'd be losing out on some information if we only put one. Cbrown1023 talk 03:19, 15 July 2014 (UTC)
Those statements are technically correct: volleyball is both a ball sport and (i.e. Boolean AND) a team sport. There is a separate aspect here about information architecture worth considering. Most properties could have captured with multiple instance of or subclass of statements, e.g. "volleyball subclass of ball sport", "volleyball subclass of team sport", "volleyball subclass of ball-over-net sport", "volleyball subclass of non-contact sport", etc. This is a bad smell. Instead of cramming everything into instance of or subclass of, domain-specific properties like "equipment used", "setting", "team or individual", "contact allowed" would allow for more precise and expressive modeling. In general, items should have 1 or perhaps a few subclass of or instance of to avoid treating them as catch-all properties. This is a loose corollary of the software design principle to "favor composition over inheritance". Emw (talk) 03:59, 15 July 2014 (UTC)
@Emw: Not as flexible as Manchester notation, but we could do something like:
--Micru (talk) 07:50, 15 July 2014 (UTC)
Micru, unsurprisingly, I think something closer to Semantic Web standards would be best. The limited UI makes that difficult, but the closer we can align to W3C standards like OWL and Manchester syntax the better.
For example, the "type of statement" qualifier prototyped in the links above casts ∀ as "mandatory statement" and ∃ as "one of these". I would prefer to use "only" for ∀ and "some" for ∃, as Manchester syntax does. The Manchester OWL Syntax has a great overview, including design considerations and implementations in popular ontology editors like Protege. The Protege UIs shown there would be a good thing to work towards, but, of course, developing a suitable UI for complex statements would likely take over a year. Qualifiers are probably the best option we have.
I was originally inclined to suggest using the generic applies to qualifier for this. However, section 1.2 of the paper linked above describes a common misreading of quantifiers that we should be careful to avoid (and not facilitate):
Coupled with the problem of containing cryptic symbols, the syntax for restrictions is a prefix syntax. That is, the restriction quantifier precedes the role/property name and optional filler. It has has been observed that this can lead users to initially read restrictions incorrectly. For example, many users initially read, ∃ hasTopping MozzarellaTopping as, "some pizzas have toppings that are mozzarella topping", compared with the correct reading, “all pizzas have toppings that are some mozzarella topping”.
The Manchester OWL Syntax, Horridge et al., page 2
Design patterns like ONLYSOME and an XOR operator (which, in hindsight, would the correct thing to use for the biocide example above, not OR) are described on page 7. We might be able to capture very simple universal and existential quantifications with qualifiers, but more complex and informative statements will likely take Wikidata development resources. Emw (talk) 12:38, 15 July 2014 (UTC)
@Emw: In the same way that perfect is the enemy of good, technicalities are the enemy of usability. Whenever there is the choice between a "perfect OWL compatible solution" and a "usable solution in Wikidata", we should be biased towards the second, because there is no point in having something perfect if our colleagues cannot figure out how to use it.
Anyhow, since more elaborate solutions are not possible at the moment, let's focus in the qualifier needed first, and later on we can discuss which items and which labels can capture the possible cases in an easy way.--Micru (talk) 13:06, 15 July 2014 (UTC)

@Delusion23, Emw, ValterVB, Micru: It's a subclass of none of them if both classes are disjoint. Following ValterVB, there could be

  • "biocide (c)" subclass of chemical substance and subclass of biocide
  • "biocide (m)" subclass of microorganism and subclass of biocide

if there is an intersection

  • "biocide (cm)" subclass of chemical substance, subclass of microorganism and subclass of biocide

The other solution is to tag items as biocide and of the other two.

I just change isotope [1]. Tamawashi (talk) 11:14, 15 July 2014 (UTC)

@Tamawashi: It's unrelated and you change does not make any sense to me. What are the instances of <isotope> if isotope is a class ? Just name one.
< isotope > subclass of (P279) miga < atom >
would mean "the first oxygen atom ever is an isotope instance". But an isotope ... of what ??? TomT0m (talk) 11:29, 15 July 2014 (UTC)
@TomT0m: "What are the instances of <isotope> if isotope is a class ? Just name one." - This makes no sense. 1) Instances don't necessarily have a name. 2) Classes don't necessarily have instances. Tamawashi (talk) 11:36, 15 July 2014 (UTC)
The statement "isotope subclass of atom" is correct. I agree with Tamawashi on that. Also in agreement is ChEBI, the reference ontology for chemicals: see e.g. oxygen-18 in ChEBI, which is marked as a subclass of ("is a") oxygen atom. Emw (talk) 12:45, 15 July 2014 (UTC)
@Emw: Absolutely not ! he modified the general isotope concept, not any isotope class ! You don't understand the difference ? TomT0m (talk) 15:33, 15 July 2014 (UTC)
TomT0m, Tamawashi, all instances of isotope are also instances of atom, thus "isotope subclass atom" is indeed correct. However, all instances of atom are also instances of isotope, so "atom subclass of isotope" also holds. Thus atom and isotope are equivalent classes. (For those unfamiliar with OWL, "A subclass of B" and "B subclass of A" entails "A = B". See the note at the bottom of the section on owl:equivalentClass for reference.) Similarly, oxygen and isotope of oxygen would be equivalent classes. If you disagree, then please suggest an example of a subclass or instance of atom that is not also one of isotope, or vice versa, or a subclass or instance of oxygen that is not also one of isotope of oxygen, or vice versa.
Thus, a better way to model isotopes is as the ChEBI ontology does, for example "oxygen-18 subclass of oxygen". ChEBI notably does not state "oxygen-18 subclass of isotope", but I think would could reasonably diverge from ChEBI here, which would enable retrieving the result set "isotope of oxygen" by querying "x rdfs:subClassOf oxygen AND x rdfs:subClassOf isotope". Emw (talk) 02:03, 16 July 2014 (UTC)
@Tamawashi: If a class has no instances, then it's equivalent to the bottom concept, whose only interest in a model is to be the bottom concept. The empty class, nothing, said differently. <Isotope> is not the bottom concept, it is a class of concept whose <Oxygen-18> is an instance of. But I'm REALLY tired to repeat that without having a convincing anwer. TomT0m (talk) 15:33, 15 July 2014 (UTC)
That is not correct, TomT0m. The most widely-cited ontology in the world, Gene Ontology (GO), has no explicitly-declared instances, but none of its classes are nothing. See for yourself: download Protege, open the GO ontology from the URL http://www.geneontology.org/ontology/go.owl, go into the 'Entities' tab, and then the 'Individuals by Type' tab. You will see no instances; there are no instances defined for any classes in Gene Ontology.
In fact, none of the ontologies in the OBO Foundry -- which GO is an example of, and which together are probably the main driver of OWL development -- have any instances. This derives from observation that:
The ontologies in OBO are designed to serve as controlled vocabularies for expressing the results of biological science. Sentences of the form 'A relation B' (where 'A' and 'B' are terms in a biological ontology and 'relation' stands in for 'part_of' or some similar expression) can thus be conceived as expressing general statements about the corresponding biological classes or types. Assertions about corresponding instances or tokens (for example about the mass of this particular specimen in this particular Petri dish), while indispensable to biological research, do not belong to the general statements of biological science and thus they fall outside the scope of OBO and similar ontologies as these are presented to the user as finished products.
Relations in Biomedical Ontologies, Barry Smith et al., 2005
This is likely what Tamawashi meant in saying "Classes don't necessarily have instances." Subclasses of an isotope like oxygen-18 obviously have instances -- e.g. an atom of oxygen-18 floating in your room -- but those instances fall outside the scope of Wikidata. Like any of those OBO ontologies used throughout biomedical research, the fact that Wikidata does not have any items about instances of oxygen-18 does not mean that oxygen-18, being a leaf node in a concept hierarchy, needs an instance of statement to avoid being inferred as nothing, i.e. \bot, the 'bottom concept' from description logic, i.e. owl:Nothing from the Web Ontology Language.
Claims like "oxygen-18 instance of isotope" are incorrect, per Semantic Web conventions as defined in OBO and other ontologies and the philosophical definitions of class (type) and instance (token) discussed in papers like Relations in Biomedical Ontologies and The Role of Foundational Relations in the Alignment of Biomedical Ontologies. Instead, we should correctly state "oxygen-18 subclass of isotope". Emw (talk) 03:39, 16 July 2014 (UTC)
@Emw: I still think you are dead wrong on this, and just one analysis of your query "x rdfs:subClassOf oxygen AND x rdfs:subClassOf isotope" to prove this : let's say isotope is an equivalent class of atom Then a reasoner could substitute the query to "x rdfs:subClassOf oxygen AND x rdfs:subClassOf atom". Then as any instance of oxygen is also an instance of atom, then the second member of your query is redundant, and your query is x rdfs:subClassOf oxygen. Which by the way can get any subclass of oxygen, not just isotopes.
On the classes with no instances : I was talking of instances in the real world, not of instances in this database.
Claims like "oxygen-18 instance of isotope" are incorrect They are useful, and I think saying CheBi does not do this but I think we can makes you out of the line to say this. Your source does not do this for a reason, you can't make them say something they do not do, and then because of something they did not do make me say something incorrect.
The discussion with pure substance of the other day were more interesting :) TomT0m (talk) 06:49, 16 July 2014 (UTC)

TomT0m, the claim "oxygen-18 instance of isotope" is incorrect not because ChEBI doesn't make it, but because it would contradict ChEBI and the usage of instance of and subclass of throughout all of the most widely used biomedical ontologies. The particular statement "oxygen-18 subclass of isotope" is not in ChEBI, but it would not contradict anything there.
Given statements like "oxygen-18 subclass of oxygen, isotope", the queries to do what we'd want are straightforward:
  • Get only classes like oxygen-17, oxygen-18 and oxygen-19:
SELECT ?subject WHERE { ?subject rdfs:subClassOf oxygen . ?subject rdfs:subClassOf isotope . }
  • Get all subclasses of oxygen and isotope:
SELECT ?subject WHERE { ?subject rdfs:subClassOf* oxygen . ?subject rdfs:subClassOf* isotope . }
Note that the only difference between those queries is the asterisk (*) that appears after rdfs:subClassOf. The asterisk enables computing transitive closure in SPARQL. As we see, it is very straightforward to disable inference to get the desired result set. Of course, many users presumably won't be directly writing SPARQL queries, but exposing some control in a user-friendly query UI like "Enable inferencing" would get the same result.
So we see that we can maintain compatibility with existing major ontologies and preserve a consistent meaning for subclass of while also enabling easy retrieval of things like oxygen-18. Emw (talk) 19:16, 20 July 2014 (UTC)
@Emw: You claim that we break compatibility with other ontologies, that's absolutely wrong. I don't add any inconsistency. I claim this way of doing, relying on non transitive queries is incorrect and non robust because it is non robust to inferences : the query does not work at all if we actually make the logical inferences. So a reasoner, who will make the inferences, with the same query, can not retrieve the actual isotope classes. And the whole point of beeing logically sound is to make inferences : our models have to be inference robust : you give us a query who works from your criteria without using inference as well, I ask you to provide a query that works with inferences turn, ie. with a real reasoner who will, him, make the inferences. This is a keypoint to good modeling : not rely on a set of conventions on how we built the model, but rely on the actual axioms : to a reasoner, there is no difference beetween a subclass of we actually put in Wikidata database and a subclass of relations he inferences, obviously : he can just logically safely add it to the set of facts. And if to classes are equivalent, then the set of facts he can infer grows a lot, or became smaller : he can safely forget one of the classes. Which totally breaks the query. Your query does not meet the inference robustness model criteria. And there is no point in using really strongly defined languages like OWL and carefully written ontologies if the whole point is not to use logics on them, this would be a really big contradiction in the way we build things. Your subclass of usage here does not make sense : it's so obvious isotope classes forms a metaclass : that is logically sounds, does not break any major ontology as the don't model the <isotope> concept itself ... Your Template:Oxygen-18 is not backed by an ontology, that's a fact, whereas a class of classes is no precedent, is logically sound and achieves inference robustness. It's therefore useful and a natural way to model things. 09:10, 21 July 2014 (UTC)

@Emw: With "Classes don't necessarily have instances." in my reply to User:TomT0m's "What are the instances of <isotope> if isotope is a class ? Just name one." - I meant exactly what I wrote, this had nothing to do with the scope of Wikidata, specifically all classes in the "abstract object"-tree have no instances and all instances should belong to some specific tree, probably physical object. User:TomT0m's statement seemed to imply if something is a class it must have an instance, and his later comments at least for me, had more indications of thinking that I find no proof for, e.g. ("If a class has no instances, then it's equivalent to the bottom concept, whose only interest in a model is to be the bottom concept." - philosphy is a bottom concept?). There are more funny statements: Q9121 - atom 1) instance of molecular entity 2) subclass of matter 3) part of matter, molecule. I changed [2]. What the relation to matter is (part or subclass), I have no time to decide right now. Tamawashi (talk) 10:26, 16 July 2014 (UTC)

I align with Emw's statement that Claims like "oxygen-18 instance of isotope" are incorrect. Correct would be: "oxygen-18 (one atom) instance of oxygen-18 (representation of all oxygen-18 atoms)".
I also align with TomT0m statement that isotope is NOT an equivalent class of atom. They are different generalizations of atomic occurrants.
And also Tamawashi is right, we should start from the very bottom and then build up, NOT the other way round.
FIRST we have something, then we decide what it is, and depending one its characteristics it can belong to several categories. The problem here is that we don't have an item that represents "instance of one atom that it is also the instance of one isotope of oxygen-18 that is also instance of matter". Instances are the joining point between all category trees, and not the other way around.--Micru (talk) 11:47, 16 July 2014 (UTC)

How to handle native language names?[edit]

What's the best way to handle native language names and labels? For example, we usually wouldn't use a Chinese, Cyrillic, or Greek script for an English label, so we'll transliterate it into Latin/English and use that (e.g., 李卓鑫 → Li Zhuoxin (Q6539344), Велимир Милошевић → Velimir Milošević (Q12749668), Αλέκος Λουμπαρδέας → Alekos Loumpardeas (Q16909134)). However, what should we do with the name in the original script? That's probably information we want to keep in Wikidata. I've been adding it in as an alias, because my transliteration might be wrong or someone might search for it that way, but do we have a better way to handle this or a property for it? This might be waiting for the multilingual datatype, but I figured I'd ask. Cbrown1023 talk 03:15, 15 July 2014 (UTC)

For one thing, the name in Russian should be used for the label in Russian. I would also support adding more string-type "name" properties, but that is a tricky and disputed issue. We currently only have birth name (P513). There is also surname (P734) and given name (P735), but they are of item-typed.--Zolo (talk) 07:19, 15 July 2014 (UTC)ced: Zolo (talk) 07:49, 15 July 2014 (UTC)
I use birth name (P513) for handling this. — Ayack (talk) 07:33, 15 July 2014 (UTC)
Sorry, typo, this is the onle I meant to cite. The description states "only if different from current name" which seems a bit strange. This property is monolingual while there is one label per language, so there has to be languages where the label is different from the birth name (we currently only have item-type properties for that and that seems a bit fragile). Beside I would split between name and surname. On the whole the name-properties ecosystem seems to remain incomplete. --Zolo (talk) 07:49, 15 July 2014 (UTC)
For Chinese names, I would use the Pinyin romanisation for people who are generally associated with or live primarily in the People's Republic of China (e.g. Hu Jintao (Q15029)). For people who are associated with or live primarily in the Republic of China (Taiwan), then I would use the Wade-Giles romanisation system, which is more popular there (e.g. Tsai Ing-wen (Q233984)). For people who lived in ancient China (i.e. pre-Republic era), I would use pinyin for their label and put Wade-Giles as their alias. If a person is better known under a certain scheme of romanisation, the practices prescribed above should be ignored, which is the case in Sun Yat-sen (Q8573). —Wylve (talk) 10:59, 15 July 2014 (UTC)
@Wylve: After reading your comment it seems that a qualifier "transliteration system" could be used to modify p513 (or a future monolingual property). Would it be useful?--Micru (talk) 11:17, 15 July 2014 (UTC)
Indeed that would be useful. However, we also need to account for "home-grown" transliteration systems or unofficial, yet somehow prominent, transliterations that don't follow any established system. —Wylve (talk) 11:24, 15 July 2014 (UTC)
That's not a problem, there can be an item for "custom transliteration" or just use the property and set it to "unknown value".--Micru (talk) 12:16, 15 July 2014 (UTC)

@Wylve, Cbrown1023, Micru: Maybe you start Wikidata:Romanization to document what is happening and then link this with Help:Label. ROC changed to Pinyin, so I would use Pinyin as a default for both. (en:pinyin: Pinyin, or Hanyu Pinyin, is the official phonetic system for transcribing the Mandarin pronunciations of Chinese characters into the Latin alphabet in the People's Republic of China, Republic of China (Taiwan),[1] and Singapore.) Tamawashi (talk) 11:48, 15 July 2014 (UTC)

@Tamawashi: I would prefer to have a generic Wikidata:Names and have a section about transliteration there. Btw, the property proposal is here: Wikidata:Property_proposal/Generic#transliteration_type.--Micru (talk) 12:16, 15 July 2014 (UTC)
@Tamawashi: I would argue against using pinyin for ROC. Although it was made official, it is not used by that many people. As a rule, I would follow however a person spells their romanised name. —Wylve (talk) 12:19, 15 July 2014 (UTC)
@Wylve: We can have them all, but yes, I also would set the one that the own person has chosen as "preferred rank".--Micru (talk) 12:27, 15 July 2014 (UTC)
@Wylve: "Although it was made official, it is not used by that many people." - Could that please be documented and sourced somewhere, be it Wikidata:Names, Wikidata:Proper names, Wikidata:Romanization or Wikidata:Transliteraton? So that maybe less people do what they think is right? Tamawashi (talk) 12:54, 15 July 2014 (UTC)
I am not clear about how this transliteration property would be used. It seems to work best as a qualifier of some string-type name property. But currently, the name is mostly expressed through two item-type properties surname (P734) and given name (P735).
reminder: previous related discussion at Wikidata:Requests for comment/Personal names. --Zolo (talk) 13:29, 15 July 2014 (UTC)
@Zolo: The transliteration property would work as a qualifier of any property with string or monolingual datatype. Nominative properties that use the item property, can also make use of this property by using a generic label property and qualifying it with "transliteration type".
Ideally the label system of item pages should be refurbished to link somehow the properties that contain labels and the item labels themselves. But since, as per Lydia, that is not going to happen any time soon, maybe someone can figure out a bot that copies the strings into the aliases (if needed). --Micru (talk) 14:02, 15 July 2014 (UTC)
@Micru:. Digging a bit the Tsai Ing-wen example. How do we say that here wade-giles name is Tsai Ing-wen and here pinyin name Cai Yingwen ? For the surname we could have
We could have to do the same with the given name (even though there may not be much to say about it. It is not like a traditional European given name that may have Wikipedia article because it is shared by many individuals, points to calendar-level saints etc. Essentially, it is just the string that was made up by the parents of Cai Yingwen so that it can be used as a name by their daughters)
So, with current name-properties the only solution I see to get from Tsai Ing-wen (Q233984) to pinyin = Cai Yingwen would be creating items about every name and surname and fetch the transliteration from them. --Zolo (talk) 16:36, 15 July 2014 (UTC)
Yes, that could work to some extent, perhaps using surname (P734) and given name (P735) as qualifiers of birth name (P513). I'm also thinking that a "name type" qualifier could be useful, because it would convey what to expect from the name (structure, order, etc). It is not the same to have a Spanish name (given name + (optional names) + 1st surname + (optional particles) + 2nd surname), than a w:Chinese name (surame + name).--Micru (talk) 18:05, 15 July 2014 (UTC)
surname (P734) is to link to articles about the surname which may have a number of different spellings. I would expect these to be in the various language labels and aliases.
The Official Name Proposed Property is awaiting the monolingual datatype. We felt string was not appropriate. This is for the exact name of the person or item in their own script. If the item has official names in more than one language or the name has changed over time then this property can have multiple values. I agree that unofficial (or official) transliterations of these names would also be useful and proposed a couple of properties for this but the proposal was not approved. One problem is that a transliteration property works best as a qualifier but you can't have a qualifier to a qualifier to tell you which transliteration system was used. Filceolaire (talk) 22:14, 17 July 2014 (UTC)

Forth and back conversions of items between class and instance[edit]

I changed instance to class [3] then had a look into the page history to obtain a link for my edits and ..., just on 2014-07-01 User:Humatiel did the opposite [4].

The editor did create a Wikidata:Database reports/Constraint violations/P279, since the item was target of "subclass of" since 2013-08-06 [5], created by User:TomT0m.

There are still 6000 unrooted subclasses listed at Wikidata:Database reports/Item classification. Can a bot inform editors that unrooted an item, i.e. that created a P279 or P31 constraint violation? Tamawashi (talk) 13:13, 15 July 2014 (UTC)

The difference between "subclass of"/"instance of" is confusing many users. After many discussions the conclusion has been that "instance of" represents something material of which there can be only one ("single exemplar of"), whereas "subclass" represents something abstract of which there can be many ("subcategory of").
I don't know how to make that more clear, but the current labels don't convey their meaning, otherwise there wouldn't be so many wrong uses. Any suggestion about how to convey better the difference?--Micru (talk) 13:47, 15 July 2014 (UTC)
@Tamawashi, Micru: Develop and make Help:Classification more visible ? It's a really important page imho. Even people like @Emw: seem to disagree with me on some topic like the type of the isotope class, after many discussions. I, (I speak for myself) am pretty confident I am at a point where I sorted this out and the instance/class/class of class scheme works pretty well for me and is really useful to model things in Wikidata, see many examples on Wikidata talk:Item classification of why I think this works and is useful. But we need to be really clear on that and not make that discussion over and over again. I'd like to make Help:Classification approved by community when it is ready for beginners and reasonably complete.
A constraint cannot do the job of explaining that to a user. @Micru: The abstract/concrete view is probably confusing. There is often several level of abstration : we divide concrete things like atoms in chemistry in classes of concrete things like "oxygen", who is a class of atom, and "oxygen-18", who is a class of atoms to, a class of oxygen more precisely. But we also use more abstract levels : isotopes for example regroups classes, not instances : we can say that oxygen-18 and oxygen-16 are isotope, but oxygen-18 is not a subclass of isotope. The definition of subclass is not "things that links abstract concept", it's more precise. TomT0m (talk) 06:33, 16 July 2014 (UTC)
@TomT0m: It doesn't seem that you have sorted out so well the distinction, otherwise you wouldn't be saying that "oxygen-18 is not a subclass of isotope". My reasoning is as follows:
  1. There are things that happen in nature by themselves
  2. By being in sensory contact with differentiable things (instances) we create representations of them
  3. By observing repeated occurrences, we create models of these things (classes)
Now I apply the same reasoning to "oxygen-18"
  1. There are oxygen-18 atoms that happen in nature by themselves
  2. By being in sensory contact, in this case through devices, with distinct oxygen-18 atoms (instances) we create representations of them (instance:oxygen-18 [one atom])
  3. By observing repeated occurrences, we create models of oxygen-18 (class:oxygen-18 [represents all atoms of oxygen-18 that may exist] )
Usually I see very often the inverse reasoning, specially from IT people, who put the class first and that is totally meaningless. Either we agree that if we want to model reality, reality has to come first, or we we'll be circling around this till the end of the times.--Micru (talk) 08:27, 16 July 2014 (UTC)
@Micru: : I have a discussion going on with Emw on this very page about this. See in particular about the proposition of query by Emw built with the <isotope> class and my answer on why it does not make sense : if <isotope> is a superclass of <isotope-18>, then the <atom> class and the <isotope> classes are indistinguishable for a reasoner, he can substitute one with the other. And I think I took part in pretty much every discussion about class/instance relationship and how they are implemented in the semantic web. I argumented and proposed to use the metaclass concept : <isotope> is a class of concepts whose instances are classes such as <isotope-18>, who exists as an entity in the human conceptualisation. What a human call an isotope is a class of atom. When we say oxygen-18 is an isotope, in common language, we don't say every oxygen-18 instance is an isotope. Otherwis the whole isotope concept is useless compared to the atom concept. The <isotope> concept by itself is useful because all atoms in an isotope class have the same numbers on nucleons. We have an isotope concept because this property is shared by a big numbers of important classes of atoms, we can regroup in a set : the set of all isotope classes, which is NOT the set of all atoms at all. I backed up by citing the french Wikipedia's definition of the <isotope> concepts which says exactly that : an isotope is a class of atoms who share the same numbers of nucleons. In short, in Wikidata we do not model reality only, we also model how human models reality. Which is a real thing too somehow. TomT0m (talk) 09:12, 16 July 2014 (UTC)
@TomT0m: if the "the <atom> class and the <isotope> classes are indistinguishable for a reasoner" it is not because the classing, it is because they do not contain enough information to make them distinct enough. We need to translate the definition in human language (an isotope is a class of atoms who share the same numbers of nucleons) into ontological language (for instance, like this, but it could be done better). Anyhow let's keep the atomic discussion in one place.--Micru (talk) 09:41, 16 July 2014 (UTC)
@Micru: No it cannot. The best stuff we have for this is to class not atoms, but substances, especially a pure substance is a substance made of only one type of atom. An isotopically pure substance is a substance made of only one isotope of one chemical element (as a class of atoms). But this concepts still do not map the isotope concept as all the kinds of isotopes types of all the atom types. But feel free to join the discussion on the query: here is my answer to Emw. TomT0m (talk) 10:18, 16 July 2014 (UTC)
@TomT0m: Substances (class) have atoms (class) as parts, and an isotopically pure substances (class) have atoms of the same isotope as part (class), but ok, I'll read the (now long) discussion.--Micru (talk) 10:40, 16 July 2014 (UTC)
I followed exactly the logic that Micru presented. In programming languages, it is usual that an instance being an object and a subclass being a subcategory. For example, an apple is an object/instance of fruit, and fruit is a subcategory/subclass of the kingdom Plantae (plants). That's why I changed it, since wikidata is to be human and machine readable. Regards, Humatiel (talk) 22:49, 15 July 2014 (UTC)
Humatiel, an instance it is not only an object, it is *the only* object that exists of something in the universe at a molecular level. One indeterminate apple is a "subclass of fruit", this apple that I have on my desk is an "instance of the class apple". Everything that is abstract (awards, movies, albums, concepts, etc) cannot be instances because there can be many of the same.
In Wikidata there is this confusion all over the place, even myself I had it until very recently, that is why I was suggesting to change the property labels to something that is more self-explanatory... but I don't know what.--Micru (talk) 23:14, 15 July 2014 (UTC)
@Micru: I french we have nature for instance of (P31) as a default label. I guess the subclass of in english would be better with something like special kinds of. How sounds "the nature of Micru is Wikipedian" in english, and "wikipedians are special kinds of humans" ? The plural for subclass of (P279) seem to me an idea worth investigating. Though instances of seems not really good, how reads Wikipedian instances of human (just random thought letting it going). Or <Wikipedians> are all <humans> with <Micru> example of <Wikimedian> ? (I'm not very aware of how natural sounds the instance word in everyday english. <Micru> yet another <Wikimedian> (no offense :)) ... <Apple> yet another <Fruit> works too. Hard to get rid of the is a easily substituted to are all in common language. <Apple> is a <fruit> works as well as <apples> are all fruit. <Apple> kind of <Fruit> or <Apples> are all <fruits> works, but can we add plural marks to all classes ??? (I like the idea, not sure it is a good one). TomT0m (talk) 07:40, 16 July 2014 (UTC)
I would go with "@TomT0m: <one exemplar of> human" and "apple <all exemplars are> fruit". More opinions on this?--Micru (talk) 08:27, 16 July 2014 (UTC)
@Humatiel: a fruit is part of a plant and not a subclass of the kingdom Plantae. BTW: there are some substantial differences between an object orientated programing language and OWL. --Succu (talk) 06:46, 16 July 2014 (UTC)

@Humatiel, TomT0m, Micru, Zolo, Emw: - To my understanding, in the tree "abstract object" there are no instances. And probably "physical object" is a tree that contains all instances. That means, two little queries would reveal already some errors in classification. There are still ~5900 unrooted classes listed by autolist2. If you could help here :-). Numbers are going down, but very slowly: Wikidata:Database reports/Item classification. And sometimes I found it hard to find an upward class, that is not "entity". Tamawashi (talk) 11:28, 16 July 2014 (UTC)

@Tamawashi: That is more or less correct, and it matches the formal definitions of 'continuant' and 'process' that Emw brought with this paper. I can help rooting, but in some cases it will be useless (as you have seen first hand), since some concepts are best defined in other terms.--Micru (talk) 12:03, 16 July 2014 (UTC)
There are individual things which are 'instance of' a class of things.
In some cases however a class is an instance of a 'type of class' as well as being a subclass of a larger class.
Oxygen-18 a class of millions of individual atoms each with 10 neutrons and 8 protons. This is a subclass of:oxygen atoms.
The class of Oxygen-18 atoms is also an 'instance of:isotope' - a special type of class.
The class of Oxygen atoms is on the other hand a subclass of:atom but an instance of:element - another type of class.
Note that 'isotope' is not a 'subclass of:element'. They are both 'subclass of:ways of classifying atoms'.
Other 'types of class' that classes can be an instance of:
  • Moby Dick subclass of:book, instance of:novel
  • Ford Model T subclass of:car, instance of:model.
At least that is how I see it. Filceolaire (talk) 21:53, 17 July 2014 (UTC)

"The class of Oxygen-18 atoms is also an 'instance of:isotope' - a special type of class." I do not agree with this. An instance has to be either a material thing (continuant) or a unique process in order to be considered "instance", in other words it has to be something that exists "here" and "now". All the rest are class relationships, either as "subclass of" or "superclass of", that we don't use. The relationship between "the class of all Oxygen-18 atoms" and "the class of oxygen isotope (classification of atoms)" is a subclass relationship.
I also don't agree with the book/novel, car/model examples. It was a mistake that we did on the Books task force that is spreading all over the place. It should be:
  • Moby Dick subclass of:novel
  • Ford Model T subclass of:car model
Of course the terms "subclass of"/"instance of" are confusing, that is why it would be better to change them. Possible synonyms for "subclass of": "is a", "nature", "all exemplars are", "all instances are". Possible synonyms for "instance of": "unique exemplar of", "real-life specimen of", "mass-energy occurrence of".--Micru (talk) 14:55, 18 July 2014 (UTC)
@Micru: You can't stay with an argumentation as poor as it was a mistake witout begin solid on this. You disagree with everybody one way or another with 3 people at least here : with Emw (talkcontribslogs), who could be your best friend here because he do not like metaclasses, @Filceolaire:, me ... We discussed this a lot, you will have to do better than just saying everybody else is wrong. TomT0m (talk) 13:50, 19 July 2014 (UTC)
@TomT0m: After the discussions with you, with Emw, and with others I have now a more coherent view that I had in the past, for that I am very grateful that you have helped me with great patience in our long discussions. I say that it was a mistake because the method used back then doesn't hold to deep scrutiny and makes the properties p31 and p279 indistinguishable from one another, as I have been argumenting here and on the other conversation below. Funny that you use argumentum ad populum (Q251695)... history is full of examples where that was a bad idea. For instance, that many people agreed that the Earth was the center of the universe doesn't make it more true, just means that there was an agreement and a legal body to keep that acceptance. In this context I do not believe in "right" or "wrong", just in approximations. If there is a general consensus then of course I will let it flow as an "accepted approximation", I am just pointing out that there are better ones. Please, do not take my disagreement personally since I consider you, Emw, and all others, allies in the mission of bringing the sum of all knowledge to all sentient beings. If you consider that this conversation is straining, let's stop it now. Sometimes it is just better to wait.--Micru (talk) 14:52, 19 July 2014 (UTC)
If you consider that this conversation is straining, let's stop it now. Sometimes it is just better to wait. That's why I went meta on the discussion and stopped argumentation at that point :) TomT0m (talk) 07:06, 20 July 2014 (UTC)

I agree and disagree with things Micru says above. Foremost, I think changing the labels of P31 and P279 from instance of and subclass of to something else would be a bad idea. Those labels have been established for over a year, have been used in academic publications and media coverage about Wikidata, and give an immediate, glanceable, conventional visual hook that illustrates a fundamental distinction used throughout the Semantic Web and philosophy at large to describe the nature of things. Changing their aliases is OK, changing their descriptions with care is OK, but we should not change their labels.

I agree with Micru that "Moby Dick subclass of novel" is preferable to "Moby Dick instance of novel". However, I disagree with the statement "Ford Model T subclass of car model". Cars models are like biological taxa in that they are not physically instantiated. Stating otherwise is a subtle mistake with significant consequences. For example, if we say "Chevrolet Malibu (Q287723) subclass of car model", then that entails the incorrect statement "The Peekskill Meteorite Car (Q7756463) instance of car model". The Peetskill Meteorite Car is not a car model, it is a car (specifically, an instance of a Chevrolet Malibu). Although Wikidata covers far fewer instances of car than instances of human, we should keep the application of instance of and subclass of consistent.

And that is why the statement "oxygen-18 instance of isotope" is problematic. Oxygen-18, cars, and humans are all types of material entity. Oxygen-18 is a type of isotope, car is a type of vehicle, and human is a type of Homo. Stating "oxygen-18 instance of isotope" creates an inconsistency in how we apply instance of and subclass of in cars and humans and all other physical things. The fact that Wikidata will likely never have an item about an instance of oxygen-18 does not mean we shold state that oxygen-18 is an instance. instance of (P31) is not simply what you use at the bottom of Wikidata's concept hierarchy. For at least physical things, outside rare and straightforward usages of explicit metamodeling, instance of should only be used on items that have a unique location in space and time. Emw (talk) 15:34, 19 July 2014 (UTC)

@Emw: I am glad that we are approaching postures. You might be right that it might no longer possible to change the labels of p31 and p279 without causing major disruption that we might want to avoid, but it worries me that some languages have departed from the canonical names and now they have come to represent different things (in German p31 is now "is a"), perhaps it will be enough making sure there is no confusion by explaining it better, or creating a tutorial.
Just a minor correction regarding your car example. You say: if we say "Chevrolet Malibu (Q287723) subclass of car model", then that entails the incorrect statement "The Peekskill Meteorite Car (Q7756463) instance of car model". I think we should examine deeper what a "car model" is. If we assume that a car model is a blue print for manufacturing an automobile with shared characteristics, then the statement "The Peekskill Meteorite Car (Q7756463) instance of car model" is not incorrect, because we have a blueprint and from there we generate something that is an automobile. Of course the statement is less accurate than saying "The Peekskill Meteorite Car (Q7756463) instance of Chevrolet Malibu (Q287723)", but still right. Same case as the example you gave me "Pluto (Q339)<part of>universe (Q1)", which is not wrong, just "less right".--Micru (talk) 16:27, 19 July 2014 (UTC)
@Micru: : this depends on the definition we use for car model. Of course we can take a car, scan it and make a copy with a 3D printer, and our copy has a model : the original car. But it's not the common definition car manufacturers uses for car model. We can define the <car model> item such that its correct wikidata use is a car model in the car manufacturing industry sense, and we get rid of this correct in a loose sense notion that is ... Uncommon. Of course if there is a use for this definition we can create an item with this definition, but let's be precise and say that car that are 3D printed with some scanned car and car model in the car industry sense are different item. The vast majority of car models in Wikidata are on the second sense. That's indeed a compelling case for class classification. TomT0m (talk) 07:52, 20 July 2014 (UTC)
@TomT0m: What is a car model in the car industry sense? Put the definition on the operation table so we can dissect it ;) --Micru (talk) 09:57, 20 July 2014 (UTC)
@emw:. My understanding is that O-18 is general would be considered an isotope of oxygen but an individual molecule of O18 would not. That sounds remarkably similar to the Ford T case.
As stated in #Issue with "instance of" for text, I would argue for a similar analysis with novels. Novel is a subclass of text, and a text is, roughly speaking, an (immaterial) string of words and Moby Dick is best viewed as an instance of novel starting with "Call me Ishmael." It can be materialized in a book, an audio recording whatever, but those are instances of book, audio recording etc, not instances of novel. Of course defining a text as a string is a very crude approximation, but a definition of text would have to do with the us of words, the existence of a plot, etc. These are not features that a material object can have It implies that the class "text" is disjoint from the class "material object".--Zolo (talk) 17:35, 19 July 2014 (UTC)--Micru (talk) 14:30, 20 July 2014 (UTC)
Agreed with Zolo on the similarity with the <car model> and <isotope> discussions, that's why the instance/class/metaclass modeing 3 layer scheme I push other and other again make sense. "oxygen-18 instance of (P31) isotope" creates no inconsistency if we make clear that isotope IS NOT a class of physical objects itself, which can be made explicit in the ontology by a disjoint with statement beetween the <physical entities> class and the <class> class, so it is absolutely not a logical inconsistency. It's not an inconsistency with the Wikidata usage if we stop pushing the Token/class philosophical principle to newbies and start pushing a Token/class/metaclass principle which actually helps a lot into understanding how to model the world. Using subclass of (P279) as the only relationship beetween classes of tokens (I don't say instances intentionally) is a mistake and logical inconsistency source and should as a result not be pushed to users. So, Emw, I think you should revise your way of thinking and starts thinking not car models as weird exceptions, but as the general case. When a scientific theory starts to have one exception scientists can be fine saying OK, that's a weird exception. When there is a second, a third and so on they'll start thinking they need a Paradigm shift (Q689971) (View with Reasonator) (paradigm shift) in their theories. I think token/class principle is such a paradigm who put our mind in a bottle, but everything is simpler if we break that bottle and shift to the token/class/metaclass paradigm. I provided several usecases for which this paradigm is useful, for me it's a compelling argument it's a solid paradigm, and more consistent than the "token/class with exception" you are proposing, who is unsatisfactory : if we got a 3 concepts principle which fits almost all wikidata, wrt. a 2 concept principle with 30 exception, the first has 3 concepts, the second 32 concepts. Ockham rasors says "the first one is better". TomT0m (talk) 07:52, 20 July 2014 (UTC)

Zolo, an actual atom of oxygen-18 is an instance of an isotope. Instances of isotopes is what is measured in e.g. isotope-ratio mass spectrometry. Thus the statement "oxygen-18 instance of isotope" is indeed inconsistent with how we use that property to describe things like humans, cars, and other physical things. It is like saying "human instance of Homo", not "human instance of taxon".
Some domains talk of types of types of things. Biological taxonomy is one, but chemistry does so less, certainly in the way TomT0m envisions:
Which is easier for users to understand, the "atom type class" model proposed above or the one without a metamodeling layer, as used in widely-cited, professionally curated chemistry ontologies like ChEBI? I do not think pervasive explicit metamodeling will ameliorate users' unfamiliarity with instances and classes. I think it will bewilder them.
Instance of and subclass of can be used for explicit metamodeling in a way that accommodates type-token distinction. The usage seen in biological taxonomy is an example. However, it is easy to do metamodeling in a way that breaks that useful distinction, as illustrated in statements like "hydrogen instance of chemistry element" and "oxygen-18 instance of isotope."
Doing so compels us to make absurd statements like "isotope is not a class of physical object" as TomT0m does. As any chemist or high school student who has completed a course in chemistry will tell you, isotopes are physical objects. Stating "oxygen-18 subclass of isotope" rather than "oxygen-18 instance of isotope" consistently applies classification properties in a way that does not compel us to paraphrase away basic scientific facts. Emw (talk) 14:13, 20 July 2014 (UTC)
@Emw: I share your opinions to some extent. I think we should not use "instance of" for cross-domain relationships, but we can attempt to cross it with a different property, even if that departs from ChEBI. The way I envision it is like this:
I don't think using metaclasses it makes it more complicated if it is explained properly (i.e. not using technical jargon).--Micru (talk) 16:27, 20 July 2014 (UTC)
Micru, I think it is essential for Wikidata to use vocabulary from the wider Semantic Web. The connection between Wikidata, the Semantic Web and ontology is that "class" is synonymous with "type" or "concept", as outlined in e.g. The Role of Foundational Relations in the Alignment of Biomedical Ontologies. The conventional term for "classes of classes" like "taxon" or "car model" is metaclass. Implying that classes are not types or concepts as your graphic above does would misalign Wikidata with most literature relating Semantic Web vocabulary to that from ontology. Styling instances as "phenomena" and classes as "substance" is also highly idiosyncratic and would lead to even obscurer philosophical wanderings than those we currently see on Wikidata.
Metamodeling in the domain of chemical elements seems generally extraneous and unhelpful. What makes "atom" a class and "isotope" and "element" metaclasses? How do "chemical substance" and "physical object" fit in; are they class or metaclasses? As explained above in my reply to Zolo, isotope is clearly a class of physical object and individual atoms are also isotopes. Why should I have to include nature of as outlined in your graphic (note: as presented it should be has nature) rather than use subclass of? What is the precise definition of has nature -- in terms used by the Semantic Web, i.e. OWL? Precisely how would it relate to P31 (i.e. instance of, rdf:type) and P279 (i.e. subclass of, rdfs:subClassOf)?
I think it would be far simpler to say something like "oxygen-18 subclass of isotope of oxygen" for such classes of isotope. We could then say "isotope of oxygen subclass of oxygen, isotope". Alternatively, we could say "oxygen-18 subclass of oxygen, isotope", and consider "isotope of oxygen" to effectively be a query result. Both of these approaches adequately model isotopes in a way that is consistent with major existing ontologies and does not pose compatibility issues with Semantic Web standards. Emw (talk) 18:09, 20 July 2014 (UTC)
Brilliant paper, Emw! It brings many insights and I dearly recommend Zolo and TomT0m to read it. I will do it again, and again. On a first read I found some gems:
  • inst: defined as bound to a specific time and space. An instance can be related to its class (a instanceOf b)
  • class: defined by its instances, which means that they can change in time. A class can be related to its upper class (b subclassOf c)
  • set: a timeless entity, idealized class, not bound to time and space. A set or a class can be related to a set (b is_a d)
The definition of what I called "nature" is what they call "is_a". Please, do notice the differences with "subclassOf" as cleverly explained in the section "Classes vs. Sets". I can go further and say that "is_a" is equivalent to "type" (well, they also say it in the paper), and "taxon rank" is a subproperty of "is a".
Coming back to my drawing, yes, you are right, it is wrong, I'll try again:
That is my interpretation of the provided paper, with the nomenclature stated in the paper. True that for atoms it looks quite redundant (although here we wouldn't use classes), however it models accurately the relationship between classes-metaclasses and between metaclasses themselves. To use subclassOf for both purposes is missleading and brings to many confussions and long discussions as we are seeing. I hope we can all reach the same interpretation and consider it both acceptable and practical.--Micru (talk) 23:18, 20 July 2014 (UTC)

Micru, that paper formally defines is a, a foundational property in major ontologies like the Gene Ontology (GO), ChEBI and others. It is referenced as the basis for is a and part of as discussed in Relations in Biomedical Ontologies, a very widely-cited, classic paper in modern ontology. That paper establishes the link between "is a" and the language of the Semantic Web: "A has_subclass B = [definition] B is_a A." -- in other words, "A rdfs:subClassOf B" equals "A is a B". This equivalence between BFO/OBO's is a and RDF/OWL's rdfs:subClassOf is noted in the draft BFO 2.0 release documentation and clear from RDF/OWL exports of GO, ChEBI and other ontologies among the Open Biomedical Ontologies.
The relation between classes and sets is a recurring topic in these discussions on Wikidata. Classes are treated essentially as sets in the Semantic Web. The foundational subclass of property is even symbolically represented with the symbol as "subset of" -- i.e. ⊆ -- in the Interpretation of Axioms and Facts in the OWL Direct Model-Theoretic Semantics specification. The relevant OWL specifications address this by noting that a class is distinct from the "extension" of that class.
It is interesting how your subsumption hierarchy (the "class tree") of atom in your interpretation of the FMA paper is virtually identical to that used in ChEBI. See e.g. the "graph view" visualization in the ChEBI model of oxygen-18 at http://www.ebi.ac.uk/chebi/chebiOntology.do?chebiId=CHEBI:33815&treeView=false#vizualisation. Of course there is no mirrored subset tree, because is a is equivalent to subclass of.
The ChEBI ontology does not explicitly account for some things we may want to, e.g. isotopes. One suspicion is that this is because all instances of atom are also instances of isotope, and vice versa. However, we might be able to model the notion that "oxygen" itself is not an isotope -- while preserving consistency with ChEBI and those other widely used ontologies -- with an approach I outline above. Emw (talk) 01:35, 21 July 2014 (UTC)
@Micru: Thanks for the link. Some notes:
  • It states that nothing can be both a class and an instance, and that conflicts with the latest OWL version. The current theoretical view appears to be that we should distinguished between the statements about the item seen as an instance of X and those of the item seen as a subclass (seen as a instance of species, gorilla beringei is an endangered species, seen as an subclass of animal, it is hairy).
  • Nothing in the theoretical part of the article says that an instance must be localizable in time and space. They indeed allude that they should in the conclusion, and for highly empirical topics like biology, it is certainly true that most instances have a location, but that would be an empirical fact, or a methodoligical advice, not a formal imperative (yes
  • You are misreading the paper, it does not say that classes are defined by its instances, quite the opposite: contrary to a set, a class "survives the turnover in its instances".
  • I like their idea that contrary to a set, a class does not have 2n subclasses (actually it think we can say that "subset" is a subclass of "class" with some special features, except that it would sound a bit odd to include the empty set and singletons in the list of subclasses). --Zolo (talk) 07:56, 21 July 2014 (UTC)
@Emw:. Ok, if individual molecules are isotopes, then clearly it is fine to call 018 and subclass of isotope. --07:56, 21 July 2014 (UTC)~
@Emw: first of all some term equivalences:
* what OWL calls "extension of a class" = what FMA calls "class"
* what OWL calls "class" = what FMA calls "set"
For the particular case of chemical elements I agree with you that there is an "almost" 1:1 relationship between class-set (real atom vs conceptual atom), but not making this distinction is problematic for several reasons. The main one is the superposition of "natures" that an element has, which collapse into one as soon as we try to access experimentally (adding instances to) one of them (cf. w:wave function collapse). Only by avoiding to use the "isotope" set, they are able to keep an equivalence between "class" and "set". It is not a lie, they just don't consider all the range of possible truths.
In our particular case, it would be perfectly fine to model chemical elements considering an equivalence between subclassOf = is_a because we don't instantiate them. For the classes that we do instantiate then metamodelling (making a distinction between class and set) is useful (taxonomy, cities, literary works, etc). But that of course forces us to juggle with properties, when by using two distinct properties (FMA-subclassOf, FMA-is_a), would remove all obscurity from our model.
The other reason to make the distinction is to avoid mixing up "extension of a class" and "class". If we use OWL-subclassOf for both cases we might arrive to the weird conclusion that a person might have the permanent nature of being an actor (!). That means that we accept instantiating permanent concepts as impermanent instances, with all the inconsistencies and headaches that it brings.
For these reasons my recommendation is to abide to the FMA definitions (instanceOf=phenomena, subclassOf=OWL-"extension of a class", is_a=OWL-"class"). This will guarantee compatibility with all existing and yet to exist ontologies, because we'll be able to cast both subclassOf and is_a to the same external ontology property (if they have so defined), or to distinct properties (if they have so defined).
@Zolo: They refer to "SNAP and SPAN: Towards Dynamic Spatial Ontology". Basically: SNAP=continuants (permanent) and SPAN=occurrants (impermanent). According to this model, yes, you can instantiate a permanent class (set), but not as an impermanent entity!--Micru (talk) 08:26, 21 July 2014 (UTC)
@Micru: That is true, but that seems to work well as long as we do not need to assume that everything is a SPAN and SNAP.
To keep things in plain language:
  • entity:
    • unlocalizable entity (concept)
    • localizable entity
      • event
      • material object.--Zolo (talk) 10:03, 21 July 2014 (UTC)

──────────────────────────────────────────────────────────────────────────────────────────────────── @Zolo: Your tree puts again together sets and classes (the "entity" you put on top cannot be both "abstract" and "real"), and as showed before that is problematic. Here there is another one:

  • unlocalizable entity (concept, set)
    • qualities: physically identifiable characteristics (shape, color, etc)
    • attributes: abstractedly identifiable characteristics (the "is_a"-ness)
    • entity: the concept that something might exist either as unlocalizable or as localizable entity
  • localizable entity (phenomenon, instance)
    • event
    • material object

@Emw: I'm sorry that I'm not following OWL conventions :( I realize that the definition of entity is self-referential... this is not my personal preference, it is just how nature appears to me... very weird... :-/ I also realize that there are more "unlocalizable entities" that only apply to living beings, but it is better to focus on this first.--Micru (talk) 12:57, 21 July 2014 (UTC)

Micru, is a is equivalent to subclass of, i.e. rdfs:subClassOf. The lead author on that paper you like is also the lead author on Relations in Biomedical Ontologies, and also the head of the team that produces the BFO, which uses that equivalence. Please take another at the first paragraph in my previous comment where I describe how is a equals subclass of in that major school of formal ontology. Please review the BFO class hierarchy, which accounts for the relation of 'entity' to 'material object' and 'qualities'. Let's use conventions from the Semantic Web and not add yet another upper ontology to it. Emw (talk) 12:55, 22 July 2014 (UTC)
@Emw: Which property shall we use to express "extension of a class"? And for metaclasses?--Micru (talk) 15:28, 22 July 2014 (UTC)
Micru, subclass of (P279), i.e. rdfs:subClassOf. Here's the definition in OWL:
The rdfs:subClassOf construct is defined as part of RDF Schema [RDF Vocabulary]. Its meaning in OWL is exactly the same: if the class description C1 is defined as a subclass of class description C2, then the set of individuals in the class extension of C1 should be a subset of the set of individuals in the class extension of C2. A class is by definition a subclass of itself (as the subset may be the complete set).
And reviewing the definition of classes in OWL:
In general classes are used to group individuals that have something in common in order to refer to them. Hence, classes essentially represent sets of individuals. In modeling, classes are often used to denote the set of objects comprised by a concept of human thinking, like the concept person or the concept woman.
The equivalence between BFO's is a relation and OWL's subclass of is used, for example, in the chemistry ontology ChEBI:
All database entries are now 'is_a' classified within the ontology, meaning that all of the chemicals are available to semantic reasoning tools that harness the classification hierarchy... To comply with our goal of increasing interoperability with other ontologies in the biomedical domain, ChEBI has provided a mapping to the upper level ontology Basic Formal Ontology (BFO) (12), version 2.0.
Feel free to verify this by downloading any of the OWL files in ftp://ftp.ebi.ac.uk/pub/databases/chebi/ontology/ and seeing the substitution of all usages of is a used through the ChEBI OBO files with rdfs:subClassOf.
Reviewing further, Role of Foundational Relations formally defines is a, and Relations in Biomedical Ontologies harmonizes that definition with Semantic Web standards from the W3C by clearly defining the equivalence of is a and subclass of. The equivalence is made explicit in the BFO 2.0 release notes. The three aforementioned works are led by the same individual and the latter two are produced by the same group. The equivalence is used throughout the world's most widely used biomedical ontologies.
Nowhere in any of those works is the discrepancy you assert between is a and subclass of alluded to. The ontologies in question that use the equivalence of is a and subclass of in practice, Gene Ontology and ChEBI, are maintained by content experts with advanced degrees and deep experience in chemistry, computer science, philosophy and ontology. If your concern with the discrepancy between sets and classes because of wave function collapse were warranted, don't you think they they would have noticed that, and raised that issue somewhere in a publication? Do you really think that's why they don't explicitly have an "isotope" concept in ChEBI? I suspect the cause was much more mundane. A discrepancy between is a and subclass of would directly contradict what that network of researchers have clearly stated in multiple major publications and spent over a decade building. Emw (talk) 04:29, 23 July 2014 (UTC)
@Emw: the last message by TomT0m about intensional vs. extensional definitions clarifies quite well the class-set duality paradox that was besetting me. Seen under that light, yes, I agree on the equivalency between subclassOf and is_a, not because any researcher or produced bible says so, but because it is possible to reconcile that duality the same way it is done in quantum physics. Any entity has thus an "existential state:indeterminate", that can collapse into "determinate existence" -when seen from the perspective of its instances (extensional view), it can represent "extension of that class"-, or into "determinate non-existence" -when seen from the perspective of its abstract meaning (intensional view) it can represent an ideal set, non-localizable in time or space. It is not necessary to declare this explicitly, because the presence or absence of instances already serves that purpose.
Do you still hold the opinion that aligning our vocabulary with CHEBI, BFO and other major ontologies by changing the label of p:p279 to "is a" would be a bad idea? As TomT0m noted, it feels more natural to to write "<isotope> is a <nuclide>" than with the current label.
About CHEBI and their lack of isotopes, why don't we stop our elucubrations and just ask them about it? :) --Micru (talk) 07:47, 23 July 2014 (UTC)

@Zolo, TomT0m: Continuing expanding the reasoning that I was exposing below, I realize that it is very similar to the 3 layer model that TomT0m is talking about:

  1. there is an idealized class of something, that at the same time is an instance of itself (see my example with the natural number "5" below)
  2. there is a class of all real occurrences of that class
  3. there are instances of the class of the real occurrences

So applying it to the case of "a novel"

  1. there is the idealized class of a novel, and the class is abstractedly equal to their instances (all thoughts about a "novel" are both a class and an instance, even if they happen with different material substrate)
  2. there is a class of all real occurrences of the class abstract novel
  3. there are instances of the class of real occurrences of the former

Now the problem is to find what is the relationship between 1-2 (I hope we can all agree that the relationship between 2-3 is clearly class-instance). For 1-2 is not that clear, because it seems that this relationship is neither of instantiation, nor of classing, but of abstraction. 2 is an materialization of 1, 1 is an abstraction of 2. It would be interesting to see if there are precedents of these kind of relationship in the literature (pun intended :P).--Micru (talk) 11:34, 20 July 2014 (UTC)zz

I don't follow you, and I think numbers are a bad example: they are already exists as a datatype in Wikidata, so I think we are fine using properties like length search or cardinality to use them. Otherwise clearly we are fine with the set membership/subset relationships for classes here : Natural numbers are a subset of numbers in the most general sense, "5" is a member of that set, so I'd be happy with
< number 5 > instance of (P31) miga < natural number >
for Wikidata, and
< Natural numbers > subclass of (P279) miga < numbers >
. I don't think it's interesting to go deeper here and enter philosophical or math foundation like number set construction problems.
Otherwise I think it's interesting to keep with instance of (P31) for linking classes to metaclasses because it is user friendly imho : in french, instance of (P31) is labelled nature (by the way Emw the english labels are not really supposed to be a reference here so I'm fine with changing the english labels, there is aliases and definition for precisions) and it's pretty intuitive to say that the nature of Porshe 911 is car model imho. It's a cool entry point, not boring the user with obscure philosophical and maybe even confusing considerations.
It would be interesting to see if there are precedents of these kind of relationship in the literature (pun intended :P) I don't see the pun, we're building something here, it has to make sense, that's all, not following absurdely some sourcable rule that may fit less well our needs. We are taking great care of standards in our discussions and are pretty standards friendly nonetheless. TomT0m (talk) 13:28, 20 July 2014 (UTC)
@TomT0m: The pun was because I was giving an example with a "novel", which is also literature... but yes, I agree with your statement, we have to be consistent mainly internally, because if we try to follow every source, the most we can aspire to achieve is inconsistency.
Ok, so we have metaclasses that are abstract concepts, classes and instances, and your proposal is to use p31 to link both metaclasses-classes and classes-instances... wouldn't that put metaclasses and classes in the same tree? It seems a bit like using "part of" or "subclass of" for everything... valid, but doesn't allow us to do consistency checks as the ones Tamawashi is doing. On the other hand, I like a lot the label "nature" for the relationship between classes and metaclasses, but I would prefer another for the relationship between classes-instances, perhaps "occurrent of", "phenomenon of" or even keep "instance of"?
If numbers are such a bad example, why do you bring it from another subsection? :)--Micru (talk) 14:30, 20 July 2014 (UTC)
Micru, "occurrent" is a direct subclass of entity in BFO, the upper ontology used in the world's most cited ontologies. It translates roughly to "event"; it is disjoint with "continuant", which includes things like material and immaterial entities. A complete rendering of BFO's class hierarchy is shown here. Thus changing instance of to occurrent of wouldn't be ideal. Changing it to phenomenon of would make Wikidata very idiosyncratic in the Semantic Web. The label instance of makes it clear that that property is based on Semantic Web conventions and the description logic that undergirds it. These disciplines use one property for instantiation -- instance-class classification -- and one property for subsumption -- class-class classification. It also makes it clear that the subject is an instance, whether that instance is an occurrent, a phenomenon, a human or a car. Emw (talk) 15:00, 20 July 2014 (UTC)

@Emw, Micru:

I lost some part of the discussion, but I'm not sure we are fine with some useful definitions of set theory : the so called set intentional definition concept : a set is defined by a property of their instances. This is an equivalent to class expressions in OWL : a class is then defined not by its instances, but by properties of its instances, see Intensional definition (Q1026899) (View with Reasonator) .

This notion is useful to understand my proposition (which is not an invention of mine btw, it's a used definition) because it fits : clearly <Oxygen-18> and <Oxygen> are both classes of atoms. But what makes one an element-class and the other an isotope-class is that we use different criteria in the intensional definition : in the element classes like <Oxygen> or <Hydrogen>, we use the property atomic number of atoms in the intensianal definition formula (class expression in OWL), and in the other one we use the number of nucleons of the atom. Now imagine we want to class, as in automatic classification in computing with learning machines or other mechanisms, ie. define a set of classes and put every class expression into some of those classes, clearly the chemical element class expression and isotope class expression definitions would become very handy and have a lot of sense. Imagine, in the sense of OWL punning, the classes item (<Oxygen-18> or <Oxygen>) becomes some kind of reification ( the senses on computer science for example, or in knowledge representation, I'm also aware of uses in logic), in some sense, is the act to treat something abstract like the other object in a model to be manipulated as the other basic objects of the model).

Then we got something very solid here : metaclasses are classes of reified (OWL2 allows this) classed using properties of their intensional definitions (as classes). We're fine with the class/token principle if we don't forget that reifications are not real world objects. It's natural to a lot of people (ie. this is exactly how the <isotope> concept is defined by french Wikipedia : an isotope is a nuclide (Q108149) (View with Reasonator) , nuclide is a type of atom. (seems the article have been rewritten since I read it the last time, it did not have the nuclide concept in it ! By the way @Emw:, were you aware of this concept ? How does this fits in your views ? It seems to me that it is exactly the atom type class concept I proposed ! Once more something I did not know but prove my reasoning solid, there is a word for this concept. So Wikidata has to be able to deal with this. Just merged the items.) I don't have time to rewrite all, but feels like a victory to me. TomT0m (talk) 17:00, 22 July 2014 (UTC)

GSoC '14- Wikidata Web Annotator : Final Testing Phase, Bug and Improvements[edit]

This is an update report for Wikidata Web Annotator under GSoC'14 I have almost made everything ready for testing. I have setup the test site here. First login using OpenID to authenticate with the Pundit software. Try selecting some text, then use annotate text option, futher use Pundit's triple composer to compose triples using selectors and vocabularies retrieved from Wikidata. Subject and Object are to filled with Wikidata Items, while Predicates with properties. Check for notebooks(in Pundit notebooks are the containers for annotations) in right most corner of the plugin by clicking on your name > selecting manage notebooks. Notebooks must be public for push to work. Now after annotating use Push to Wikidata button as shown in the figure below

Push to Wikidata Button

​ This will open a new window where you can select which annotations are to be pushed. You must authenticate with Wikimedia before doing that as Wikidata is fed by using your Wikimedia username. After this select the annotations you want to push and the click on the 'Push annotations' button, status column will show you about the progress of the push.

I have tested it but not completely. I have checked for most of the cases like: 1. If claim exists-> then try to push references only. 2. If not create claim -> then push references. These are just example of cases I have checked upon, there is much more. You all are welcome to test the Annotator so we can further move on to package it as plugin. Discuss the issues, bugs etc. Also test page may take time to load.

Workflow

Example page which has been fed with this annotator is https://www.wikidata.org/wiki/Q2336535. In this item, claim for P277 has been wholly added with references via Annotator. While claim for P106 was already existing and I have added references to it via Annotator.

Here is flowchart showing the workflow: ​ For closer look check here http://www.gliffy.com/go/publish/5918718

Repository for all these can be found here 1. Main source including modified Pundit : https://github.com/apsdehal/WikidataAnnotationFeeder 2. Login and API handler: https://github.com/apsdehal/WAL 3. Middleware App: https://github.com/apsdehal/Bajo

Please feel free to add issues or Pull requests there. --Apsdehal (talk) 14:14, 15 July 2014 (UTC)

Maybe I have missing something but Pundit (Q2336535) is "indigenous surveyors who explored regions to the north of India for the British..." and you have added programming language (P277). Why? --ValterVB (talk) 18:24, 15 July 2014 (UTC) ps URL is breack.
@ValterVB: I think it is because Apsdehal was using Q2336535 as a sandbox (see item talk page).--Micru (talk) 08:48, 16 July 2014 (UTC)

Interwiki links in Arbic wikipedia and Commons[edit]

Hello.There are Interwiki links in Arbic wikipedia and Commons (example).Who can make a global bot to expulsion these links to here? --ديفيد عادل وهبة خليل 2 (talk) 14:36, 15 July 2014 (UTC)

There is an bot-flag request underway for commons interwiki links at Wikidata:Requests for permissions/Bot/SamoaBot 39. For the arabic wikiedia interwiki links you might wan´t to ask at Wikidata:Bot requests.--Snaevar (talk) 14:25, 22 July 2014 (UTC)

Multiple claims when a person has received an award multiple times?[edit]

When a person has received an award multiple times (for example in different years), should one then tag this with multiple claims like this:
award received (P166)="some award", qualifier: point in time=2002
award received (P166)="some award", qualifier: point in time=2010

or use multiple qualifiers for one single claim?
award received (P166)="some award", qualifiers:point in time=2002, point in time=2010

Personally I think the first approach would be correct, but then I also saw the second one already on Wikidata. --Bthfan (talk) 21:31, 15 July 2014 (UTC)

I have used the second (ex. Farciot Edouart (Q11284)) for the OSCAR, but maybe the first is better for the source. --ValterVB (talk) 21:43, 15 July 2014 (UTC)
I tend to use the first format when dealing with things like position held (P39) or member of sports team (P54), because those tend to be more complicated with both start dates/end dates and other qualifiers. Awards are pretty straight-forward right now, so we might be able to get away with using multiple qualifiers on one claim, but I think generally it'd be better if we stuck to separate claims for each to allow for the possibility of more qualifiers in the future. Cbrown1023 talk 23:57, 15 July 2014 (UTC)
Second is better, because additional qualifiers may be used (per Cbrown1023) -- Vlsergey (talk) 05:50, 16 July 2014 (UTC)
The award is awarded twice and consequently it should show up twice. This does not happen with the second option in Reasonator for instance. Thanks, GerardM (talk) 09:26, 16 July 2014 (UTC)
The first one is better because it allows for more qualifiers for each award. /ℇsquilo 13:00, 16 July 2014 (UTC)

Many wrong coordinates[edit]

I've created a list with items, which have either wrong coordinate location (P625) or wrong country (P17) values: User:Pasleim/Implausible/coordinate.
Feel free to fix some of them. There might be some false positives, if I got the country borders wrong or if the item is spreading over several countries.--Pasleim (talk) 22:31, 15 July 2014 (UTC)

Nicely done! I've struck the ones I've fixed. A couple were confusing so I skipped them. --Haplology (talk) 06:03, 16 July 2014 (UTC)
Guess any bot should be programmed not to import the coordinate 0°0'0" N 0°0'0" E. All the Thailand articles I fixed had that nonsense coordinate, because the article had an empty coordinate template. Ahoerstemeier (talk) 07:37, 16 July 2014 (UTC)
Q1469808 is an Antarctic research base owned by Ukraine, so that the coordinates are obviously outside the Ukrainian borders. I would welcome all ideas how to solve this.--Ymblanter (talk) 12:02, 16 July 2014 (UTC)
Heritage sites shared by two countries are also difficult: Q3370031. Plus rivers which flow through multiple countries: Q14383 :( 130.88.141.34 15:12, 16 July 2014 (UTC)
There are also changes of country like Q46475 which depending on the start/end date belong to a country or another. Antarctic bases like Q403043 are also hard, can it be said that they are in a country?--Micru (talk) 15:32, 16 July 2014 (UTC)

One issue appears to be confusion over what "Country" means in each context. Does it mean "Located in", "Belongs to", "From here", "Made in", all the above? 130.88.141.34 07:56, 17 July 2014 (UTC)

So the best would be to delete country (P17). For "located in" there is is in the administrative territorial entity (P131). There is also applies to jurisdiction (P1001). Tamawashi (talk) 23:49, 17 July 2014 (UTC)

I updated the list and removed for now all items with multiple P17 claims. --Pasleim (talk) 19:35, 21 July 2014 (UTC)

WEF gadgets update[edit]

This is an gadget update notification.

External links gadget update

(«WEF: Links» in "Tools" menu at the left)

  1. all changes are stored using single edit query (second can be used for removing claims, if any were deleted), not query-per-edit as before
  2. support qualifiers (usually -- language (P407)). language (P407) qualifier used by (no label) (Q17116619) to select and display links (for example, if we have Russian and English links with normal rank, and French one with preferred rank, template will output Russian and French only)
  3. add encyclopedia links support (via own properties or described by source (P1343))
  4. add VIAF lookup support

Gadget is available in ruwiki directly (from prefferences). In all other projects (incl. Wikidata) it can be added to common.js with the following line:

mediaWiki.loader.load( '//ru.wikipedia.org/w/index.php?title=MediaWiki:WEF_ExternalLinks.js&action=raw&ctype=text/javascript&maxage=86400&smaxage=21600' );
New gadget to edit person data

(item «WEF: Person» in "Tools" menu at the left)

Gadget is available in ruwiki directly (from prefferences). In all other projects (incl. Wikidata) it can be added to common.js with the following line:

mediaWiki.loader.load( '//ru.wikipedia.org/w/index.php?title=MediaWiki:WEF_PersonEditor.js&action=raw&ctype=text/javascript&maxage=86400&smaxage=21600' );

Source code is available here: https://github.com/vlsergey/WE-Framework

Gadgets are localized in Russian and English. One of the following languages is used:

  • language from user preferences
  • content language of the project
  • English (fallback)
  • Russian (fallback)

Property and entity names are obtained as localized labels from Wikidata and stored in local storage cache.

Also, creation of new gadget is a matter of hours. If any project need a similar gadget (movies, cities, countries, languages) -- feel free to ask here or at my talk page. -- Vlsergey (talk) 05:47, 16 July 2014 (UTC)

Interesting and useful gadget I think! But somehow loading localized labels does not work in all cases fine though. I just tried with German language for the person gadget, when I click on "Update labels" in the "General" tab, sometimes it loads the German labels for the bottom three properties and sometimes it loads the German labels for the other properties in that tab. And sometimes it does not display any labels at all, just the property number.
Actually, I just logged in into Wikidata (I was only logged in into Wikipedia at the time I tried it) and now it works fine. So, problem resolved I guess? Not sure what logging into Wikidata changes here. Or it was just some random glitch. --Bthfan (talk) 07:32, 17 July 2014 (UTC)
@Bthfan:, labels are not loaded immediately, it takes some time to send requests to Wikidata server and load all labels and descriptions. But i will try to reproduce the case and check what can occur. -- Vlsergey (talk) 07:59, 17 July 2014 (UTC)
@Bthfan: I was able to reproduce the problem and fixed it. -- Vlsergey (talk) 15:38, 17 July 2014 (UTC)
@Vlsergey: Many thanks for doing this, it is so useful! The only thing I missed is some link to the property here in wikidata. Sometimes they are not translated, so I have to come to WD and look for the page here. I also missed these properties for people: movement (P135), notable works (P800) and this in general: Gran Enciclopèdia Catalana ID (P1296). Another gadget for taxon info would be useful too.--Micru (talk) 08:13, 17 July 2014 (UTC)
@Micru: i've added movement (P135) and notable works (P800) to person form; Gran Enciclopèdia Catalana ID (P1296) and Sandrart.net ID (P1422) to external links editor (on "Encyclopedias" tab). Let me check, what property types are required for taxon editor. -- Vlsergey (talk) 17:25, 17 July 2014 (UTC)
@Vlsergey: Wonderful! I have some problems when using the search function of Gran Enciclopèdia Catalana ID (P1296), for some reason it uses the English name in Wikidata instead of using the text I input on the box. It would be nice if CANTIC (P1273) had the search function enabled too, it looks something like: http://cantic.bnc.cat/index_nps/index?text=$1&index=1, where $1 is the search string with white spaces converted into "+" signs.--Micru (talk) 09:30, 18 July 2014 (UTC)
@Micru: All search on external websites uses the titles of wikipedia pages, not the value fields. So for Gran Enciclopèdia Catalana ID (P1296) it now uses the title of the page from cawiki first (if present), and after it -- title of the page from 'enwiki'. Also i've added search button for CANTIC (P1273). -- Vlsergey (talk) 17:23, 18 July 2014 (UTC)
@Vlsergey: I've translated the gadget in French. Could you add fallback for it please?
Thanks. — Ayack (talk) 10:01, 17 July 2014 (UTC)
@Ayack: thank you for the translation! Please, also, have a look at the commons file used by both gadgets, it may need some translation as well: [6]/[7]. -- Vlsergey (talk) 17:25, 17 July 2014 (UTC)

Taxon editor[edit]

@Micru: please check the new Taxon Editor:

mediaWiki.loader.load( '//ru.wikipedia.org/w/index.php?title=MediaWiki:WEF_TaxonEditor.js&action=raw&ctype=text/javascript&maxage=86400&smaxage=21600' );

I'm not a specialist in biology, so feel free to propose any changes to grouping/splitting/etc on the form (at github, for example: [8]). -- Vlsergey (talk) 08:59, 18 July 2014 (UTC)

Josve05a (talk)

FelixReimann (talk)
Infovarius (talk)
Daniel Mietchen (talk)
Soulkeeper (talk)
Brya (talk)
Klortho (talk)
Tobias1984 (talk)
Delusion23 (talk)
Alexander Vasenin (talk)
Notified participants to Wikiproject Taxonomy.

Some remarks: botanist author abbreviation (P428) and author citation (zoology) (P835) belong to the "person editor", not to the "taxon editor". I would also add ecoregion (WWF) (P1425).--Micru (talk) 09:23, 18 July 2014 (UTC)
@Micru: botanist author abbreviation (P428) and author citation (zoology) (P835) moved to "person editor", ecoregion (WWF) (P1425) added to "taxon editor". Thank you! -- Vlsergey (talk) 09:36, 18 July 2014 (UTC)

Issue with "instance of" for texts[edit]

Currently, Help:Sources say that items like. Even though there has already been long discussions about text-items, it seems that we have failed on this particular point. It is recommended that items like Hamlet or the Bible should be marked as instances of book (Q571). This is not correct. "Subclass of Q571" would not completely correct the issue as Q571 refers to "set of written, printed, illustrated, or blank sheets, made of ink, paper, parchment, or other materials, usually fastened together to hinge at one side.". What the Bible and Hamlet are is not a book but a text. I am not sure whether is should be instance or subclass of text (I would think "instance of" would be more convenient, but possibly not completely correct). Also, "text" seems a bit vague, so we could use some of its subclasses. What Hamlet is is a tragedy, what Ode to a Nightingale (Q3349126) is is a poem. So, it seems that the p31 (or p279) value should be the same as the p136. So I would argue that p136 is useless (at least for works) and that its content should be merged with p31, and that "instance of book" should be only used for iotems about real material books.---Zolo (talk) 08:10, 16 July 2014 (UTC)

@Zolo: I agree with you, it was a mistake from the early times that we should correct. We are also discussing here if there is a better label for "instance of/subclass of" which would avoid misunderstandings as these. It could also be considered to have a specific property for creative works. Notification Books task force participants--Micru (talk) 08:40, 16 July 2014 (UTC)

According to WIDAR autolist there are 38,000 or so items with instance of = book. Would many be solved by linking to Q8261 (novel), Q179461 (religious text), Q5185279 (poem) etc. instead? The novel is a more abstract concept e.g. JK Rowling has written seven Harry Potter novels which have been distributed via the medium of books. Therefore Harry potter and the Chamber of Secrets = Q8261. Non-fiction texts would be more difficult to specify without a new item. 130.88.141.34 08:56, 16 July 2014 (UTC)

We should probably have a generic "non-fiction text" item but many coulds have more specific values (biography, essay, handbook, etc.)--Zolo (talk) 09:25, 16 July 2014 (UTC)
In general, I am in favor of adding instance of (P31), even if the item used is overly generic or slightly "off". Items without instance of (P31) are hard to find, if you look for something specific; but making instance of (P31) more specific is much easier if you have an "anchor" to work with. Maybe not the best example, but instance of:book in en:category:Poems gives us ~270 candidates. Autolist2 can add and remove statements, so selecting the poems in that list (or de-selecting the non-poems) and changing instance of (P31) is straight-forward. In the future, we could even have a tool list us "usual suspects", such as "instance of:book" automatically, and suggest "popular" replacements. --Magnus Manske (talk) 12:21, 16 July 2014 (UTC)
@Magnus Manske: The thing is "instance of book" is supposed to mean a material book like Lorsch codex (Q538377). These call for properties items about text do not (for their physical location and their material description). I am in favour of adding a generic p31 in the absence of a generic one, but I think it should be text (Q234460) rather than book (Q571), so that we do not mix differnt classes of objects. --Zolo (talk) 15:11, 16 July 2014 (UTC)
I agree. There are a few physical books which have their own articles and these particular tomes should be 'instance of:book'. Items about a text should be 'instance of:novel, poem, etc.' and 'subclass of:book, ebook, etc.' Filceolaire (talk) 21:34, 17 July 2014 (UTC)
@Filceolaire: Text is an abstract object if it is not a physical text, so "instance of" is wrong as it is for most books. Tamawashi (talk) 14:16, 18 July 2014 (UTC)
Tamawashi An abstract object can be an instance of a class of abstract objects. A book is a physical object so a book and a text must be in different classes. A whole taxonomy for books has been developed by librarians but I've lost the link. Filceolaire (talk) 19:39, 18 July 2014 (UTC)
@Filceolaire:. I don't think texts should be subclasses of books,~ebook etc. either, as they are conceptually independent from their physical materialization (they would remain the same text if they were written on a wall or recited aloud). However, in the terminology of Help:Sources, instances of {Q|3331189}} could be subclasses of books or ebooks. --Zolo (talk) 20:43, 18 July 2014 (UTC)
@Filceolaire: If an abstract object can be an instance of another abstract object, then how can editors decide whether to use "instance of" or "subclass of"? Tamawashi (talk) 02:18, 19 July 2014 (UTC)

@Filceolaire, Zolo, Tamawashi: There cannot be instances of abstract concepts in Wikidata, although instances of abstract concepts do exist in real life. For example, let's take a look at The Raven (Q22726):

  • if I memorize it (the process of transfering it to my pasive memory), I encode it in my neuronal circuits (as material substrate)
  • if I recite it (the process of activating passive information), other sentient beings or sensing devices (recorders) can become affected by the organized patterns of sounds that is my performance and perhaps create new instances of it
  • if I write it down, then again I am modifying a material entity (paper) to encode passive information that can be eventually decoded by someone else.

All those are instances, but if I want "something" to represent all of them, then I need a class, therefore <The Raven (Q22726)>subclass of<literary work (Q7725634)>. Same applies to music, movies, etc. which might be confusing to users (even for me it was confusing until not long ago), that is why I suggested above changing the label of p:p279/p:p31 to something easier to distinguish.--Micru (talk) 08:32, 19 July 2014 (UTC)

@Tamawashi, Micru: I really do not understand why an instance should be something with a location in space and time. Beside that would not be identical with being a material object. A football match can be located in space and time but it is not a physical object.
It seems much more useful to define instances and subclasses in terms of logical implications (like in Help:Basic membership properties). This is the cleanest way to use these properties for making inferences. Instance = the item has this feature. Subclass = Instances of this item has these features. A text is a defined sequence of word. Everything that is a defined sequence of word is an instance of text. Everything whose instances have these features is a subclass of text.--Zolo (talk) 08:55, 19 July 2014 (UTC)
@Zolo: There are not only material (or boundary-defined) objects, there are also processes. A football match is a process that is voluntarily engaged by agents (players) according to agreed rules.
There are two views: either we have models and we "create" things from those models (instances of a model), or instances exist on their own in real life and we create models from them. In practice it doesn't matter much, because you can express reality one way or the other and still reach the same reality representation. However, for the common mortal (non-IT people), the notion that the reality comes from the representation of it, it is, to say it in plain terms, a bit weird.--Micru (talk) 09:50, 19 July 2014 (UTC)
@Micru: as you say, the way we see the causal structure of the World does not really need to show in the use of the properties. I am just saying that the simplest and most useful definition of a an instance is "something that has the features defined by a class" An irrational number is a number that cannot be expressed as a ratio of integers. "Irrational number > 3" is not a number, so not an instance of irrational number. Other the other hand pi is an instance of irrational number. Adding requirements about physical location does not appear to have any practical benefit. Moreover, OWL now allows something to be both a subclass of X and an instance of Y. It is hard to see how an object with a physical location can be a class, so it must be that something without one can be an instance.
If we define a text as a unique sequence of words, or something like that, an instance of text must be an abstract object, not a physical one. This would make "Ode to a Nightingale" an instance of poem, and that would tell us that it is a unique poem, not a type of poem the way ode (Q178985) is. It would also permit additional constraints like "values of the edition (P747) property must be instances of text". -Zolo (talk) 11:26, 19 July 2014 (UTC)
@Zolo: If you define an instance as "something that has the features defined by a class", then what is a class and a superclass?
"Irrational number > 3" is a "subclass of:mathematical concept", when you write it down or, when you read about it, think about it, or otherwise create it in your mind, then you have an "instance of:Irrational number > 3". Pi, as any other irrational number, is also a "subclass of:irrational number concept". When I write down "Pi", that is an "instance of:irrational number concept", when I calculate it and write down the value that becomes an "instance of:approximation to irrational number".
The instances of the neurological processess of John Keats combined the linguistic patterns that he had acquired through his life to give rise to the first instance of "Ode to a Nightingale" in his mind. He penned it down and created another instance of "Ode to a Nightingale", then it was printed and more instances of "Ode to a Nightingale" were created, each one of them had a different material support, but all of them have a strong simmilarity and for that we can speak of a "the class of all instances that exist of 'Ode to a Nightingale'".
If you think about it, the only reason we have the sepparation between "instances" and "classes" is because for some things we have the factual certainty that they exist (instance, token, empirical evidence, phenomenon, the actual thing, reality) while with the others we imagine or suppose that they exist (class, type, kind, rational evidence, nuomenon, fabrication). Even the concept of gravity is a class that we have inferred after seeing instances of its effects repeated times. That is the basis of the scientific method, create classes from the observation of repeated instances. If you don't make that distinction, then it is equivalent to say that there is no difference at all, and in that case we can just use one property for everything and rely on bottom concepts.--Micru (talk) 12:08, 19 July 2014 (UTC)
@Micru: An instance is "something that has the features defined by a class and a subclass is something whose instances have the features of a class. Pi is a unique number and it is irrational, so it is an instance of irrational number. "Irrational number > 3" is not a itself a number so it cannot be an instance of number, but everything that meets the conditions "is a number", "is irrational" and "is > 3 is an instance of irrational number. So "irrational number > 3" is a subclass of irrational number.
Is the mental representation of an armchair -or the neural configuration producing this mental representation- an instance of armchair ? No because a mental representation cannot have a back and two arms. But just the same, how can a mental representation be an irrational number or be > 3 ?
That said there are certainly issues with allowing something to be both a class of something and an instance of something else. How do we know that the the statement "is > 0" means "this item is > 0" for pi (Q167) and "all instances of this item are > 0" for natural number (Q21199). That may not be trivial, but it appears to be a practical issue as opposed to the logical dead-end that seems to be created by requiring instances to be phyiscally localizable objects. --Zolo (talk) 14:02, 19 July 2014 (UTC)

@Zolo: Saying that "an instance is something that has the features defined by a class" is the same as saying that you have an object because you have a general representation of it... It doesn't seem rational to say "I know that the Mona Lisa is a painting because I always knew what a painting is". When I was little and I had no knowledge of language I wouldn't have been able to recognize the Monalisa as a painting, so most probably someone showed me a painting saying "this is a painting, and this, and this, and this", so that is why now I have a model (class) of what a painting is based on my exposure to the instances of that class and why now I can say "Monalisa instance of painting".

Pi is a pattern that happens in nature and we have idealized it (class). "Irrational number > 3" is a fabrication that you made up in your mind and I can replicate it almost identically in my mind (I say "almost" because I cannot have knowledge of your subjective experience while you are evoking that representation, so I cannot compare it to mine), for you that is enough to call both instances "the same", even if they are not represented with the same mind-substrate.

A mental representation of an armchair is the closest our mind can be to that existing armchair, that is why we call it "instance" and also because we can agree with others that we are referring to exactly the same object in time and space. Even if we now that it is a lie (our "mind object" or "object representation" can never be the object itself), it is "less lie" that pretending that things can exist without material substrate. In the same way, a mental representation is the closest we can be to an irrational number. If there wouldn't be matter where to store that representation, that representation wouldn't exist.

You cannot know the answer of "pi > 0" because pi is an abstraction and as such it has no value, but you can ask if "approximation of pi > 0", because you have executed some algorithm that gave you Approximations of π (Q1130396) and you can use that as a value for the comparison. The comparison itself "z = (x > y)" is a subclass of all comparison algorithms (passive process) that give a binary value of z as a result of the execution of the comparison process. The algorithm has an agent that executes the comparison process. x and y have to fulfill certain characteristics for the agent to be able to perform the comparison, but most probably the conditions are to have enough memory to store the representation of both x and y, and enough processing capacity to manipulate both representations. "z = (x > y)" as such may represent 3 superposed states: indeterminate, true, false. The agent selects which one is active at a given time. But all in all you need an instance of that algorithm (implementation, be it in your brain or in a computer) to be able to use it, which requires matter, energy, and takes space.--Micru (talk) 16:01, 19 July 2014 (UTC)

@Pasleim, DavidMar86hdf, Zolo, Micru, Filceolaire: And now classification as an instance [9]. Tamawashi (talk) 16:53, 19 July 2014 (UTC)
@Micru: I have a feeling that we do not speak the same language here. This is not about the psychological origins of knowledge. If we want to know if Mona Lisa is a painting, we just need to check whether it matches the citieria set in painting (Q3305213) or in some extrernal source.
"irrational number > 3" does not mean "someone's idea of a irrational number larger than 3", it means the set of real numbers that are larger than 3 and cannot be written as a quotient of real numbers. The neuronal circuit creating your idea of an irrational number may well be completely different from the one creating mine, but it is really irrelevant."Irrational number" is a mathematical object defined in mathematical terms. How the minds encodes it does not change a thing to its definition or its constituents. --Zolo (talk) 17:40, 19 July 2014 (UTC)
@Zolo: No, no, it is not about the "psychological origins of knowledge", it is plainly about the "origins of knowledge", or just about "knowledge". You are saying that it is possible to have information without matter to represent it - I am saying that is impossible. You are saying that if you write the number 3 in one paper, and the number 3 in another paper, you don't have two instances of the number three - I am saying that is impossible. You speak from the perspective of mathematics, where things do not need to exist in order to be imagined - I speak from the perspective of ontology, where even the concept of non-existence needs a material substrate to be represented. Our postures are very different, but it is not necessary to keep talking about this. Let's keep working on productive tasks and may some insight come to any of us that allows us some day to come closer :) --Micru (talk) 18:04, 19 July 2014 (UTC)
@Micru:, yes maybe it is better to focus on other things, but I am absolutely not thinking that we can have information without matter, just that:
a) The simplest and most productive definition of instance is "something that has the features defined by a class"
b) The simplest and most productive definition a subclass is "something whose instances have the features of a class".
c) A mathematical concept should be defined using the definition given by mathematicians.
d) Mathematicians define pi as "the unique real number whose value is equal to the ratio of a circle's conference to its radius", and prime number as "any natural number greater than 1 that has no positive divisors other than 1 and itself."
It directly follows from the points above that pi is an instance of real number and prime number is a subclass of natural number. These are plain logical implications. It does not purport to say anything about the nature of knowledge.
Obviously, we could use choose other definition for "subclass" and "instance". We could say that an instance is something that has a location in space and time. But the latter definition has serious drawbacks, like not making the distinction between a number and a set of numbers, or between an individual novel and a literary genre. Given that definition b solves this issue, I really do not see any reason why we should not use it. --Zolo (talk) 08:02, 20 July 2014 (UTC)

@Zolo: After reading your comment, I think we are closer that we imagine we are. I agree with your inderdependent definition of class-instance. For a class to exist (in a representation system) it needs an instance, and for an instance to exist (in a representation system) we need a class to represent it with. Our difference is in where this inderdependece relation originates. You say that it can exist "in the void", where I say that they need to be "grounded", that is, that instances must exist first in space and time. The definition of Pi adapted to ontological language is "there is an (idealized) class that represents a unique (idealized) result equal to the ratio of an (idealized class of) circle's conference to its (idealized) radius". It is important to understand deeply that the definition refers to classes, not instances, even if the result seems to be a unique number, because it is not. I think what is confusing is that you can have an instance of Pi (the idealized class) and also an instance of Pi (the class of all aproximations). For instance if I write: π, π, π, π, π, π, π, those are 7 instances of the idealized class of Pi (the ones that we can only think about abstractedly), but if I write 3.14159, 3.14159, 3.14, 3.14, those are 4 instances of the class of all Pi aproximations (the ones that we can process). If there was not a Pi (the idealized class), I wouldn't be able to write instances of that idealized class of Pi. Notice also that all instances exist in time and space, where the class is a supposition that more instances might exist.

As of "the latter definition [considering instances in time and space] has serious drawbacks, like not making the distinction between a number and a set of numbers, or between an individual novel and a literary genre." I don't see those drawbacks. Consider the following:

  • there is an idealized class of a set
  • there are instances of that idealized class of a set
  • there is a class of all real occurrences of a set
  • there are instances of a real occurrence of a set

So now a practical example with a natural number "5"

  • there is an idealized class of 5: the mathematical five
  • there are instances of that idealized class of 5: representations of the matematical five (5, 5, 5, 5, 5....)
  • there is a class of all real occurrences of 5: all fives that can happen in real life
  • there are instances of that class of real occurrence of 5: five apples here on my table

Of course it wouldn't be practical to type in all this structure in Wikidata, because we just need the mathematical idealized class, but that one is clearly a class, not an instance, both a class and and instance otherwise we wouldn't be able to instanciate it. Remember your own definition: for a class to exist there must be an instance, and the other way round.--Micru (talk) 09:48, 20 July 2014 (UTC)

@Micru: sorry, it is growing way too long for the Project chat, but we do not agree at all:
  • Concepts and real objets are too disjoint classes. A material object cannot be an instance of concept and vice versa. Concepts do not have a material existence, they only have a defintion.
  • Numbers are concepts, so no material object can be an instance of number.
  • Five apple are five instances of apple or an instance of group of apples (with the property: quantity of apples: 5), but it is absolutely not an instance of 5.
  • Writing "pi" on a paper does not create an instance of pi, just an instance of a depiction of the pi symbol. --Zolo (talk) 14:57, 20 July 2014 (UTC)
@Zolo: Yes, it is getting too long for the project chat. You say that concepts do not have material substance and I agree, but they can happen simultaneously in two different places. True that "concepts and real objets are too disjoint classes". Concepts are also both a class and an instance of themselves (recurrence), I corrected my words above to reflect this. Classes (group of apples) and their instances (five apples) do not show that property. What bothers me is the wish to use the same property both to relate classes with instances and concepts with classes. If they are two disjoint groups, why to use the same property?--Micru (talk) 15:26, 20 July 2014 (UTC)
@Micru: If something meets the formal definition of instance, I do not see why it should not be called an instance. That concepts and material objects are disjoint classes is actually rather fortunate in that respect. This way the superclass tree of the p31 value of an item can tell us with certainty whether the item is a material object or not. --Zolo (talk) 15:50, 20 July 2014 (UTC)
@Zolo: I have made a drawing above. If the red lines were instances, then we would have instance of instance, which doesn't seem right.--Micru (talk) 16:40, 20 July 2014 (UTC)
@Micru: Yes, as you had already noted, concept is an instance of concept and number is also an instance of concept so instance is an instance of an instance of concept. But that's ok, ask user:tomT0m about metamodelling :). But individual texts and individual numbers are really not classes so this is not an issue here (no a number cannot exist at several places at the same time, because a number, according to the standard definition, really does not exist in a spatiotemporal sense). --Zolo (talk) 17:20, 20 July 2014 (UTC)
@Zolo: A book may contain text and a text maybe published as a book. (But of cause you can publish a collection of pictures without text as a book and not every text is published as a book.) If you want a perfect definition for properties do not use Wikipedia. Why: Because WP combines articles in different languages. Abstract concepts are not existing like a single person and therefore they will always set apart. (German WP: book = printed book, German language: book = printed books and e-books, Unesco: only publications with more than 49 pages are "books" etc.) In our (WD) definition a "book" - book (Q571) - is a kind of work, the published book is called "edition" - edition (Q3331189). Changing basic definitions is like shooting a running horse and than wonder why this horse didn't win the rase. --Kolja21 (talk) 11:08, 20 July 2014 (UTC)
@Kolja21: we may not always be able to get perfectly equal definitions among projects, but here pretty much every Wikipedia and "authority control" link define books as a sequence of sheets etc. Even if we add ebooks to the definition, we have something very different from book seen as a "kind of work, the published book is called "edition", the published book is called "edition". Beside we need an item about real material books like Lorsch codex (Q538377)) and it only makes sense that it is the one with the links to en:Book,, etc.
@Kolja21:It is true that we should try not to change the structure of Wikidata too often. If you do not want to change the statements made in other items, and the documentation pages, we can also move the sitelinks, statements, and labels to a new item, and move the content of text (Q234460) to Q571. That will affect less pages, and, with me at least, it would be perfectly ok. --Zolo (talk) 14:57, 20 July 2014 (UTC)
@Zolo: text (Q234460) has not been used for works/books yet. So there is nothing to move. For the book In Mexico (2010) Wikidata tells us that the author is a person called "text" but that is just one of the usual bot errors.[10] --Kolja21 (talk) 15:21, 21 July 2014 (UTC)
@Kolja21: Items like On the Origin of Species (Q20124) are marked as "instances of Q571". The Wikipedia and authority control links in Q571 say that a book is a material object made of leaves. On the Origin of Species (Q20124) is not a material object made of leaves. So we must either change the value indicated in On the Origin of Species (Q20124) or the Wikipedia/authority control links in Q571. --Zolo (talk) 17:04, 21 July 2014 (UTC)
@Zolo: As I said before you will never have a 100 percent match. Your definition of book is one possible defintion, but not the one of WD. Take for example the English description: book = "medium for a collection of words and/or pictures to represent knowledge, often manifested in bound paper and ink, or in e-books". (An e-book as you know is not made out of leaves.) This is the official English description - today. Tomorrow an editor might change it. And of cause the German description is different. So forget about the English word "book" and the rest of the 200 or 300 possible labels, descriptions and Wikipedia links. It's item Q571 and we are happy with it. --Kolja21 (talk) 23:20, 21 July 2014 (UTC)
@Kolja21: I am not sure that you have correectly read what worte. I said that (almost ?) all sitelinks to Wikipedia said a book is made of leaves, so it would make sense adopt this definition in the corresponding Wikidata item. And even when Wikipedia links do not match 100%, descriptions in Wikidata should. If the German and English descriptions in Wikidata do not match, then at least one of them should be changed. Beside, this is not an issue of not being 100% right. As I, and Filceolaire, and Micru, and other before have noted it is just 100% wrong to say that "The Origin of Species" is an instance of book - whether we include e-books or not. If we changed the definition of Q571 to make it mean "text" instead of "book", it may be 95% right. --Zolo (talk) 07:33, 22 July 2014 (UTC)
A text? Are you serious? If we have problems with a difinition of a simple thing like a book, don't even think about talking about a text. This term is so complicated you can write a doctor theses about it and we still wouldn't agree what a text is. The concept of WD is to connect properties with items. "Freiheit (German) = Freedom (English)" is not the same as "2 + 2 = 4." Move sitelinks as often as you like, split them up, rename them and later merge them again, but stopp changing the definitions of properties. We lost hundreds of working hours because of chaos resulting in these kind of changes. We said The Origin of Species is an instance of "book". That is a definition. It's good and it works. If we would have said it's a "work", it might be a the title of an opera. If we would have called it "literary work" it would exclude many types of books. If we would have said it's a "text" it could be anything, from DNA to an email. --Kolja21 (talk) 15:14, 22 July 2014 (UTC)
As it stands now, should be considered plain wrong for any of you. The German Language knows the word Schriftwerk implying a work predominantly textual in nature but not necessarily an intellectual or artistic work (think of a phone book which essentially is a database, i.e. the effort relies on the act of collecting). However this would also include individual articles (published in journals or newspapers) like essays, a generalization not everybody might find appropriate. The main obstacle seems to be to translate Schriftwerk into enough languages to pin down the meaning of the corresponding item here. -- Gymel (talk) 15:33, 22 July 2014 (UTC)
@Gymel:. Would "written work" sound ok in English ? Also, note that both the translation and the generality issue will not be too prominent if, as proposed, we recommend using more specific subclass like "poem" or "phone directory" (best avoid the word book :) ) for the value of p31.--Zolo (talk) 07:22, 23 July 2014 (UTC)
@Gymel, Zolo: I think we need a property to say "<book> embodiment of <literary work>". If you don't like that label it could be said as: "gives material form to", or "materialization of".--Micru (talk) 07:55, 23 July 2014 (UTC)

New editor removing sitelinks[edit]

I don't know who they are but they are still there. And their vandalism often remains unreverted [11]. Please join the fight against them ! And it should be done regulary. From what I understand only the last 500 of them can be found easily, after that they disappear from recent changes and from the visible World alltogether. --Zolo (talk) 09:19, 16 July 2014 (UTC)

Or you can watch this page: User:KrBot/Lost links. — Ayack (talk) 09:33, 16 July 2014 (UTC)
@Zolo: and it's something easily detectable (sight). TomT0m (talk) 10:41, 16 July 2014 (UTC)
@Ayack: I did not know the report page, it is actually much better. --Zolo (talk) 12:00, 16 July 2014 (UTC)
I am working on it every day but more help is certainly welcome.--Ymblanter (talk) 12:15, 16 July 2014 (UTC)
@Ymblanter:, yes actually the situation seems better than I thought. It sometimes takes one or two days before being reverted, but at the end most of them get done. The most problematic cases seem to be with languages that are inintelligible to most contributors, like Japanese. --Zolo (talk) 07:14, 17 July 2014 (UTC)
There are some cases when an editor replaces a link and the old one is not attached to Wikidata. In these cases I usually do not act, except if I know the language AND the subject.--Ymblanter (talk) 07:25, 17 July 2014 (UTC)
I didn't know about that helpful page until today. Thanks KrBot, nicely done! I have looked at the Japanese items. Two needed new items, one was probably a confused IP user (two areas in Scotland with the same name), and the other was probably just random vandalism. FWIW, I have been watching RecentChanges for some time now, and it is exceedingly rare to find vandalism or mistaken edits for Japanese, there is not so much need to worry about Japanese at least. --Haplology (talk) 14:05, 17 July 2014 (UTC)

Wiki Loves Monuments UK[edit]

This year's Wiki Loves Monuments UK would like to try something new: Instead of managing objects in large tables on Wikipedia (shudder), they should be proper Wikidata items! We are talking about English, Welsh, Scottish and Northern Ireland grade I and II* buildings; a first estimate is 50,000 objects. Quite a few of them are already on Wikidata, and all of them should be automatically notable due to their grade status. I will soon get the "official" lists of objects, write a tool to find the ones we already have, and to create the ones we are missing. At the very least, objects created through this tool will have name, coordinates, grade status, and possibly short descriptions and "administrative unit". Through the previous and upcoming WLM, most of them should also get an image or two. I just wanted to get everyone here up-to-date, and prepare you for a potentially large influx of building items without Wikipedia articles. I think this will be a major win for both Wikidata and WLM! --Magnus Manske (talk) 10:42, 16 July 2014 (UTC)

Welcome to Wikidata:Cultural heritage task force--Ymblanter (talk) 12:17, 16 July 2014 (UTC)
I see you're lacking UK data there ;-) --Magnus Manske (talk) 12:27, 16 July 2014 (UTC)
I took the liberty to fix your title. This year Wiki Loves Monuments won't be an international event although some countries are running local events. Not sure about the UK.
Over the last weeks I've been importing Rijksmonument (Q916333). I'm now at about 20.000 out of 65.000. I'm using the data from the Monuments database. Shouldn't be too hard to do this for other countries too, the UK is in the database. Multichill (talk) 16:54, 16 July 2014 (UTC)
Thanks for the welcome here! As part of the process we will be updating the UK data using the most recent information from the official listing authorities. --62.7.191.176 07:21, 17 July 2014 (UTC)

Is there a tool / bot to detect REDIRECT's linked to items?[edit]

Hi! Comala, Colima (Q1113251) lists the REDIRECT en:Comala, Colima and Comala (Q5150494) lists the page en:Comala. The question is not about merging the WD items. The question is how such / similar items can be detected. gangLeri לערי ריינהארט (talk) 12:41, 16 July 2014 (UTC)

Ping! @Magnus_Manske, @Lydia_Pintscher_(WMDE), @Ladsgroup / Amir . The question sis quite simple. Please answer! Thanks in advance! לערי ריינהארט (talk) 07:46, 18 July 2014 (UTC)
@לערי ריינהארט: If you link the usernames or if you use the template {{ping}}, the users get a message via Echo. Otherwise, it's likely that the addressed users don't see the post. --Pasleim (talk) 21:19, 20 July 2014 (UTC)
@Pasleim As far as I know wiki syntax checker can be tricked using UNDERSCORE's instead of spaces in user names. template:ping does the same. לערי ריינהארט (talk) 21:25, 20 July 2014 (UTC)

This paragraph should not be archived unless the question is answered. לערי ריינהארט (talk) 19:37, 22 July 2014 (UTC)

Hungarian honoured artist[edit]

Hoi, The Hungarian and Russian honoured artist categories were intertwined. I have created a new item for the hungarian category and artist. I have added statements for all humans in that category and I removed them from the Russian category.. There may be more stuff in the Russian category that is wrong I can not tell.

I can do either of two things. I can clear the current data and add data based on the category or I can do nothing further. Thanks, GerardM (talk) 16:56, 16 July 2014 (UTC)

Gerard, would you please post the links.--Ymblanter (talk) 19:31, 16 July 2014 (UTC)
Q17370480 and Q4187990 GerardM (talk) 19:37, 16 July 2014 (UTC)
There are not so many links. You can drop the matter, I will take care of them. Thanks for noticing.--Ymblanter (talk) 20:15, 16 July 2014 (UTC)
<grin> what did you do, it is also a question for future reference ... ? GerardM (talk) 20:59, 16 July 2014 (UTC)
I went through the list of articles which refer to Q4187990 (this is an article not a category btw) and checked that all of them are honoured artists of Russia, not Hungary.--Ymblanter (talk) 06:37, 17 July 2014 (UTC)
There is a category as well ... GerardM (talk) 23:27, 18 July 2014 (UTC)

removal of persondata[edit]

On enwiki, every few months somebody ask a question about persondata's existence. The answer I usually give is that it probably will go away with Wikidata now here. I see one bot, User:SamoaBot, grabs data from the persondata's "short description" parameter. On SomaoBot's page, it states it still acquires this parameter.

Long story short, from a Wikidata prospective, is persondata still useful or can it be dropped? Bgwhite (talk) 18:29, 16 July 2014 (UTC)

It can be dropped. -- Vlsergey (talk) 18:57, 16 July 2014 (UTC)
'running intermittently' means there are no currently known obstacles to that task, and that I could run it at my discretion, but I do not recall having run that task in these months. However, tons of data have already been imported, and now it is up to the English Wikipedia community to decide whether they still want those data within articles. My RFBot in 2013 was not accepted. --Ricordisamoa 21:53, 16 July 2014 (UTC)
Before ripping the persondata out of all the enwiki articles, would it be possible to convert the persondata to a lua template that grabs it back from Wikidata? That would be very useful to have, especially for new stubs and articles where the places and day/month of birth or death have yet to be added to the article. Jane023 (talk) 08:58, 17 July 2014 (UTC)
That's something that should be done in infoboxes, not the persondata template. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 11:17, 17 July 2014 (UTC)
This may be true for English Wikipedia (I don't know), but e.g. the German Wikipedia does use persondata templtes, but generally no infoboxes for biographical articles. So any effort put into generating benefit from the connection of persondata templates and Wikidata might at least help the German Wikipedia, even if the English Wikipedia has other ways to produce the same results. --YMS (talk) 12:04, 17 July 2014 (UTC)
Concerning the persondata template in de.wikipedia: For those capable of reading German, please read the last two sections at Hilfe Diskussion:Personendaten/Archiv/5 (Wikidata and Organisation eines reibungslosen Umstiegs). The gist of the argument: Please no wikidata values for that template. --HHill (talk) 12:41, 17 July 2014 (UTC)
Speaking as someone who uses Persondata on their own personal project, I worry that my useful source of data is going to disappear. Whilst I agree Wikidata is the better solution, I have yet to see an explanation on how Wikidata keeps the data accurate. For example, when someone dies, Persondata is updated within minutes! What process is in place to update Wikidata? Periglio (talk) 01:21, 18 July 2014 (UTC)
There are currently 3177 people categorized in en:Category:2014 deaths (11 articles in the category are not about people nor animals). There are currently 3174 people and 6 animals in Wikidata with a 2014 death date + a link to en.wikipedia, so the difference seems negligible. Wikidata relies a lot on Wikipedia to add these dates, but I do not think it uses persondata anymore, so removing them would not hurt. But I may be wrong on that.
What process is in place to keep en.wikipedia persondata accurate ? If there is a good one, it can probably be transferred to Wikidata. -Zolo (talk) 07:28, 18 July 2014 (UTC)
Do WikiData accept "death date" without a reference? 2.110.98.101 23:37, 19 July 2014 (UTC)

──────────────────────────────────────────────────────────────────────────────────────────────────── I've noticed (and corrected) two cases just recently (highlighted as bugs in the revision history statistics display) where date of birth was present in Persondata but not in wikidata, even though date of death was in both. We should probably deal with such cases before removing information from the affected articles. --Mirokado (talk) 04:30, 20 July 2014 (UTC)

An RfC on en.WP has now been opened. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 22:17, 20 July 2014 (UTC)

What to do when an item refers to two things in one Wikipedia article?[edit]

Q4010540 The Wikipedia article on this topic is about both the fictional character named Vext as well as a comic book series entitled Vext. In part, this is due to parameters in the infobox that collapse together fictional characters as well as series when they have the same name (note also that sometimes, there will be more than one comic book series with the same title—they are usually disambiguated by volume). What do we do on Wikidata about this? Is this data item about the character, the series (there is only one volume), both, or neither? —Justin (koavf)TCM 17:25, 17 July 2014 (UTC)

In my opinion, the item, in general, should be about the larger concept. An article about a city and the same name surrounding county would link to an item about the county. In this case the item linked to the article would be about the comic series with a separate item for any characters notable enough to have their own items (do we know enough about them to make three statements?)
On the other hand, in this case the article starts by talking about the character, not the series so maybe...
On the third hand the article goes on to describe 17 other characters as well so by the end of the article it does seem to be about the whole series.
All the above influences whether the enWP article is linked to the wikidata item for the series or the wikidata item for the character. Either way we still need separate items for these as statements about one are not true for the other.
Hope this helps a little. Filceolaire (talk) 21:12, 17 July 2014 (UTC)
On the specific question you asked: This is what is known as the "Bonnie and Clyde problem". A wikidata item should not refer to 'two things' though it may refer to a pair of things or a bunch of things where we can make statements about that pair or that bunch. If a wikipedia article refers to two things then the item linked to that article will in turn have 'consists of' statements linking separate items for those things. Where an article discusses one class/group/big thing and also a bunch of other things that can be considered to be 'part of' or 'instance of' or 'subclass of' the group/class/big thing then the Wikipedia article should link to the wikidata item for the group/class/big thing. Filceolaire (talk) 21:22, 17 July 2014 (UTC)
@Filceolaire: But this isn't like w:Coen Brothers/Q56095 where there are two things (Joel and Ethan) who have some separate properties (e.g. their birth years) and some common properties (e.g. awards that they have won together or occupations that they have shared separately): it's not like a comic book series is an aggregate of comic book characters. One is a publication, the other is a fictional entity and they both just have the same name. I would guess that the series is the "main" topic but that's more or less arbitrary, really. —Justin (koavf)TCM 07:40, 18 July 2014 (UTC)
Note: I actually looked up that Wikidata item and there are parts segregated out to Q13595531 and Q13595311 even though neither has an independent Wikipedia article, Commons category/gallery, Wikiquote listing, Wikivoyage guide, or Wikisource bibliography... —Justin (koavf)TCM 07:42, 18 July 2014 (UTC)
Yes Justin except I didn't talk about the 'main' topic. I said the article should link to the group/class/big thing and that is the series. Each of the characters is featured in the series (and listed on the page). Each individual comic is a part of the series (and referred to in the 'publication history' section of the article). so the series is the group/class/big thing and the wikipedia article should link to the item for that.
I can imagine a language wikipedia with an article about Captain America which includes discussion of his appearance in his own comic, the Avengers, movies, TV series etc. where the article should link to the item for Captain America and not to one of the comic book series. Either way Wikidata still needs a separate item for each series, each character, each edition. Wikipedia can combine these topics in articles in many different ways and which of these items each article links to is then a separate question. Filceolaire (talk) 09:47, 18 July 2014 (UTC)
@Filceolaire: Ah. Thanks. That makes sense. So I can/should split this single item into two? —Justin (koavf)TCM 03:09, 21 July 2014 (UTC)

Main page redesign[edit]

Hey folks,

As part of my internship and continuing efforts to improve outreach for Wikidata, I'm now working on updating the Main page. I've put together a summary of areas for improvement and recommendations for a redesign at Wikidata:Portal_Redesign; please review and comment when you can. You can leave feedback or other suggestions at Wikidata:Portal_Redesign#Feedback—ideas and brainstorming on possible visuals to include and ways of featuring fresh and dynamic content (without the need for constant maintenance by users) are very welcome!

FYI, I am also now working on a redesign mock-up of the Main page and will be sharing this in the upcoming days so there will be another opportunity to give feedback before any changes are implemented. Looking forward to your comments! -Thepwnco (talk) 17:52, 17 July 2014 (UTC)

Date format for years<100[edit]

Please, see comment on en:Module talk:Wikidata#Date error for years of 1st century. It seems an inconsistent date format for dates of the 1st century. --Vriullop (talk) 19:42, 17 July 2014 (UTC)

Unreliable English Wikipedia[edit]

@Ladsgroup: I checked ca. 10 claims P313 that had Svilengrad - they were all wrong. Correct is Svilengrad Municipality, example: [12]. Most or all are based on edits by Dexbot. The algorithm that made these may have caused hundreds of wrong claims. Tamawashi (talk) 00:42, 18 July 2014 (UTC)

when the bot is importing wrong data from Wikipedia, please at first fix it in that Wikipedia (specially when the wiki is the biggest) and please inform me exactly where the issue happened. I use lots of checks to catch the errors but if people don't tell me when bad data have been imported I can't fix it. By the way 10 errors in more than 200K data is acceptable for me. Amir (talk) 01:57, 18 July 2014 (UTC)
@Ladsgroup: it was 10 errors in 10 items checked for that set. I wrote "The algorithm that made these may have caused hundreds of wrong claims." - I don't know which algorithm you used here, the bot edits don't reveal that. I don't even know what item(s) in English WP you used as source(s) for your claims, so even if someone would like to "fix it in that Wikipedia" he may be unable to know where. And, no, I will not fix errors in that Wikipedia. For geography of Bulgaria it is unreliable. Tamawashi (talk) 10:35, 18 July 2014 (UTC)
It's error in places that their P131 is a municipality and they are in Hungary, I can say it won't be more than 500. Can you check some of places like this and list them to me? (somehow, e.g. using Autolist) and after that I will fix them Amir (talk) 00:33, 19 July 2014 (UTC)
@Ladsgroup: Maybe next time you ping, I just came back here because Kameno Municipality has the same issue. It is Bulgaria, not Hungary. The villages are not in Foo, but in Foo Municipality. How can you say it won't be more than 500? BG Wikipedia lists ~250 templates for municipalities. So, if there are more than two items per template I would assume it could be more than 500. The templates also indicate the type, Село=village, Град=town. Could the bot add an algorithm number to each edit and the algorithm be published with that number somewhere? So users could see how many items might have been affected. Or at least put a description, e.g. importing from municipality templates from BG Wikipedia. Tamawashi (talk) 04:33, 19 July 2014 (UTC)

Class vs instance - Languoid[edit]

I made:

subclasses of languoid (Q17376908). Current links to that item: Special:WhatLinksHere/Q17376908.

Now using languoid for data testing:

  • Subclasses in autolist2: 159 [13]
  • Instances in autolist2: 8028 [14]

Since a language has no location at a given point in time, the instances are probably all false. Tamawashi (talk) 14:05, 18 July 2014 (UTC)

Correct. Spoken/read language is an active process and written language is an unexecuted process, and as such they cannot be instances (unless we refer to a particular word written or spoken). The funny thing is that the label of Property:P31 in German is "ist ein(e)" ("is a"), label that is normally reserved for Property:P279. It would make more sense to have more user-friendly labels, but I'm not sure if we have enough social capital to find consensus.--Micru (talk) 14:20, 18 July 2014 (UTC)
What about e.g. dialect continuum (Q215844) (View with Reasonator)? Until now I've been using instance of: language (Q315) (View with Reasonator) (also known as "dialect") for dialects too, is this incorrect? It has "dialect" as an alias, so I thought that's what you were supposed to use... πr2 (t • c) 19:27, 18 July 2014 (UTC)
Dialect continuum is one type of Languoid. There are a number of dialect continuums which are 'instances of' 'dialect continuum'. This means 'dialect continuum' is a class, not an instance and should have the 'subclass of (P279)' property, not instance of (P31). Filceolaire (talk) 20:35, 18 July 2014 (UTC)
I edited 'Languoid' to add a description, from Wiktionary, and make it a subclass of 'abstract object'. Filceolaire (talk) 20:35, 18 July 2014 (UTC)
I think we are safe with the class/token relationship using the following definition : a language is a set of conventions used by humans to communicate. Then languages are localized in time and space : at a given moment a set of people can communicate because they share these conventions. Then <modern french> is the set of conventions in use right now for french speaking people. <québécois> is very close, so those two languages belongs in a tight language family (a class of languages). TomT0m (talk) 14:03, 19 July 2014 (UTC)

Class vs instance - programming language[edit]

Hi there,
I see that you are actually removing instance of (P31) and add subclass of (P279) instead on items about software or programming languages. I would ask you for stopping that, because its wrong. Your edits would be reverted like I did already ([15]). There is a consent that instance of (P31) would be used for this. Greetings, --#Reaper (talk) 00:00, 20 July 2014 (UTC)
PS: During writing this may already got obsolete. (See here.) Posting just for information.

@Reaper35: - please show where I added "subclass of a software" recently. Tamawashi (talk) 00:07, 20 July 2014 (UTC)
I'm sorry, looked to fast at the item labels. I must correct myself, I only mean programming languages. --#Reaper (talk) 00:17, 20 July 2014 (UTC)
@Reaper35: - Thank you. At least one editor here, that corrects himself. Tamawashi (talk) 00:21, 20 July 2014 (UTC)
@Zolo, Micru, Filceolaire, Emw: - see above, another instance of use of "instance of" instead of "subclass of". Sadly some abusive admin blocked me, so I cannot talk about that in the project chat, but it would belong to Wikidata:Project chat#Class vs instance - Languoid. Tamawashi (talk) 00:07, 20 July 2014 (UTC)
The topic in the project chat is a bit different in my understanding. At least to say: subclass is rather to understand as "(sub)type of", so subclass won't fit for these topics and its items. I'll take another look at the long discussion at WD:AN tomorrow. Good night, --#Reaper (talk) 00:17, 20 July 2014 (UTC)
@Reaper35: Why? It is about languages, abstract objects. Apart from that, many have different versions, e.g. there is no single PHP language. Tamawashi (talk) 00:21, 20 July 2014 (UTC)
There was already an discussion about abstract objects (or/like virtual things) and "instance of" at the beginning of both properties a bit more than a year ago. The consent is, that also these object could be an instance. (Otherwise there could never be an instance of an (programming) language (or software too, btw.)) At least there is only one actual definition of (as example) the PHP language. Older Versions aren't part of / would be "ignored" by this project (Wikidata), as long there no special item or other need for them. (Btw: The definition and its official implementation is here placed into one item, too. Unlike C++ or C. But thats would be ok.)
So this would also fit for the "real" languages (ok, at least I'm not familiar with this science) and would also fit with websites and similar you edited recently.
Greetings, --#Reaper (talk) 00:37, 20 July 2014 (UTC) (PS: No need to ping me (twice too) on this page, I'm watching it now.)

User:Reaper35 claims Java is an instance of a programming language:

I see that you are blocked but replying quickly before I forget. I am not a computer scientist, but from the definitions I find on Wikipedia. Java is an instance of programming language. Java SE8 is an instance of computer language version. Generic Java is an instance of programming language dialect and the bits of memory that allow my computer to run Java is an instance of machine implementation of a programming language (or whatever it is called, it is not an instance of programming language, as a programming language is an abstract object). --Zolo (talk) 08:08, 20 July 2014 (UTC)

@Zolo, Micru, Emw: Wikipedia is not reliable source for how classification is done in Wikidata. Zolo, is Java SE8 a subclass of Java? If yes, then Java is a class, right? Tamawashi (talk) 01:17, 21 July 2014 (UTC)

Help with moving task forces to WikiProjects[edit]

Hi all,

I'm putting the call out for folks to help move/rename their task forces as WikiProjects. I'm asking because right now we are using both names in a seemingly arbitrary way and it is important to keep this consistent so as not to confuse people (in fact, this issue came up as something that should be discussed on Help:FAQ when I polled the community about sources of confusion when using Wikidata).

This issue was previously discussed on Project chat about a year ago (see here for the discussion) when it was proposed to rename task forces as WikiProjects because of the naming conventions of other sister projects (i.e. "WikiProject" is the term used on Wikisource, Wikipedia, etc.) and because the term "task force" was not a concept easily translated into other languages.There was also a bot request for assisting with this process but it appears it was decided in the end that we should just move task forces one by one.

For those wanting to help, I recommend that something like "XX task force" be renamed to "WikiProject XX". The WikiProject should then be removed from Category:Task forces and added to Category:WikiProjects.

I will take on moving Wikidata:LGBT_task_force and am happy to edit appropriate Help pages to reflect and document this change, but need help with the rest. Please let know if there are questions or concerns related to this proposal. Thanks. -Thepwnco (talk) 17:12, 18 July 2014 (UTC)

The page Wikidata:Task forces needs moving. The categories which sort the WikiProjects should probably be called e.g. Category:Geographical WikiProjects, the same as done on en.wikipedia, rather than Category:WikiProjects Geography which seems to be the pattern we're heading for now. Delsion23 (talk) 08:38, 19 July 2014 (UTC)
Thepwnco: I made you a todo list (query). It currently contains about 650 pages. Multichill (talk) 10:43, 19 July 2014 (UTC)

Wikidata is a Wikimedia Project. Now inside there are WikiProjects. The latter have pages in the project namespace of the Wikimedia Project "Wikidata". The project namespace of the Wikimedia Project "Wikidata" is named Wikidata. Tamawashi (talk) 16:45, 19 July 2014 (UTC)

@Delusion23: thank you for pointing this out - I have now added those recommendations to Wikidata:WikiProjects (which did not have much in the way of information on category use).
@Multichill: thank you for the to-do list! Do you think this is worth adding to a new section or subpage of Wikidata:WikiProjects?
-Thepwnco (talk) 00:11, 22 July 2014 (UTC)
And Category:Completed_task_forces --> Category:Completed_WikiProjects please... --Liuxinyu970226 (talk) 10:51, 23 July 2014 (UTC)

Difference[edit]

What is the difference between female animal (Q43445) and female (Q1304841), and male animal (Q44148) and male (Q1957296)? Delsion23 (talk) 17:54, 18 July 2014 (UTC)

Hmm... no idea why the English label was changed from female to female animal... --Stryn (talk) 20:09, 18 July 2014 (UTC)
I think I have worked it out. The latter are about the diminutive word. I have changed the en labels and descriptions to reflect this. Delsion23 (talk) 20:57, 18 July 2014 (UTC)
No. It's not a diminutive. In English Male and Female are used to refer to both animals and to people. In some other languages there is a distinction, with different words used. See the description of sex or gender (P21) for how these items are used with that property. Filceolaire (talk) 23:12, 19 July 2014 (UTC)
See Wikidata:Project chat/Archive/2013/03#Statement p21 (sex) should not use q43445 (female) as value. LaddΩ chat ;) 19:19, 20 July 2014 (UTC)
@Filceolaire: I thought we use female (Q6581072) for female human, not female (Q1304841)? Q6581072 is the one linked to through P21. 130.88.141.34 08:55, 21 July 2014 (UTC)

Linked Data Interface[edit]

According to the Linked Data Interface data about an item or property can be obtained from its URI, where the file format of the data can be selected through content negotiation by extending the URI with a postfix to indicate the file type, such as .json, .rdf, or .nt. For example http://www.wikidata.org/entity/Q42.rdf.

However, the information contained in http://www.wikidata.org/entity/Q42.rdf basically contains only label-related data in all languages. What is missing now, are statements that relate this item to other items like "place_of_birth Cambridge". The corresponding wiki URI, i.e.https://www.wikidata.org/wiki/Q42, contains more informations - in particular the section about Statements.

Linking only to labels shouldn't be called "Linked Data" since the labels are dead ends or sinks, i.e. literals do not have properties that would point to other items. In other words, the network work of this "Linked Data Interace" consists of isolated islands, because the connecting properties are not rendered.

Either I am missing something here or the "Linked Data interface" should be supplemented with all the missing properties. Pingking (talk) 23:25, 18 July 2014 (UTC)

Yeah it is simply not finished :) Help welcome. Tracking happens at bugzilla:48143 --Lydia Pintscher (WMDE) (talk) 07:08, 19 July 2014 (UTC)

Wikidata for non-WMF wikis[edit]

Hi from Czech Republic! I would like to ask wikidata devs, if it will be possible (or already is) to use wikidata on non-WMF projects. The outsider wiki I come from uses already medua from commons and cooperates happily with wikipedia in other matters and would like to extend this collaboration with WMF projects by including information stored at wikidata in our infoboxes. Thank you for your answer and a have nice weekend. --Wesalius (talk) 05:41, 19 July 2014 (UTC)

It is not possible for a 3rd-party wiki to access the data here in the same way Wikipedia does yet. It is a long-term goal to enable this. What is possible is for you to hack something up that uses the API and gets the data. --Lydia Pintscher (WMDE) (talk) 07:06, 19 July 2014 (UTC)

Increase the minimum support votes for WD:RFA[edit]

I think we are now such a large community that it'd be safe to raise the number to 10 or 12. --Ricordisamoa 13:03, 19 July 2014 (UTC)

I agree. I've also thought about if we could make some rule to not accept users without any edits on Wikidata to vote. --Stryn (talk) 17:32, 19 July 2014 (UTC)
  • This needs an RfC, it can't be raised by a project chat discussion alone. But I do agree - I would even suggest 15 (12 is not that much greater than 10) and 20 for bureaucrats (do not change CU and OS).--Jasper Deng (talk) 18:04, 19 July 2014 (UTC)
not broken and does not need the change. If community is big enough, there will be enough votes pro and contra. If we have big community and not enough vote -- it means nobody cares. So, no big deal. I don't like the idea to force people to care. But, i agree to Stryn, votes should be required to have some qualification. -- Vlsergey (talk) 18:26, 19 July 2014 (UTC)
I don't really think we need a change in support levels but I won't be against implementing them. re. voting requirements - yes. John F. Lewis (talk) 18:35, 19 July 2014 (UTC)
+1. --AmaryllisGardener talk 20:15, 19 July 2014 (UTC)
I can not recollect recent closing problems because of the low participation, and I would not object rising the number of minimum votes, on the other hand I do not see any particular need. It should go through an RFC anyway after this preliminary discussion.--Ymblanter (talk) 20:58, 19 July 2014 (UTC)
I would think 12-13 is enough. Jianhui67 talkcontribs 01:38, 20 July 2014 (UTC)
Per Ymblanter. --Rschen7754 06:18, 23 July 2014 (UTC)
Can't see much of a need, but wouldn't be vehemently opposed to it. TCN7JM 06:21, 23 July 2014 (UTC)
Don't see a need either. More important is that there aren't too many "oppose" votes. Lymantria (talk) 07:15, 23 July 2014 (UTC)

Compare categories[edit]

Is there any tool that can compare the contents of categories on different Wikipedias? For example, items which have links in sv:Kategori:Diptera yet also links in en:Category:Plants. This would check for incorrect links. Delsion23 (talk) 22:42, 19 July 2014 (UTC)

Try User:JVbot/wikipedia-sync.py.
syntax: wikipedia-sync.py -qid:Q123 -once JAn Dudík (talk) 09:18, 22 July 2014 (UTC)

Disambiguation pages mixed with normal pages[edit]

This autolist page contains items that claim "instance of = Wikimedia disambiguation page" but also have other common properties not usually in items that are disambiguation pages. There are over 1,800 that need working on. Most will be because normal articles and disambiguation pages have been mixed together and so need separating. Any help in tackling this would be appreciated. Cheers. Delsion23 (talk) 00:22, 20 July 2014 (UTC)

This list appears to overlap with this list of Kazakh articles on places, which are contained in disambiguation items. Over 1100 of them :( Delsion23 (talk) 13:28, 20 July 2014 (UTC)
BTW: The autolist list there is already down to 1650 items. --Bthfan (talk) 10:05, 21 July 2014 (UTC)
Down to 1603 now :) 250 down in less than 2 days, great progress everyone! Cheers. Delsion23 (talk) 18:30, 21 July 2014 (UTC)
The mostly easy ones are probably done now ;) (though there are still quite a few left). Most items I fixed were items where multiple Wikipedia links about different places were within the same item. Easy items to fix were disambiguation pages about places in USA or at least having an English article (or German as well in my case). There you could compare census data or various geodata IDs from the Wikipedia articles to sort this out which Wikipedia article belongs to which item. Places/location infoboxen were also useful to compare the various Wikipedia pages in a structured way. I also noticed a few geodata errors in the Wikipedia articles and notified the Wikipedia editors of this. --Bthfan (talk) 19:01, 21 July 2014 (UTC)
Good idea. If the mistake is in the Wikipedia article, it will only creep its way back into Wikidata over time. I've found that the tt and ce.wikipedia ones have been the most challenging due to neither language being translated by Google translate. Down to 1,553 now :) Delsion23 (talk) 22:33, 21 July 2014 (UTC)
htwiki seems to have quite a few bad articles regaring location/places in the USA. Looks like they did some mass-import of places a while ago, those articles lack almost any details (you sometimes don't even know what location an article is now about, those roughly say "[foo] is a town in California, USA" ;). These articles I just kept on the disambiguation page item. --Bthfan (talk) 22:50, 21 July 2014 (UTC)

Selecting male is no longer easy and obvious[edit]

Hoi, the "male" option for "sex/gender" is no longer part of the options on the first list. It is an additional value.. This makes it rather tedious to make humans male.. Thanks, GerardM (talk) 09:59, 20 July 2014 (UTC)

@Lydia Pintscher (WMDE): I think that there is a group working on this in the development team. They are the same people that made it so that common properties appear based on popularity (a great improvement for finding and entering missing statements). I think the same thing will eventually be true for items for each property e.g. the two most common items "male" and "female" will appear at the top for sex/gender. This makes it easier for users to find, and less likely for newbies to select the wrong ones. Delsion23 (talk) 10:09, 20 July 2014 (UTC)
Correct :) --Lydia Pintscher (WMDE) (talk) 10:13, 20 July 2014 (UTC)
@Lydia Pintscher (WMDE): Eventually is of little use, now; especially for such a high-usage parameter/ value pair. This needs a fix, ASAP, please. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 15:28, 20 July 2014 (UTC)
@GerardM: A work-around is to tag them as "male animal" (the two should be merged anyway; comments passim) and a script can later convert them. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 16:37, 20 July 2014 (UTC)
@GerardM, Pigsonthewing, Delusion23, Filceolaire, Stryn: Please ensure you read read Wikidata:Project chat/Archive/2013/03#Statement p21 (sex) should not use q43445 (female) as value before merging these sex items. LaddΩ chat ;) 19:14, 20 July 2014 (UTC)
@Laddo: I've read it a long time ago, and I'm well aware of it. Thanks anyway for reminding. --Stryn (talk) 19:17, 20 July 2014 (UTC)
I am also aware of it and have no desire to merge the items. Cheers. Delsion23 (talk) 19:53, 20 July 2014 (UTC)
Fair, just wanted to be sure you guys knew about these older discussions. LaddΩ chat ;) 22:26, 20 July 2014 (UTC)
The reason seems to be that male was added as an alias to male animal. I suggest to remove that again. --Lydia Pintscher (WMDE) (talk) 10:09, 21 July 2014 (UTC)

Finding Twitter users[edit]

How may find (in a human readable form) all the people (not organisations/ brands) who have a Twitter account listed in Wikidata? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 15:53, 20 July 2014 (UTC)

Here is a list on autolist that includes everyone with "instance of = human" and "website account on = Twitter". Hope this helps. Delsion23 (talk) 16:31, 20 July 2014 (UTC)
That's very helpful, thank you. I wonder if it's possible to go one step further, and get a list of their Twitter IDs? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 16:38, 20 July 2014 (UTC)
In AutoList you can specify values that you want to see.. Thanks, GerardM (talk) 17:55, 20 July 2014 (UTC)
How? πr2 (t • c) 04:42, 22 July 2014 (UTC)

Many cities, towns, places, rivers, etc. from Germany are located in Europe too ...[edit]

Hi! @Andreasmperu / Andreasm, @Ladsgroup / Amir: claim[17:183] AND claim[30:46] is just a query. I analyzed more then ten thousand places searching twin places.
While most items about communities from Italy contain is in the administrative territorial entity (P131) and the proper province and region those from other countries as France contain region and department on an "arbitrary base". Some places from Greece, etc. did not contain the country (P17) statements. Some city subdivisions called suburbs, districts etc. depending on the countries do not contain all relevant is in the administrative territorial entity (P131).
questions:

a) What is the today policy about is in the administrative territorial entity (P131) values. Should all relevant values be added? These values are hierarchical. Are they validated?
b) Is there a bot support for adding / validating the values.
c) Which is the best place to maintain a to do list?

Regards gangLeri לערי ריינהארט (talk) 19:28, 20 July 2014 (UTC)

User:לערי ריינהארט - Wikidata:Administrative territorial entity provides some documentation. If you have a query that checks if P17 is there, then add it to the page. Maybe you also have other observations that could be added to the page. There are now 1724 subclasses. Maybe Wikidata:Database reports/Administrative territorial entity has some value for you. Regarding P131 - why not use the talk page there? Tamawashi (talk) 00:47, 21 July 2014 (UTC)
Thanks @Tamawashi for your comments. The title is ironic / sarcastic / cynic etc. It seems that WD is a source of intergalactic wisdom ... Why do we need 1,600++ different abstractions at Wikidata:Database reports/Administrative territorial entity? Probably to optimize querries ... לערי ריינהארט (talk) 19:43, 22 July 2014 (UTC)
There a lot of different meanings of some terms, e.g. what "town" means, depending on the country/state. Maybe some could be removed though, not sure. --Bthfan (talk) 20:04, 22 July 2014 (UTC)

Two dates of death[edit]

Yashiki Takajin (Q3545819) has two dates of death, Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 22:14, 20 July 2014 (UTC)

I resolved that and added a source for the correct date of death --Bthfan (talk) 07:26, 21 July 2014 (UTC)

Rugby identifiers[edit]

Can we create in bulk some rugby-related properties? All of them would be String datatype:

  • Barbarians
  • All Blacks
  • British Lions
  • French Barbarians
  • Pro12
  • ErcRugby
  • ItsRugby

Were they already proposed by someone else? When and how to make such a proposal? AuthorityControl/People? --Ricordisamoa 22:28, 20 July 2014 (UTC)

What do you mean with rugby-related properties here? Those names look like rugby teams to me, so you could just create new items for them...? Or do you mean certain properties are missing to describe those teams? --Bthfan (talk) 06:43, 21 July 2014 (UTC)
@Bthfan: those are also names of identifiers for rugby players, as used by it:Template:Link statistiche rugbisti a 15. ✓ proposed as "person". --Ricordisamoa 15:22, 21 July 2014 (UTC)

Property proposals are at WD:Property proposal. There are Property:P861 and Property:P858 already, these kind of things can be done. 130.88.141.34 09:24, 21 July 2014 (UTC)

Blocked by two admins, where was the "pattern of local abuse"?[edit]


You are currently unable to edit Wikidata.

You are still able to view pages and data entries, but you are now not able to edit, move, or create them.

Editing from Tamawashi has been blocked (disabled) by John F. Lewis for the following reason(s):

Inappropriate use of user talk page while blocked: Refusal to listen/acknowledge comments by users, see also this AN discussion

This block (ID #6387) has been set to expire: 23:52, 20 July 2014.

Even if blocked, you will usually still be able to edit your user talk page and email other editors. To discuss the block, you may contact John F. Lewis or another administrator.


When following the link labeled "blocked", one comes to the page "Wikidata:Blocking policy", which says:

Administrators may block user accounts or IP addresses:

  • To prevent local abuse where a pattern of local abuse has been established. Local abuse includes, but is not limited to:
  • When consensus to block a user or IP has been developed on a discussion page, with at least a week of discussion and clear consensus on administrators' noticeboard.

QUESTION: Where was the pattern of local abuse? Where is "local abuse" defined? Tamawashi (talk) 00:29, 21 July 2014 (UTC)

@Tamawashi: First thing which people never understand; we don't have a blocking policy. Now, refusal to work in a co-operative and community environment by repeatedly removing questions and warnings is just a major no. The AN section in question pretty much shows me a consensus among the administrators that there is an issue here which we can't resolve because you refuse to communicate with us. On a side note; unless you have something real to discuss - I ask you to stop carrying on the behaviour which resulted in your talk page access being revoked. Thank you, John F. Lewis (talk) 00:33, 21 July 2014 (UTC)
My reply:
  1. OK, no blocking policy, admins can do what they like to. Thank you for at least admitting this.
  2. I ask you to stop carrying on the behaviour which resulted in your talk page access being revoked. Can you define such behavior? Oh, but wait, there is no blocking policy, i.e. talk page access can be randomly removed, it wouldn't follow any defined behavior.
  3. "consensus among the administrators that there is an issue here" - ah yes? A consensus among those with superior rights to keep users down that don't follow their random commands?
I made 300000 contributions in one month, I am at 200000 in the second month. I am discussing EVERY content issue, 1) if there is disagreement 2) if I am aware of the existence of such 3) if the discussion is lead in English. It is pure defamation to talk of "refusal to work in a co-operative and community environment". You are now at least the second admin to post defamatory statements. 01:06, 21 July 2014 (UTC)
The comments by these two admins are not defamation under any standards. Please refrain from using that word lacking a proper context. Now, we do have a blocking policy and is called common sense. We are here to take care of a wiki and make sure everyone can work and discuss matters on a timely and friendly manner. This is not the behaviour you have shown, and as such you received warnings and, ultimately, a block. — ΛΧΣ21 01:11, 21 July 2014 (UTC)

@Tamawashi: I'll be completely blunt about this. In addition to John's and Hahc21's comments above, you appear to not assume good faith nor assume that we do with you. "Defamation" and "trolling" accusations (such as here, against @TomT0m:) fundamentally fail to do either, without substantial evidence. As for blocking you, I do not know of any administrator who opposed the block of you, nor did I see any other users comment to that effect. It is true that our blocking policy needs to be actually fleshed out, but en:WP:NOTTHEM behavior is invariably frowned upon by the community and not an appropriate use of your talk page while blocked. My honest advice for you is to just drop the stick, because your behavior, if you choose to continue it, will lead to further blocks.--Jasper Deng (talk) 01:49, 21 July 2014 (UTC)

User:Jasper Deng - you again fail to properly parse text. It was TomT0m that made the troll allegation against me. So, now the way you treated me, go to TomT0m and treat him like that? Let's see, if you are fair, and apply your standards to everyone. "en:WP:NOTTHEM" - First, this is Wikidata not English Wikipedia. Second, why do you refuse to talk about your behavior? It was a separate issue. My behavior is mine, yours is yours. Putting the NOTTHEM sticker all the time makes you look a little bit like a hypocrite, since when a discussion about your behavior is started then you use NOTTHEM to evade the discussion. Regarding "drop the stick" - well, you chose to attack me in the first place, by stating "I will also say that Tamawashi's been not so careful with his editing of items" at the AN [16], without providing evidence that I was not careful. Tamawashi (talk) 02:04, 21 July 2014 (UTC)
Can we please desist from this matter now? No more mentions of it, no more accusations etc. that goes for both of you (Tamawashi and Jasper Deng). I shouldn't really have to say this but if you make another accusation whether provoked or not Tamawashi, I'm afraid we may have to block you again for carrying on behaviour immediately after your block expired. Take this as a final warning please. John F. Lewis (talk) 02:10, 21 July 2014 (UTC)
May I ask, why threatening with a block is only applied to Tamawashi, and not to other people? Can they make accusations against Tamawashi, but if he accuses them of unfair accusations, then he gets blocked, irrespective of whether he was right or not? I would like to drop the whole matter, but I FEEL TREATED UNFAIRLY. Tamawashi (talk) 02:17, 21 July 2014 (UTC)
This is nothing about unfair treatment. The mention of a block with regards to you is only because you came off a 24 hour block and immediately started this thread and began to accuse another user of trolling, therefore you came off a block and started to be disruptive again. If Jasper Deng just came off a block and did this; he would get the same response as you. John F. Lewis (talk) 02:21, 21 July 2014 (UTC)

Section break[edit]

This still is not answered: "Where was the pattern of local abuse?" - Any idea why? Any diff? Tamawashi (talk) 02:13, 21 July 2014 (UTC)

There does not need to be a pattern of local abuse. The block comes up 'consensus to block at WD:AN' and general discretion I exercised by WD:UCS. Now as I said, please desist from this matter now. John F. Lewis (talk) 02:15, 21 July 2014 (UTC)
Thanks for the clarification that "pattern of local abuse" was not the reason. Tamawashi (talk) 02:22, 21 July 2014 (UTC)
Great, can we drop the matter now and get back to focusing on editing? John F. Lewis (talk) 02:23, 21 July 2014 (UTC)

Profile[edit]

Oluwole 'Segun Michael A native of Iloro-Ekiti in Ijero LGA area of Ekiti State, South West Nigeria.

Item needs splitting?[edit]

Does the item Jeanna Giese (Q261564) need splitting? Some link are about a person, the others are about a protocol. 130.88.141.34 10:08, 21 July 2014 (UTC)

✓ Done into Milwaukee protocol (Q17394776) --Ricordisamoa 15:56, 21 July 2014 (UTC)
Thanks :) 130.88.141.34 16:05, 21 July 2014 (UTC)

ResearchGate[edit]

Does anybody know if Reasearch Gate has some identifiers we can extract? The have a lot of pages about institutions: https://www.researchgate.net/institution/Universitaet_Potsdam (University of Potsdam (Q153012)), and also about people https://www.researchgate.net/researcher/21718598_R_P_Feynman (Richard Feynman (Q39246)). The number that appears before the name, doesn't seem to be an identifier, because it is missing from many pages, but the number alone will also link to the correct page (https://www.researchgate.net/researcher/21718598). They also seem to have some machine created pages that have misspellings: https://www.researchgate.net/researcher/2033091646_R_P_Frynman. -Tobias1984 (talk) 12:29, 21 July 2014 (UTC)

Use of Wikidata content in Wikipedia so far[edit]

Hello, since I just started using Wikidata, for me the question arises, how widely is content from Wikidata used in Wikipedia pages yet?  – The preceding unsigned comment was added by Alepfu (talk • contribs).

Have a look at the templates in the category Category:Templates using data from Wikidata (Q11985372). It might not be complete, but you get a good impression. --Pasleim (talk) 15:20, 21 July 2014 (UTC)
@Alepfu: (edit conflict) From the looks of the results of this RfC on enwiki, it's appropriate to modify existing infoboxes to permit Wikidata inclusion when there is no existing English Wikipedia data for a specific field in the infobox, but it's inappropriate to use Wikidata in article text on English Wikipedia. So, Wikidata's not widely used on WP yet. --AmaryllisGardener talk 15:23, 21 July 2014 (UTC)
The situation in French Wikipedia but I think it is similar in others: Wikidata is not much used in infoboxes yet. But it is relatively widely used for external links like VIAF identifier (P214) and for geographical coordinates --Zolo (talk) 20:27, 21 July 2014 (UTC)

Adding TimeValue with Autolist[edit]

I want to add some dates using Autolist and and Wikipedia categories. Problem is I don't know how to format a date statement to efficiently add this information to Wikidata, I can't find any documentation for this and the interface doesn't allow me to see how they are formatted. How would I, for example, format a statement to add date of publication (P577) to a bunch of film (Q11424)s that were released in 1944? Is it at all possible? Väsk (talk) 15:59, 21 July 2014 (UTC)

@Väsk:. I do not think there is any way to add a date from autolist at the moment. Autolist2 is an independent tool developed by a volunteer (user:Magnus Manske), not a built-in feature of Wikidata, so it cannot always support all the features of Wikidata's software. :). --Zolo (talk) 13:38, 22 July 2014 (UTC)
You can do it with QuickStatements; you'll need to prepare a "tabbed" text to paste in. --Magnus Manske (talk) 15:44, 22 July 2014 (UTC)

MediaWiki:Gadget-DraggableSitelinks.js[edit]

It is an improved and working version of User:Inductiveload/scripts/draggableSitelinks.js by Inductiveload. If you agree, I'd make it a real gadget. --Ricordisamoa 16:49, 21 July 2014 (UTC)

first draft for Main page redesign ready for review[edit]

Hello again,

thanks for everyone who left feedback on the Main page redesign planning page. As promised, I've put together a first draft for the new Main page at Wikidata:Portal_Redesign/draft and would love to hear your comments and suggestions - please leave your feedback on the talk page.

Major changes include the following:

  • moved language links down the very bottom of the page (currently they eat up a lot of important real estate at the top of the page)
  • updated the "quick links" at the top of the banner/header (currently we just link to Statistics)
  • added in a visual right at the top of the page - this is just a placeholder and could either function as a banner OR a rotating/dynamic slideshow (see Wikivoyage for an example of a main page with both a banner and slideshow)
  • added in other images and screenshots; updated link for "Wikidata in action" item to one in Reasonator that is more visually appealing
  • added a new section, "Learn about data," that will ideally communicate what structured data is to those who don't already know (note: supporting content is still under construction)
  • added a new section, "Discover," with the intention of highlighting and/or featuring tools, applications, queries, WikiProjects initiatives, showcase items, etc.
  • removed "Use Wikidata on your wiki" section (it's important content but I don't think necessary to have on the main page, at least not at this stage in the project)
  • merged "News" and "Staying up-to-date with development" sections together
  • re-named "Contribute" section to "Get involved" - got rid of some of the links and clustered them together to make it a more manageable list
  • made "Welcome!" section one column that takes up half the page instead of one that spans the entire length of the page
  • made all sections one colour - green. This was just easier to work with as I moved sections around to different places but I thought I'd leave as is while collecting feedback

Just a note - the placeholder image at the top (File:Wikidata_POTW_Candidate_11_(Plain_Map).JPG) comes from Wikidata:Picture_of_the_week - other ideas for infographics or images that would make good banners are more than welcome. While I like the placeholder one and think it's both visually appealing and relevant to conveying some of what Wikidata can do, it also does look very similar to the Wikivoyage banner.

Looking forward to your comments. -Thepwnco (talk) 20:01, 21 July 2014 (UTC)

The draft of the main portal seems good but the picture is more similar to the propagation of a virus than something about knowledge. I would prefer another picture or at least to modify the colors of the proposed picture. Snipre (talk) 11:23, 22 July 2014 (UTC)
The draft of the main page is a step in the right direction. I have three pointers. Firstly, like snipre mentioned the map picture is neither relevant nor helping the overall look of the page. Secondly, I´m not getting the point of the arrows between the links to Introducing Wikidata, Project chat, etc. Personally I would just put each of those links to their own line. Thirdly, I would swap the placements of the discover and get involved boxes. At least at my current resolution (1024x768) the placement of those two boxes create a big area of whitespace. Also, the position the discover box is at currently seems to be in a prominent place. I don´t really get why, since a minority of users are interested in the technical side of things.--Snaevar (talk) 14:09, 22 July 2014 (UTC)

Again mobbing by a user[edit]


Programming a subclass of object orientation but not of activity?[edit]

Moved to Wikidata talk:Item classification after concerns of venue by User:Pasleim Tamawashi (talk) 21:30, 21 July 2014 (UTC)

Language links incomplete in mobile view[edit]

In normal page view, en:Tin Can Cathedral shows a link to the Danish (dansk) page da:Blikdåsekatedralen and vice versa, which seems good.

But in mobile view, en:Tin Can Cathedral language links does not show a link to the Danish page. However, English seems to be available from the Danish mobile view. Seems to be a glitch in how mobile view, at least in English, is processing the Wikidata. Dl2000 (talk) 21:33, 21 July 2014 (UTC)

This needs a bug in Bugzilla. Thanks, GerardM (talk) 11:29, 22 July 2014 (UTC)

General question regarding subjects of important works[edit]

Hi I have been updating the items for people portrayed by Frans Hals, but I notice that it's sometimes a bit blurry whether an article is about the person portrayed or about the creative work (in this case, a portrait). To solve these, I have been creating two items, one for the person and one for the creative work. See Aletta Hannemans: Q16859703 and Q17275957. With the portrait painting, I want to link to the pendant (for which I use the property "part of"). With the person, I want to link to her family (and soon also, the house where her brewery was). My question is, does she really need two items in this case? I want to be sure before I do more of these. Thanks in advance. Jane023 (talk) 07:11, 22 July 2014 (UTC)

Yes, you should create two separate items like you did here. Multichill (talk) 17:28, 22 July 2014 (UTC)
OK Thanks! Jane023 (talk) 20:20, 22 July 2014 (UTC)

Wikinews Sitelinks (Phase 1)[edit]

Greetings everyone!

As planned Wikinews will be the next sister project supported by Wikidata. We're starting with language links again and will later enable data access. Phase 1 is currently scheduled for August 19th. Depending on how that goes the next phase will scheduled. Coordination is happening at Wikidata:Wikinews and it'd be awesome if you added your name there if you're active on Wikinews. We need people to act as local ambassadors to make sure everything goes smoothly.

Let's make this deployment even better than Wikiquote was!

On behalf of the development team, John F. Lewis (talk) 18:57, 22 July 2014 (UTC)

How will we deal with linking news stories and notable events? --Jakob (talk) 19:23, 22 July 2014 (UTC)
Over at en.wn, I remember we discussed Wikidata. Perhaps aspects of the discussion are slipping my mind. My memory (however flawed it might turn out to be) is we generally didn't see much utility for it in mainspace; and from what I know of other-language Wikinews category spaces, they often have such disparate category hierarchies that the prospects didn't seem very promising there.
I'd be interested to know quite what is meant to be done. To the best of my definitely limited knowledge Wikidata seems to be a vehicle of encyclopedic structure, while Wikinews is about as different from an encyclopedia as a wiki can get. Just for example, our concept of newsworthiness is very different from Wikipedia's notability; and Wikinews is usually more interested in covering under-reported stories (and otherwise unreported ones, via original reporting) than in covering the biggest stories of the day, which typically have less need of our coverage. (Keep in mind, being involved with news, I'm supposed to be skeptical aobut everything; I'm also supposed to have an open mind, and with all the bad news in the world I surely need to stay open to a positive outlook.) --Pi zero (talk) 12:15, 23 July 2014 (UTC)
I might mention, Wikinews categories quite often do not correspond, or correspond closely, to Wikipedia categories, as the very purpose of a news category hierarchy is different from that of an encylclopedia category hierarchy. Sister-linking Wikinews categories tends to be highly customized, and likely to only get more so over time as we get around to various types of hierarchy upgrades we've been planning. --Pi zero (talk) 12:39, 23 July 2014 (UTC)

What are the aims of the Wikidata project for Wikipedia interlanguage links?[edit]

I am looking for a Wikidata page that describes the aims of the Wikidata project for Wikipedia interlanguage links. What should it provide? Of course it should provide links between different languages of Wikipedia. But what characterizes two articles of two Wikipedias-languages to have an Interwiki link on Wikidata? Could anybody give me a Wikidata link to such a description? Thank you! -- Tirkon (talk) 20:32, 22 July 2014 (UTC)

Does Help:Sitelinks help you? --YMS (talk) 07:32, 23 July 2014 (UTC)