Wikidata:Project chat/Archive/2016/04

This page is an archive. Please do not modify it. Use the current page, even to continue an old discussion.


Wikidata weekly summary #201

Instance of

  • "Identifying problematic statements in Wikidata via multi-level modeling theory" is a very important paper for us and shows us why it's a bad thing that our use of (predominantly) instance of (P31) is a bad thing. --Izno (talk) 13:48, 21 March 2016 (UTC)
    • And let me clarify: Using the instance of (P31) relation for things which are not instances (except where we use it as a metaclass statement). --Izno (talk) 16:02, 21 March 2016 (UTC)
      • @Izno: I think I understand your latter point, but for the benefit of me and others please can you explain/ give examples? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 16:35, 21 March 2016 (UTC)
        • @Pigsonthewing: Immediately to mind is any video game, where we use instance of (P31) video game (Q7889); this is not a true statement because video games (and books) as we document them on Wiki* are not instances but instead classes, where there are no useful instances to document. We have similar issues in disease (as I came across yesterday) and in genes, and in chemicals. An instantiation of a gene, for example, is the individual molecule representing a gene in my or your body, but we document genes on Wikidata as instance of gene rather than the proper subclass of gene. An area where we got it right is in our treatment of biological taxonomy, which treats taxa as instance of a taxon rank and subclass of organism (by inheritance). (Side note: That particular domain did do something wrong in using domain-specific properties where P31 and P279 would have worked, but that's not an interesting part of this discussion.) --Izno (talk) 18:43, 21 March 2016 (UTC)
          • Side note: Only a minor sidekick. --Succu (talk) 21:02, 21 March 2016 (UTC)
          • Obviously we've had extensive discussions about this sort of thing before - This recent RFC for instance. I don't believe we either have a consensus on the correctness of this sort of instance/subclass distinction, or even that the exclusive multi-level modeling described in this article (while interesting) is really a valid description of the world that we ought to emulate. If you've ever read, for example, Douglas Hofstadter's books (I'm thinking specifically of I Am a Strange Loop (Q3616600)) you'll realize this kind of sharp level distinction is not the way the world actually works, and intermixing of levels is natural. That's not to say that some of the modeling in wikidata is not currently wrong and in need of fixing - the "earthquake" example in the paper is a good one. But I'm not so sure about genes or chemicals, or even books or video games. I Am a Strange Loop (Q3616600) subclass of (P279) book (Q571) feels wrong, while instance of (P31) feels right. The confusion is I think because "book" has two meanings in English: a physical object that you can throw at the wall, and an abstract entity described by its title, author, and content irrespective of physical manifestation. I believe book (Q571) describes the latter, and instance of (P31) is the correct relationship in that case. But such ambiguities are almost inevitable and it may be quite correct for a wikidata entity to represent more than one meaning in this sort of sense. ArthurPSmith (talk) 20:18, 21 March 2016 (UTC)
            • @ArthurPSmith: The video game token can be considered to be (exactly as a normal game) as the class of all concrete games one ever played, the concrete experience. Then a video game like "Quake 1" is indeed a class of game. Each video game is a type of game experience, specifically related to a single set of code/data, more specifically, according to philosophycal definition virtual is a kind of potentiality that becomes fulfilled in the actual., like a crop is a virtual plant (see en:Virtuality_(philosophy) where we can see mentioned Charles Sanders Pierce, the same one of the type/token distinction) So to use instance of (P31) for video games I'd put something like :
              < Quake > instance of (P31)   < virtual game experience >
              , which seems totally philosophically justified and justifies as well the use of a class as if it was an instance, here it's a potential instance. Same constructions can be constructed for genes, I think. author  TomT0m / talk page 14:58, 22 March 2016 (UTC)
      • @Izno:, you've writeen examples of what you consider wrong. Can you explain why you consider it wrong? --Infovarius (talk) 11:08, 1 April 2016 (UTC)
    • I was wondering where they got the TBL sample from and was reassured that its P31 wasn't what they suggest.
      --- Jura 16:41, 21 March 2016 (UTC)
      • That may be because I fixed most of their samples (an IP got to one of them before me). --Izno (talk) 18:43, 21 March 2016 (UTC)
        • There is no trace of it for Q80#P31, but the value at Property:P106#P1647 is (currently) probably wrong and could have led to the sample.
          --- Jura 19:00, 21 March 2016 (UTC)
          • Right, they were also using subproperties in their queries. No, I don't think it's wrong--indeed, TBL is a computer scientist (and he is an instance, not a subclass), so that statement is right. But I'm not sure about computer scientist subclass of profession. I think it would be correct to say computer scientist subclass of professional. --Izno (talk) 23:24, 21 March 2016 (UTC)
            • That should be two separate items, one for "computer scientist" (person) and another one for "computer scientist" (profession). However, since it is too much effort to maintain two items for each, I think it is an acceptable compromise to just use "computer scientist" (profession) and instanciate as human.--Micru (talk) 08:27, 22 March 2016 (UTC)

The main problem is the definition of "instance of". For me an instance is "the most detailed item in a hierarchy in term of properties use". So even if it is possible to describe a single molecule of water in a glass or a single grain of rice in a sac, we don't have the properties to describe these two elements among the other ones. The location property is not able to define the exact position at time t and to distinguish one molecule/grain among the others. We can't describe this level of precision so why do we have to bother about that level in the hierarchy ? Then second question do we want to create items about unique molecule/grain of rice ? If not why again do we need to think about that level ?
An ontology is always depending on the purpose we want to reach. So we should adapt the classification according to what we want to do and not according to what it is possible to do. So if we don't want to describe water molecules we can consider water (Q283) as an instance of chemical compound (Q11173).
Then about the famous problem B is an subclass of A and C is an instance of B. Something it is worth to says that C is an instance of A too because this is a good way to be sure to have all instances of A in a query. Here the query tool should be able to deduce that all instances of B are instances of A too. Currently I don't know if SPARQL can do that with a simple query or if we have to specify that in the query request. Snipre (talk) 13:25, 22 March 2016 (UTC)

"For you": This is not the consensus meaning of P31, so I'm not sure why you even mention it as a point.

The problem is when we introduce a hypothetical "frozen water molecule". This is clearly a subclass of "water molecule", but if we say "water molecule" instance of "molecule"... we have an issue. This is clearly wrong.

I broadly agree with "purpose of reach", but our purpose is to be used. And to be used usefully, we must indicate that an item of interest is classified correctly. This is basic Help:Membership properties.

Your comment about "famous problem" is not in keeping with consensus best practice here (except in a handful of cases where there are sources calling out an item as a subclass/instance of one thing and a subclass/instance of another thing, for which we have Help:Ranks and Help:Sources and Help:Qualifiers, which should all be used together). I don't see a reason why tool support should have an impact here--SPARQL is good at its job. --Izno (talk) 13:48, 22 March 2016 (UTC)

I summarized points that looks a lot like this article one in Help:Classification, that I proposed as a project guideline. It's the same except they don't explicitely use the en:Type–token_distinction explicitely as the first level, and that they use more levels that I use in this document, but indeed their example of taxon ranks feels right to me at forst reading. author  TomT0m / talk page 14:58, 22 March 2016 (UTC)
@Izno: Thanks for your kind reception of my opinion in an open debate and I can say the same about your opinion when speaking about my opinion: I'm not sure why you even mention it as a point.
But your next words are more interesting. Your arguments are contradictory: you agree about purpose to reach but just after your speak about hypothetical water molecule. Isn't it better to solve now the question if we want to start to create items about special states of molecule ? You can't fix purposes if you want to keep everything open in case of hypothetical things. And if you are afraid to change the classification of some hundreds of elements in the case of higher granularity, I suppose you never heard about bots and their capacity to modify items.
And by the way how can you use instance now for humans ? Don't you plan one day to create more than one item for each human in order to specify their developement or particular states like X as child or X as teenager ? With your reasonning of hypothetical items you shouldn't use in any case instance because you can always think about one hypothetical state.
At least my definition doesn't required a help page to be explained: the instance in a set of items of the same species is the more detailed item according to the number of properties used to characterize it. And with that definition you know that the last element in an hierarchy is always an instance. Snipre (talk) 18:36, 22 March 2016 (UTC)
That's a self contained definition : you relate the definition of an instance only to the way we model stuffs in Wikidata. So this is pretty close to "an instance is an instance" ... That also fails to capture that both an instance of human and an instance of molecules are both concrete object. Plus, what do you do with instances of type of administrative division like french commune ? You just fail to capture that there is instance of relationship that are not really on the last level, don't you ? Last but not least : why would we need two properties if we just have to check if we are on the most detailed level to see wether or not we are speaking of an instance, and this would be way simpler to maintain. Why bother having instance of (P31) at all with your (non) definition ? author  TomT0m / talk page 21:21, 22 March 2016 (UTC)
TomT0m It is just a definition. But everyone can now see that even with a help page on how to use instance of/subclass of, the current approach is completely useless because no one understands it or takes care of it (the cited paper is really a proof than the current system doesn't work). You can always refer to big theories, standards or what you want, if people don't use it for any reasons, you are stuck.
I am pretty sure your reference to type–token distinction is not understandable for a certain proportion of contributors. WD is a collaborative project with hundreds of regular contributors and thousands of occasional contributors so if you want that contributors use a specific system it should be very, very simple. I don't want to speak hours about that because for me only the results count: the good system is the one which used and allows to provide the results we want. Snipre (talk) 12:50, 23 March 2016 (UTC)
@Snipre: Which result do we want ? A mess ? author  TomT0m / talk page 12:54, 23 March 2016 (UTC)
And please note that we can see all that as variations on things we already does and use a little better defined. And I really think anyone can understand those pretty simple principles, otherwise I would really not push it. There is a lot of nice features that allows this to be understandable and beautifully regular, and this regularity makes all this simpler. It answers questions for us. author  TomT0m / talk page 12:58, 23 March 2016 (UTC)
@TomT0m: The mess already exists so please see the reality: you promote one system since a long time with which results ? A paper citing WD as a mess in term of classification. That's facts. You are not responsible of that, this is not a judgment. So what can be done to improve the situation ?
We have several possibilities: continue with the current system or change it. My comment was just a proposition of change with a new perspective, not completely defined I agree but something different. My proposition is not accepted and even denigrated ? Not a problem I don't want to spend time for that thing. But now this is time to show that your proposition can improve the situation in WD in term of classification. I don't take care about what you think I take care about what people are doing: do contributors use the system described in Help:Classification in their contributions in WD ? Do they understand how they can use it ? That are the only real questions. Snipre (talk) 13:23, 23 March 2016 (UTC)
Well, I propose something to make it cleaner, with strong and clear principle. I want to spend time explaining and try to convince it adds something. It takes time but I'm not afraid of it. It's abvious newcomers don't understand this fully at first, but I don't think your definition helps them. Plus you did not understand the concerns about "if we adopt this, why should we have two properties ?" without any answer to this, I think you better should trat them as a synonym and don't try to get in the way of people who try to make sense out of it, this would be a good compromise and smoother for everybody. It would not remove anything. author  TomT0m / talk page 13:31, 23 March 2016 (UTC)

@Snipre: You might find this paper useful: Instances of Instances Modeled via Higher-Order Classes, Douglas Foxvog. Cheers, Bovlb (talk) 14:42, 26 March 2016 (UTC)

I tested anti-pattern 1 using the SPARQL:
select distinct * where { ?Z wdt:P31/wdt:P279* ?A . ?Z wdt:P279+ ?A . }
and I got 358,840 pairs. Here are the top few:
Bovlb (talk) 17:56, 26 March 2016 (UTC)
I tried again grouping by the ancestor:
 select (MIN(?Z) AS ?MINZ) (COUNT(*) AS ?COUNT) ?A where { ?Z wdt:P31/wdt:P279* ?A . ?Z wdt:P279+ ?A . } GROUP BY ?A ORDER BY DESC(?COUNT)
and got 2476 different values.
Click [expand] to view the content
Example descendant Count Ancestor
winner (P1346) 1143217 entity (Q35120)
defendant (P1591) 781330 object (Q488383)
defendant (P1591) 418608 abstract object (Q7184903)
economy of Djibouti (Q1000084) 406948 physical object (Q223557)
Commodore Amiga (Q100047) 371441 artificial entity (Q16686448)
Mandarin roll (Q1002439) 351902 work (Q386724)
Mandarin roll (Q1002439) 284272 goods (Q28877)
Mandarin roll (Q1002439) 284210 product (Q2424752)
Matkaoppaat (Q1000000) 283112 mental representation (Q2145290)
Matkaoppaat (Q1000000) 277544 concept (Q151885)
Matkaoppaat (Q1000000) 238642 point (Q44946)
Matkaoppaat (Q1000000) 238642 primitive notion (Q6453739)
Matkaoppaat (Q1000000) 196042 point in time (Q186408)
Matkaoppaat (Q1000000) 195706 occurrence (Q1190554)
Buddhism in Germany (Q1000850) 147904 process (Q3249551)
Buddhism in Germany (Q1000850) 129505 biological process (Q2996394)
Marshal of the Sejm (Q100172) 128843 behavior (Q9332)
Marshal of the Sejm (Q100172) 127007 activity (Q1914636)
China Railways HXN3 (Q1005425) 120327 mode of transport (Q334166)
China Railways HXN3 (Q1005425) 120298 finished good (Q3245975)
China Railways HXN3 (Q1005425) 120276 vehicle (Q42889)
Marshal of the Sejm (Q100172) 106677 human behaviour (Q3769299)
Marshal of the Sejm (Q100172) 102689 occupation (Q12737077)
carbon-11 (Q1014081) 73535 matter (Q35758)
Commodore Amiga (Q100047) 67926 artificial physical object (Q8205328)
economy of Djibouti (Q1000084) 57328 manifestation (Q286583)
Commodore Amiga (Q100047) 47839 tool (Q39546)
ghat (Q1010155) 46454 location (Q17334923)
ghat (Q1010155) 46454 position (Q23008351)
ghat (Q1010155) 46255 geographic location (Q2221906)
ghat (Q1010155) 44255 geographical object (Q618123)
carbon-11 (Q1014081) 42923 isotope (Q25276)
carbon-11 (Q1014081) 42602 chemical substance (Q79529)
GFA League First Division (Q1007931) 41325 phenomenon (Q16722960)
GFA League First Division (Q1007931) 41293 phenomenon (Q483247)
GFA League First Division (Q1007931) 41172 social phenomenon (Q602884)
GFA League First Division (Q1007931) 40724 event (Q1656682)
Commodore Amiga (Q100047) 39677 device (Q1183543)
Saint-Pourçain AOC (Q10214) 38838 base material (Q214609)
Bürgl Hut (Q1021576) 38451 construction (Q811430)
Bürgl Hut (Q1021576) 37666 architectural structure (Q811979)
GFA League First Division (Q1007931) 36358 competition (Q476300)
carbon-11 (Q1014081) 34591 simple substance (Q2512777)
Buddhism in Germany (Q1000850) 33530 class (Q5127848)
Buddhism in Germany (Q1000850) 33238 class (Q16889133)
bucolic poetry (Q1003292) 33036 identifier (Q853614)
carbon-11 (Q1014081) 32984 chemical element (Q11344)
economy of Djibouti (Q1000084) 32829 system (Q58778)
bucolic poetry (Q1003292) 32121 information (Q11028)
Saint-Pourçain AOC (Q10214) 29456 part (Q15989253)

Primary Sources Tool broken

I want to contribute with the Primary Sources Tool but it's not working properly at the time. F.e. it provides the same birthdate as it is set already in Jacques Spiesser (Q1678111), but with different references. I know that the tool is not maintained anymore but can anybody fix that? Queryzo (talk) 11:51, 28 March 2016 (UTC)

If you accept the references without accepting the claim, the new references will be added to the old claim and the PST claim will go away. --Izno (talk) 12:03, 28 March 2016 (UTC)
yes, but it didn't even work, when there are no references: Marceline Day (Q782604). Queryzo (talk) 15:20, 1 April 2016 (UTC)

Koppen climate classification

Polar and alpine climates

I'm working on implementing the infrastructure for Köppen climate classification (P2564) and I have something of a problem on my hands, and I'm not sure how best to deal with it (though I have a stopgap now--someone can change how I've dealt with it).

The problem is the E-group of classifications. From what I can observe based on my reading of the en.Wikipedia pages, the entire group is polar and alpine climate (Q23662318). I created it to faithfully reproduce the classification system. This item has no links as a result.

I also have on hand polar climate (Q193372) and alpine climate (Q859654), both of which have many links. A reading of the en.WP article for "polar climate" implies it is the same as the whole group. And similarly for "alpine climate".

Should we set the classification up such that it uses the joint class at polar and alpine climate (Q23662318) or should we pick one or the other of polar climate (Q193372) and alpine climate (Q859654) to set as our representation of the classification? My concern with the latter option is that ice cap climate (Q5985406) and Q23662321 need to take the correct subclass relationship, which I believe to be the "joint" item at polar and alpine climate (Q23662318). --Izno (talk) 16:46, 30 March 2016 (UTC)

Tundra climate

I have created a Q23662321, with no links. However, the en.WP article at tundra (Q43262) seems to represent some of the climate, though is more about the biome as a whole (and it is represented as such in Wikidata, as a "biome"). I have currently listed the new climate item as the correct item to use. Do you agree with this decision? --Izno (talk) 16:46, 30 March 2016 (UTC)

you need to specify a tundra climate is a class within the Koppen climate system, so a new class item is needed. Michiel1972 (talk) 10:19, 31 March 2016 (UTC)

Use of Koppen climate classification as instances

Another: I have used the system item, given its English name, to indicate that each of the classifications is an instance of (P31) the classification as a whole. Is this okay, or should the classification item be a separate "class/type" item? --Izno (talk) 16:54, 30 March 2016 (UTC)


@Izno: I urged to create the climate zone items before, but I wasn't heared: here.--Kopiersperre (talk) 10:35, 31 March 2016 (UTC)

@Kopiersperre: Right, and that's what I'm filling out on the property talk page now and how I bumped into this problem. Do you have a comment specific to my question? --Izno (talk) 11:07, 31 March 2016 (UTC)
  • I'm not really convinced about the approach to re-purpose existing items about articles that might not even mention the classification for a use within a scheme. I think that was a problem already mentioned in the creation discussion. You might find better input at the relevant WikiProject.
    --- Jura 07:43, 2 April 2016 (UTC)

Rhodopsin (Q423107) and RHO (Q14859555)

seem to be related. Is it that all sitelinks for Rhodopsin show up at RHO or are there more sublte twists? -- Gymel (talk) 12:08, 31 March 2016 (UTC)

From my understanding the first one is about the protein and the second is the gene which encodes the protein. I think some WPs treat both in one article and depending on the focus the article is linked to the protein item or to the gene item. I can't read all languages to define which is the more appropriate item. Snipre (talk) 14:21, 31 March 2016 (UTC)

Same with Q39798 and INS (Q21163221). Seems as @ProteinBoxBot: is systematically(?) moving sitelinks from proteins to the responsible genes. As a layman I find this a bit surprising. -- Gymel (talk) 21:54, 1 April 2016 (UTC)

Coordinates and OSM

I've made a list of mountains in the Balearic Islands: ca:Usuari:Paucabot/Muntanyes2. The problem is that the coordinates are not good enough as they can be some meters away from the summit: Zoom in here to see it. Is there any way to (easily) import and substitute OpenStreetMap coordinates? Or, much better, is it planned something to share OSM data? Paucabot (talk) 18:04, 29 March 2016 (UTC)

This is tricky. OSM is licensed under the Open Database License (ODbL) (thematically similar to CC-BY-SA), while Wikidata is licensed under CC0. Importing substantial portions of data from OSM into Wikidata is not allowed unless Wikidata is able to license these imported data under ODbL. —seav (talk) 19:39, 31 March 2016 (UTC)
Thanks, Seav (talkcontribslogs). By the way, I found this: Wikidata:OpenStreetMap. Maybe we should explain this there? Paucabot (talk) 08:57, 1 April 2016 (UTC)
@Seav, Paucabot: This is just about the coordinates of 89 summits. Does that small amount qualify as a "substantial portion of data from OSM"?--Pere prlpz (talk) 23:04, 2 April 2016 (UTC)
Please see for what the OSM community considers "insubstantial". —seav (talk) 23:19, 2 April 2016 (UTC)

Call for ideas : wikipédias and wikidata references

Hi people, wikidata datas are (chaotically) more and more used by wikipedias, and some - long standing - problems could become more important in the near future.

One of them I think about for a long time : how to use wikidata references in wikipedias articles. I think we have a set of partial solutions that does not meet some of the higher quality standards people can manage to use into some carefully sourced wikipedia articles : how to mix references that comes from Wikidata in a consistent way with sources coming just from the article. I want to know if other people encountered these kind of problem already, how, and get some ideas from community (and @Lea Lacroix (WMDE):?)

One of the main problem is reference deduplication that we have no real automated way to know if a reference in some part of the article is actually used in another part of the article, at several levels.

Problems sketch

For example if a URL/document is used in several statements on the same items. Say we want to show the references in an infobox. We could have stuffs like

  place of birth: New-York[example1 1]
  date of birth: date[example1 2]
  place of death: New-York[example1 3]
  date of death: date[example1 4]
"recording reference" possible solution 
This first problem could be dealt with in a lua generated infobox if the infobox is generated in a single lua template call because there is a way to record every url used in the references and generate an ID for them and generate the following code into
place of birth: New-York[example1bis 1]
date of birth: date[example1bis 1]
place of death: New-York[example1bis 1]
date of death: date[example1bis 1]
  1. 1.0 1.1 1.2 1.3
  2. But this is not a perfect technical world, and this can not be extended in other parts of the article if the same url is used. We actually cannot record the references in the whole article because of some technical requirement for the Visual Editor (he must be able to pre-render templates without side effects so that its rendering cannot be invalidated by changes in other parts of the page).
    the "computed/conventional id" solution 
    This solution is used for things like so called "harvard references". In harvard references, every book has an id conventionally made from the authors first names and the name of the year. This id is not used in the "name" attribute of the ref tags, but to generate an anchor to the book in the bibliography.
    This allows to decouple the citation to the cited book, and to cite specific pages of the book without duplicating the informations about the book itself. It would be hazardous to try to use such an id in the "name" of references because it coud easily lead to ugly error message if the content of the ref tag is different :
      place of birth: New-York[example1ter 1]
      date of birth: date[example1ter 1]
      place of death: New-York[example1ter 1]
      date of death: date[example1ter 1]
    1. 1.0 1.1 1.2 1.3 Cite error: Invalid <ref> tag; name "url1" defined multiple times with different content Cite error: Invalid <ref> tag; name "url1" defined multiple times with different content Cite error: Invalid <ref> tag; name "url1" defined multiple times with different content
    But, although we have many identifiers for books or articles that goes from isbn, Qids, doi, etc. it seems it's hard to use them at community scale in a consistent way, that the harvard ids may generate collision which may make necessary to overload the automatically generated one ...

    So it seems to me we don't really have a perfect solution with current software and that we may benefit changes in the mediawiki references management system ... any thought ? What could we do to make easier to use Wikidata references in Wikipedias ?

    PS: That said I think I may have The solution : wrap every article into a lua call so that we can share the references used into the article and treat that with a custom parser and the "expandtemplate" lua function. This would look like this :

    {{referenced article|
    a '''referenced article''' is an article in which references and bibliographies are deduplicated<ref name="fisher"> {{U'|TomTom}}, 1st of April</ref>.
    {{infobox examples}}
    {{deduplicated bibliography}}

     – The preceding unsigned comment was added by TomT0m (talk • contribs).

    Have you seen w:hy:Սթիվեն Հոքինգ? (especially ref N4 and N6)--ԱշոտՏՆՂ (talk) 18:10, 1 April 2016 (UTC)
    I´m working on references too, but I have not thought about integrating references from the module and from the Wikipedia text. A possible solution is maybe to store some reference names in Wikidata. This way it is possible to combine the references at least by hand, because this way I knew the references names used in the module. The references names I´m creating are build from the item q-number plus a counter. --Molarus 20:24, 1 April 2016 (UTC)
    Why can't we just use whole reference for reference name? Like this: <ref name=""></ref>. It works in hywiki (see W:hy:Մոդուլ:Sources especially line 543)--ԱշոտՏՆՂ (talk) 22:57, 1 April 2016 (UTC)
    On svwiki I use "hash". See line 315 in sv:Module:Wikidata2. If different parts of the page uses different references with different hashes, the parser of the page combines them in one reference in the page. I do not know how our developers have managed to do so. They have probably spent many years in Hogwarts to accomplish thIS!
    table.insert(reference, mw.getCurrentFrame():extensionTag( 'ref', s, {name = ref.hash} ) )
    -- Innocent bystander (talk) 07:35, 2 April 2016 (UTC)
    @Innocent bystander: I assume you mean ref.hashreference.hash ? What would be interresting is the code that compute the "reference" object. author  TomT0m / talk page 09:21, 2 April 2016 (UTC)
    @ԱշոտՏՆՂ: what you show up is just using the uri, not the whole reference. This will break if the same URI is used in two references but some datas are added in addition to the reference, because the content of the ref tag could be different (say, same uri, but a different consultation date in the two cases) the same way in my "exampleter". author  TomT0m / talk page 09:21, 2 April 2016 (UTC)
    Yes Tom, 'ref' here is a loop over every 'reference' in the 'references'. I am a little surprised that it works, but it actually does. The only time it breaks down, is when the code that sets the reference together looks differently in different parts of the page. If there is two infoboxes which uses separate algoritms to set the references together and they both uses the same "hash", it breaks down. But that has only happened to me when I have made a copy of the algoritms to tests some improvements. -- Innocent bystander (talk) 09:39, 2 April 2016 (UTC)

    Don't forget that references can be to paper sources. See, for example, Galileo Galilei (Q307), the occupation property (astronomer), has a reference to the book The Hunt for Planet X (Q20888754). Also, because different articles have different citation styles, it will be virtually impossible to find one way to code an infobox that would allow it to be used in a wide variety of articles. I'm working on w:Template:Infobox zodiac/sandbox, but the only reason that one is manageable is because it will only be transcluded in 12 articles. Jc3s5h (talk) 12:15, 2 April 2016 (UTC)

    @Jc3s5h: Yes, I know. Actually when the work has an item, the Qid could be very handy has an anchor to identify it in the bibliography on articles that uses harvard style references - this is the default id used for fr:Modèle:Bibliographie. As this template use itself the Qid this totally make sense. This is usable in fr:Modèle:Référence_Harvard_sans_parenthèses or fr:Modèle:Référence Wikidata. But there is a lot of statement thathave references without any dedicated item. author  TomT0m / talk page 13:46, 3 April 2016 (UTC)

    Line and Ligne

    In English wiki there are two articles: and they are described by items Q649848 and Q1630774. They are for British Line Unit and French Line Unit. In Russian wiki there is only one article about all kind of Line units:Линия_(единица_длины) but now it is associated only with Q649848, but should be associated with both of the items. And both English wiki pages Line_(unit) and Ligne should have link to this Russian article.

    But I am new to wikidata, and I do not understand at all how to do this. (Not only technically, but I think I miss some part of general idea too) Can you please explain me this step by step, so I can will be able to do it myself next time. Thank you! --Nataraj (talk) 08:27, 3 April 2016 (UTC)

    This is a general problem with the wikidata model, which assumes a 1:1 relationship between concepts and Wikipedia articles. Each item in Wikidata can be associated with a maximum of one page on each Wikipedia, and each Wikipedia article can be associated with a maximum of one Wikidata item. I believe interwiki links can be worked around by adding them the old fashioned way (subject to local wiki policy), but I don't know how (or even if it is possible) to do everything fully in this situation. Thryduulf (talk: local | en.wp | en.wikt) 10:38, 3 April 2016 (UTC)

    How many entities should be in Wikidata dumps?

    According to main page, there are 17.209.354 items in Wikidata. I downloaded a Wikidata entities json dump a few days ago and I expected to find the same number of items there. However, I counted about 20,568,199 lines, 20,565,957 of which are items (e.g. entities with an id starting with "Q").

    Although my count could be wrong I think it isn't. Therefore, is there any reason to be more items in the dump than in statistics?

    If it matters, the dump I used was latest-all.json.bz2 from . --Pere prlpz (talk) 23:35, 2 April 2016 (UTC)

    Redirects? - Brya (talk) 06:01, 3 April 2016 (UTC)
    Yes in the dump there are also redirect --ValterVB (talk) 07:16, 3 April 2016 (UTC)
    That makes sense, because redirects and merged items also account for the difference between the official number of items (17M) and the much higher Q numbers of recent items (23M). It also seems related to de way items are ordered in dump, in two series of increasing order - although entities order is not supposed to be meaningful. I'll try to check.--Pere prlpz (talk) 09:49, 3 April 2016 (UTC)
    I haven't been able to find any duplicate nor any redirect marked somehow as a redirect, but it's hard to search in such a big file (69 Gb).
    This problem occurred a couple of years ago: . I left a comment in phabricator just in case it is the same problem again.--Pere prlpz (talk) 21:57, 4 April 2016 (UTC)

    Removing Wikidata bots from watchlists on other projects?

    I like having Wikidata on my Wikipedia watchlists, but the Wikidata bots just seem to clutter everything up. Is there a way to just remove the Wikidata bots while leaving the Wikipedia bots and the rest of Wikidata on my Wikipedia watchlists? Cheers, Irn (talk) 17:18, 3 April 2016 (UTC)

    @Lydia Pintscher (WMDE): Probably something for WD:Contact the development team, but I'll ping you here just to see if you have an opinion. --Izno (talk) 20:41, 3 April 2016 (UTC)
    @Irn: Without checking, it's probably possible to use CSS to hide the lines where it's a bot from Wikidata. --Izno (talk) 20:42, 3 April 2016 (UTC)
    I have sv:Användare:Larske/Testsida6 in my Watchlist. There I now see 7 different edits by User:Sarah Layton in 5 different lines. The edits affects 3 different items, Alexandrine of Mecklenburg-Schwerin (Q57264), Christian X of Denmark (Q156617) and Louise of Sweden (Q232402). Why do I have 5 and not only 3, one for the latest edit in the linked items? -- Innocent bystander (talk) 10:29, 4 April 2016 (UTC)
    I think we need phabricator:T51315 to solve this. --Lydia Pintscher (WMDE) (talk) 10:10, 5 April 2016 (UTC)

    @Izno: and @Lydia Pintscher (WMDE): Thanks for your responses, but I'm not really sure what to make of them. How do I go about using CSS to hide the lines from Wikidata bots? And I checked out the phabricator link, but I'm not really sure what to do there. Cheers, Irn (talk) 13:06, 5 April 2016 (UTC)

    Links to redirects

    To my knowledge it was once discussed here that links to redirects shall not be changed, and Help:Redirects#Links to redirects also states that. However apparently at least one bot operator seems to have more knowledge, and his bot changes redirects, causing exactly what was the help page states - lot of other items modified without any real need after a wrong page merge, and thus lot of additional cleanup work - see here @Ivan A. Krestinin:. Was there ever a discussion which changed that policy how to handle redirects? And if yes, why nobody bothered to update the help page? Ahoerstemeier (talk) 10:00, 4 April 2016 (UTC)

    Since I didn't know of that page (which is not a policy, but a policy proposal), I have been fixing redirects when I came across them. This has the background that labels of redirects are not rendered, which makes redirect less than useful on item and property pages. --Srittau (talk) 11:06, 4 April 2016 (UTC)
    Wouldn't that be better to fix the software, instead of changing the data to workaround the problem? Don't know if there is already a bug report for this, as I am not really familiar with the bug reporting here. Ahoerstemeier (talk) 09:07, 5 April 2016 (UTC)

    Wikidata weekly summary #203

    Adding Wikidata for Brandemix

    Hello WIkiData Admins,

    I am new to wikidata and wanted to include the basic details of the business Brandemix Deepaksachdeva123 (talk) 10:52, 5 April 2016 (UTC)

    See Srittau's note above. Mahir256 (talk) 16:12, 5 April 2016 (UTC)

    Upcoming coordinate import

    So a few days ago, I imported some coordinates from German Wikipedia, but I was a bit ... indiscriminate about it. So now I am preparing a larger import, taken from all Wikipedia. Yup, that's right. Currently, there are >120K coordinates; could be ~150K by the time it's finished. But this time, each and every item I add to has (a) no coordinates, and (b) instance of (P31):subclass of (P279):geographic location (Q2221906). So, it should be significantly higher quality. I will start adding them today or tomorrow.

    I am posting this not only to alert you to watch out for this (edits will be done by User:Reinheitsgebot), but also to ask a question. Apparently, on some wikis, bots have been adding pages; in this case, I am mostly looking at hi.wikipedia. There are lots of village-level entries, which is good in itself. However, many of these have the same coordinates; the top shared coordinate is used for 1285 articles. Should I include these? They may be somewhat off, but still OK-ish, AFAICT. Or should I try to filter out the commonly used ones? Which "usage level"? 5x total? or 5x by the same Wikipedia? Or should I just skip Hindi Wikipedia (Q722040) and Gujarati Wikipedia (Q3180306) altogether? --Magnus Manske (talk) 14:33, 31 March 2016 (UTC)

    No, I think if we are sure the coordinates are not correct we should not add them.--Ymblanter (talk) 14:44, 31 March 2016 (UTC)
    Agreed. It's much easier to find items without coordinates than to find items with wrong coordinates. --Srittau (talk) 14:50, 31 March 2016 (UTC)
    BTW, could you not import coordinates from ruwiki. Apparently they have some issue with the DMS/decimal conversion.
    --- Jura 14:53, 31 March 2016 (UTC)
    I know, I should have answered at botreq page, but as you're posting here... (with little background, so this sounds logical) Isn't it possible to add correct precision? See this import and dewiki article. Precision as you can see is quite different. Why it's really important to have the correct precision? Well, because it is (or at least should be) correct :) Geo-people most probably can give you more serious reasons for having correct precision. And previously bots have been adding coordinates with correct (at least kind of correct) precision (that's why, I'm asking about this precision thing). Or problem is somewhere else? About Jura1's mentioned ruwiki problem see also this. --Edgars2007 (talk) 15:01, 31 March 2016 (UTC)
    So I am using the geo_tags database table on the Wikipedias, which does not store precision. I will be using QuickStatements to add them, which uses the default precision of Wikidata (that is, it does not specify a precision).
    Also, suppressing hi, gu, and ru Wikipedias as sources, statement count goes down by 50%. Pity. --Magnus Manske (talk) 15:10, 31 March 2016 (UTC)
    Maybe the ru thing isn't relevant (see prev. discussion) @Ivan A. Krestinin: could more be imported?
    --- Jura 15:37, 31 March 2016 (UTC)
    ruwiki error rate is not looked as significantly different from another wikies. So coords from ruwiki can be imported. Some other bots loosed precision during import. Please be more accurate. Also please do not import coord templates without "display=title" flag. — Ivan A. Krestinin (talk) 21:24, 31 March 2016 (UTC)
    Related possible import: Category:Articles with OS grid coordinates (Q6377660). Some of these articles don't have any coordinates or no statements at all (see Wikidata:Database reports/items without claims categories/cywiki).
    --- Jura 15:40, 31 March 2016 (UTC)
    I think it would be mostly correct to add coordinates for instances of event (Q10290214) as well, there are quite a few battles, train accidents, mass killings... that need coordinates. --Zolo (talk) 14:44, 1 April 2016 (UTC)

    Update: I have begun an import of ~30K coordinates. These have been filtered "conservatively". If there are significant issues, please block User:Reinheitsgebot. The coordinate gathering script has not yet completed, so there is a huge batch of coordinates from Swedish Wikipedia to be imported; I hope that one doesn't have the "Russian issues". Plus, all languages that come alphabetically after "sv" are still to be done. --Magnus Manske (talk) 15:47, 1 April 2016 (UTC)

    @Magnus Manske: I am not aware of any "Russian issues" on svwiki, but many coordinates comes from GeoNames, and they are often rounded to minutes, and are therefor sometimes wrong. Lsjbot is trying to compensate for that, see my note below. -- Innocent bystander (talk) 05:58, 4 April 2016 (UTC)

    @Magnus Manske: found one issue, I will call it "Chuvash issue". You can search here "0°0'0.000"N, 0°0'0.000"E". Probably those are not the only involved items. And/or your script could take into account cordinates with are at the center. There are not so many objects :) --Edgars2007 (talk) 06:51, 5 April 2016 (UTC)

    Update 2: Now importing half a million (!) coordinates from Swedish Wikipedia. Looks reasonable so far, ping me (Twitter, email, here) if I need to cancel it. --Magnus Manske (talk) 15:02, 5 April 2016 (UTC)

    So most coordinates have been imported now. There are still ~76K left, mostly from these wikis:

    • ruwiki: 30,674 pages with coordinates
    • guwiki: 18,153 pages with coordinates
    • ltwiki: 13,624 pages with coordinates
    • hiwiki: 8,536 pages with coordinates
    • glwiki: 1,991 pages with coordinates

    But above I hear the ru/guwiki ones might be substandard. What to do? --Magnus Manske (talk) 14:05, 7 April 2016 (UTC)


    FYI: There are a lot of articles with Coordinates in Lsj's latest project. It has used the coordinates from GeoNames. One problem we have detected with GeoNames is that they have introduced many rounding errors. The bot has tried to adjust the coordinates when such mistakes has been detected by the help of an algoritm in the bot code. The pages with such adjustments can be found in sv:Category:Artiklar med robotjusterad position. -- Innocent bystander (talk) 08:06, 1 April 2016 (UTC)

    Guide for importing data into Wikidata

    Hi all

    Myself, User:Jens_Ohlig_(WMDE) and some others are working on a how to guide for people wanting to import data into Wikidata, the audience is existing Wikidata editors with a non technical background (i.e not programmers) and it would be a kind of partner page for the Wikidata:Data_donation page I rewrote. At the moment it is just a collection of existing tools and ideas, any help would be appreciated.

    Many thanks

    John Cummings (talk) 09:28, 7 April 2016 (UTC)

    Nontrivial request: Recovering unclear pdf document

    (also cross-posted to Wikipedia talk:Graphics Lab)

    Hello, I didn't even know Wikidata existed until 5 minutes ago, and I don't know if this is the correct forum etc. so apologies.

    I'm working on a complete rewrite of w:Bengal famine of 1943 in my personal sandbox. The central document for this topic (though it's biased) is the Woodhead Commission Famine report. It's a available in pdf format here. I can save that into .txt format (hurray!), and have written a little Python program that finds keywords from a large number of similar text files and stores quotes into separate files.. however, the scan quality or the Famine Commission report is so poor that extended stretches are simply gobbledygook.

    This is a nontrivial request: Is there a PhotoShop guru (or similar) who could sharpen the MANY pages into significantly better & more scannable pdfs? Not all pages could be fixed, because some show the curvature of the book pages etc., but I think many many could be improved.

    I have downloaded an evaluation copy of PhotoShop etc and tried to use Sharpen and Levels or Layers whatever to make each page more machine readable, but I don't know how to do it for an entire (large!) report, and I don't know how to scan them or save them to text instead of image (printing every page and scanning each manually is obviously much too much work). I also have a family life and work etc. and learning how to do all these things would just take too much time.

    Does anyone have suggestions?

    In theory, this service might be valuable for other old documents scanned to pdf, but i dunno how much demand there would be for such a service.

    ThanksLingzhi (talk) 04:51, 8 April 2016 (UTC)

    @Lingzhi: This is not relevant for Wikidata. You may get more help at commons:Commons:Village pump or at wikisource:Wikisource:Scriptorium/Help. Good luck! Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 11:41, 8 April 2016 (UTC)
    thanks!Lingzhi (talk) 13:35, 8 April 2016 (UTC)

    P2037 Github username

    hi, should GitHub username (P2037) be added as a property or as an identifier? if it's an identifier, my follow up question is how do I search for identifiers i the query tool? Example query: slash h8vragg Ambrosiani (talk) 17:34, 8 April 2016 (UTC)

    @Ambrosiani: hmm, I assume you meant qualifier instead of identifier, right? If so, then it should be used as a property. Some piece about querying qualifiers in SPARQL: Wikidata:SPARQL query service/queries#Working with qualifiers. --Edgars2007 (talk) 17:55, 8 April 2016 (UTC)
    Sorry, I mean "Statement" or "Identifier" (the two different places where I can add properties). Ambrosiani (talk) 18:33, 8 April 2016 (UTC)
    Oh, that sounds more logical :) AFAIK, you don't have to worry about that. At least, putting everything in statement part works completely fine. System will put in right place. Quering at SPARQL is the same. For the record, Github username goes to "Identifier" section, as it has "External identifier" datatype (you can see it at property page). --Edgars2007 (talk) 18:47, 8 April 2016 (UTC)

    Splitting {{Property proposal}} from {{Property documentation}}?

    Currently, {{Property documentation}} is used on both property discussion pages and on property proposal pages. This made sense, but I think the time has come to split this template in two, since their use has diverged quite a bit since their creation. The overlap of fields to copy when creating a new property is very low now, usually only one or two fields, since more and more of the fields have been moved into property statements (where they belong). On the other hand there are several code paths in the template now that check, whether it's been used on a property talk page or elsewhere, and does different things based on that.

    Therefore, I suggest to create a separate template for property proposals that can be separately maintained. Advantages from my point of view:

    • The use of {{Property documentation}} and all its fields, which are partly targeted at property discussion pages, confused me when I proposed my first property. A separate template with a separate documentation can fix that.
    • Clearing up the confusion that exists, because fields are used differently, depending on the use of the template.
    • Removal of unnecessary (and therefore also confusing) fields from either template.
    • Better error detection, for example missing or spurious fields on each template.
    • Easier code paths/easier to maintain templates.
    • Easier way to evolve and improve the templates in the future.


    • You have to replace the word "proposal" with "documentation" when creating a property. Considering the cleanup required anyway, I consider this a minor point.
    • The translations of both properties have to be maintained separately. I think the overlap of both templates will become very small over time, though.

    Please note that currently there already is a template {{Property proposal}}, which I consider ill named anyway and would need to be renamed. Since its use is not very widespread, I don't think this would be a problem though (about 60 uses).

    Suggested course of action:

    1. Rename {{Property proposal}} to {{Property proposal link}}.   Done --Srittau (talk) 09:34, 6 April 2016 (UTC)
    2. Duplicate the existing {{Property documentation}} as {{Property proposal}}.   Done --Srittau (talk) 20:27, 6 April 2016 (UTC)
    3. Replace {{Property documentation}} with {{Property proposal}} on the proposal pages and the archive.
      • Partially done. I replaced the uses on the current proposal pages and in the current archive as well as the templates for new proposals. Someone needs to allow the latter to be translated, though. Also missing are the other 46 archive pages. If a bot operator could help here, it would be much appreciated. (Replace "{{Property documentation" with "{{Property proposal".) --Srittau (talk) 21:01, 6 April 2016 (UTC)
      • Now   Done, thanks to User:Pasleim. --Srittau (talk) 20:05, 7 April 2016 (UTC)
    4. Clean up the LUA code of both templates.

    When the templates have been refactored and un-entwined like this, we can start to discuss cleanups and improvements to those templates. --Srittau (talk) 13:35, 3 April 2016 (UTC)

    Do it. --Izno (talk) 20:40, 3 April 2016 (UTC)
    A near-identical proposal is already being discussed, at Wikidata talk:Property creators#Property Template change?. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits
    Thanks for mentioning that, I should have done so in the proposal above. I repeated it here again for a larger audience and with a concrete plan of action. --Srittau (talk) 18:08, 4 April 2016 (UTC)
    A pointer is sufficient; please don't split discussions. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 20:51, 5 April 2016 (UTC)
      Support I agree this is the right thing to do now. ArthurPSmith (talk) 17:32, 4 April 2016 (UTC)

    Good to see you working on documentation. You might want to use the (quite hidden) footer feature. Take for example OMIM ID (P492). Would be nice to have the docs at the bottom too so you don't have to click on the talk page. You can set it at Property talk:P492/footer. Probably it's just Module:Property documentation needing a small modification so it loads the information. Multichill (talk) 19:35, 7 April 2016 (UTC)

      Done, further discussion can take place on Template talk:Property proposal and Template talk:Property documentation. --Srittau (talk) 16:27, 14 April 2016 (UTC)

    This section was archived on a request by: Srittau (talk) 16:27, 14 April 2016 (UTC)

    Any guideline about taxon names?

    Lots of taxons, such as family with only one genus, are merged into one article on most of Wikipedia projects. Should we keep them as one item or separate ones? If the answer is latter, which one should we used for links?

    An example is Q3866060 and Q229084. On most of Wikipedias they are described in the same article (with a few exceptions, like But apparently the links are now in different items despite they're describing the same thing. This kind of situation is super common. I assume there must be some guideline about this but I cannot find it. --Fireattack (talk) 08:54, 8 April 2016 (UTC)

    From a taxonomic point of view, they are not necessarily the same so they should be separate. Thanks, GerardM (talk) 10:32, 8 April 2016 (UTC)
    In taxonomy there are often multiple points of view: when one taxonomist considers a family to have only one genus, there often is another taxonomist who considers that family to have more than one genus. It is important to document any taxonomic position with good taxonomic references. Given that there are multiple projects that have pages on both Moschus and MOSCHIDAE, it is a good example of a non-issue. But, anyway, pretty much by definition, a page on a family is different from a page on a genus. - Brya (talk) 10:53, 8 April 2016 (UTC)
    In taxonomy there is a clear taxonomy of the publications. It includes dates and sources. The notion that we have multiple projects that are not aware of this is a non-issue. It is important that Wikidata is correct in this. Thanks, GerardM (talk) 13:31, 8 April 2016 (UTC)
    Maybe it would be of your interest: Wikidata:WikiProject Taxonomy. Paucabot (talk) 11:38, 8 April 2016 (UTC)
    Specifically Wikidata:WikiProject Taxonomy/Tutorial#Taxonomy changes. Paucabot (talk) 11:41, 8 April 2016 (UTC)
    @GerardM, Brya, Paucabot: Guys, thanks for your answers. I totally agree that they should be separate. However I think you didn't address my real concern: which one should be used to storage all the SITELINKS, for projects (mainly Wikipedia) that merged these two items. Status quo: Some of them are linked to one while some others are linked to another. Obviously we could do better. Maybe always link to the one with higher rank? (this is just an example.) I am here to seek for guidelines to deal with this kind of situation. I also read the pages Paucabot provided but I didn't find info about this very topic. --Fireattack (talk) 19:08, 8 April 2016 (UTC)
    Why the „higher rank”, Fireattack, and not the lowest rank? You'll need a species name to establish a genus name. --Succu (talk) 20:47, 8 April 2016 (UTC)
    No why, I said it's just an example...--Fireattack (talk) 06:14, 9 April 2016 (UTC)
    It is often difficult to determine what the real concern of a poster is. As to the placement of sitelinks, the practical course is to accept these at face value: if the title of the page is MOSCHIDAE, then put the page in the item MOSCHIDAE. Hopefully, the opening sentence and the taxobox will be in agreement with the title. If one was going to determine, on a case by case basis, what the gist of the page is, one would hit two obstacles: 1) others might not agree on what the gist is, and 2) pages are dynamic, and would have to be re-evaluated after each substantial change. - Brya (talk) 03:57, 9 April 2016 (UTC)
    Well, following the title of the page is a solution which I guess is okish from the point of view of Wikidata (which is still not ultimate solution: where should articles like w:en:Musk deer go?). But in Wikipedia, I don't think it's a good practice to not link (essentially) same inter-language pages together.--Fireattack (talk) 06:18, 9 April 2016 (UTC)
    Well, in w:en:Musk deer both the opening sentence (sort of) and the taxobox claim it is about the genus, which is supported by the list of species. So, the obvious placement is in the item Moschus. If the opening sentence is taken at its most literal, it belongs in neither, but in a "Musk deer, instance of common name", which is impractical.
            And connecting everything that would be desirable from the perspective of each individual Wikipedia will have to wait for an appropriate software solution which allows customizing. - Brya (talk) 06:55, 9 April 2016 (UTC)

    Who added a betawikiversity link?

    Which is affected by phab:T54971? --Liuxinyu970226 (talk) 03:34, 9 April 2016 (UTC)

    The item is Q23685749. The link isn't displayed, although it's there in the JSON. - Nikki (talk) 10:30, 9 April 2016 (UTC)

    WikiCite applications closing soon

    A reminder that applications to attend WikiCite 2016 – an event that should be of interest to Wikidatans active on source-related work – close this Monday April 11. We have a limited number of travel grants to support qualified participants. If you wish to join us in Berlin to participate either in the data modeling or engineering effort, please consider submitting an application --DarTar (talk) 16:09, 9 April 2016 (UTC)

    Deleting descriptions

    Can we have some tool/API call or whatever to remove all descriptions for an item? Basically, I'm talking about former disambigs, that we're transformed into articles at (all, of course) wikis, like Josh Cobb (Q6288763). It is painful to remove them manually (even in list of headers tool). I know, that I'm not the only one, who would like to have such tool. --Edgars2007 (talk) 17:43, 3 April 2016 (UTC)

    MediaWiki:Gadget-dataDrainer.js is a great tool for that. --Stryn (talk) 18:31, 3 April 2016 (UTC)
    Hah, I'm not in needed user groups :D Then maybe somebody could delete descriptions for linked item? --Edgars2007 (talk) 18:45, 3 April 2016 (UTC)
    Ah, I see. Earlier it worked for autopatrolled users as well, but as this user group doesn't exist anymore, meh... then you should request for adminship maybe ;) --Stryn (talk) 18:51, 3 April 2016 (UTC)
    My rationale would be lame :) But if seriously, then probably not now. --Edgars2007 (talk) 19:57, 3 April 2016 (UTC)
    You can also use the gadget if you're a rollbacker tho. Sjoerd de Bruin (talk) 11:37, 10 April 2016 (UTC)

    Anyway, if somebody wants to get some fun, here is the list of items (308 currently), that are not instance of (P31)=Wikimedia disambiguation page (Q4167410), but have English description "Wikimedia disambiguation page". Of course, double check is needed. --Edgars2007 (talk) 09:17, 5 April 2016 (UTC)

    'Changed data type from string to external-id'

    Hi all, I'm running a bot to move some control authority identifiers into wikidata (namely this one: P1607). However, as pywikibot does not support 'external-id' I'm not able to find any straightforward way to do id. May I ask to revert the data type change in P1607 for some hours so that I can update the items? They're no more than three hundred items. Any other solution or suggestion is also appreciated. Best regards --Discasto (talk) 23:32, 8 April 2016 (UTC)

    Forget it, I just tweaked the pywikibot core code. Thanks --Discasto (talk) 09:12, 9 April 2016 (UTC)
    Has this been merged into pywikibot core? I guess other people will also run into this issue if it has somehow been forgotten by the pywikibot team! ·addshore· talk to me! 09:45, 9 April 2016 (UTC)
    @Discasto: @Addshore: I wonder what you ran into. As far as I know Pywikibot has full support for this new datatype and I haven't run into any problems myself. Did you run the latest (git) version? What was the problem exactly? How did you fix it? Multichill (talk) 14:24, 9 April 2016 (UTC)
    Hi all, thank you for your questions. Initially, I thought it was just a question of an old pywikibot distribution. So yesterday I removed the pywikibot core code and downloaded the last version (this one), which is said to be 2.0rc3 ( says the version is 2.0b3, I assume both strings, 2.0b3 and 2.0 release candidate 3 are the same). I got the same issues with the following code:
    DIALNET_PROPERTY        = u"P1607"
                item = pb.ItemPage(repo, title=page_id)
                #print type(dialnet_id)
                try :
                except :
                if DIALNET_PROPERTY not in
                    try :
                        dialnet_claim = pb.Claim(repo, DIALNET_PROPERTY, datatype='external-id')
                        pb.output('Adding %s --> %s' % (dialnet_claim.getID(), dialnet_claim.getTarget()))
                    except :
                        print "Couldn't add claim"

    I don't remember the exception text (after removing the try/except statements), but mentioned 'external-id' not being supported.

    So I edited and included one code line and a half. I added line 3802 (within class Property):

                 'external-id': basestring,

    And modified line 4317 (former 4316). From

            elif self.type in ('string', 'url'):


            elif self.type in ('string', 'external-id', 'url'):

    Once done that, the bot was able to accurately insert any dialnet identifier I asked.

    Hope this clarifies the issue and helps the pywikibot developers (unless I was doing anything wrong). Best regards --Discasto (talk) 21:47, 9 April 2016 (UTC)

    I guess you're referring to this line Discasto. The pywikibot stable version doesn't include the latest features. I recommend you switch to the git version. You can do this with the command $ git clone --recursive pywikibot-core assuming you already have git on your system. Multichill (talk) 20:48, 10 April 2016 (UTC)
    Thanks for the info. I don't have git on my system, so that I will have to live with my "stable" version. Odd to notice that such a stable version has got such significant flaws :-( --Discasto (talk) 21:31, 10 April 2016 (UTC)

    Please merge

    Hi, please help me to merge this items: Q8341916 and Q772399 (duplicate). I can't understand how to merge items here.-- 09:35, 10 April 2016 (UTC)

      It's done, thanks. The second one ended in zero (Category:Cannibalised people (Q7723990)). Strakhov (talk) 10:03, 10 April 2016 (UTC)

    [1] and [2] should be the same person. Unfortunately, two different Wikidata items. --Jobu0101 (talk) 15:41, 10 April 2016 (UTC)

    Those firstly needs to be merged at Wikipedia. 1 WD item per 1 Wikipedia article. --Edgars2007 (talk) 18:58, 10 April 2016 (UTC)
    I've turned Q20564129 into a wikimedia duplicated page on wikidata. Someone will need to merge the wikipedia articles on azwiki, which looks fairly straighforward for an Azerbaijani speaker, as they are virtually identical. Silverfish (talk) 00:38, 11 April 2016 (UTC)

    WikiProject Museum

    Hello everybody,

    I want to create a project for museum, with propreties recommended and rules (like the difference between the buildings and the museum). But I do not have enough time and skills to lead this project. If you are interested…   --Tubezlob (🙋) 18:09, 10 April 2016 (UTC)

    I've made that : Wikidata:WikiProject Museums --Tubezlob (🙋) 13:42, 11 April 2016 (UTC)

    seleccion chilena

    alguien modifico los seudonimos de la seleccion chilena y tambien chile esta en el rankinf fifa 3 no 5 alguien que lo arregle gracias  – The preceding unsigned comment was added by Sergio12341234 (talk • contribs) at 14:42, 11 April 2016‎ (UTC).

    Google translates this as "someone modify the pseudonyms of the Chilean national team and also Chile is in the fifa rankinf 3 5 someone to fix it thanks". Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 15:12, 11 April 2016 (UTC)

    Panama papers

    What property to use to indicate that someone is mentioned in the Panama papers ? Thanks, GerardM (talk) 06:53, 5 April 2016 (UTC)

    I can't find anything really suitable. We have participant of (P1344), part of (P361), and present in work (P1441) which are all close but probably wrong. There was a proposal for a "mentioned in" property that was not done, but that would have had the domain of fictional entities only.
    In the short term you could probably do something with significant event (P793) but unless I'm missing something a new property to denote this sort of thing (mentioned in reports, leaks, etc) will be the best solution going forwards. Thryduulf (talk: local | en.wp | en.wikt) 16:02, 6 April 2016 (UTC)
    Ok, This is what I have done for now; it is easy to update as it now exists as data. Thanks, GerardM (talk) 10:31, 8 April 2016 (UTC)
    From the News I've heard that Anders Wall (Q6230054) also is mentioned in these papers. But I am not sure it qualifies as a "significant event". -- Innocent bystander (talk) 09:24, 10 April 2016 (UTC)
    If I had to choose between significant event (P793) and participant of (P1344)/part of (P361)/present in work (P1441), I personally would use present in work (P1441), which doesn't say, how big deal it is for particular person. For some big deals we could use significant event (P793). --Edgars2007 (talk) 09:29, 10 April 2016 (UTC)
    Note that at present present in work (P1441) is intended to be used for fictional entities only. I'll leave a message on the property talk page so those watching it are aware of the suggestion to broaden it. Thryduulf (talk: local | en.wp | en.wikt) 00:17, 12 April 2016 (UTC)

    Importing from Japanese Wikipedia the name in kana

    The Property:P1814 (name in kana) references the Japanese pronunciation of a name. This information is available in the Japanese Wikipedia pages, so it could be automatically extracted in order to be batch-imported into Wikidata. I wrote a small program to do the extraction (more details below), but I don't know if it would be technically possible to do the importation part (and if this kind of batch-import is generally acceptable). If it is possible, then I could open a discussion in the Japanese forum to check with native speaker that my extraction method gives accurate results and how it could be improved.

    About the extraction: I wrote a Java program (to be cleaned-up, improved a bit and published) that processes the Japanese Wikipedia pages from a database dump. At the beginning of most pages, the name in Kanji (Chinese characters) is present in bold, followed by its pronunciation in kana characters. Example for Tokyo: "東京都(とうきょうと)は".

    • For 33% of the pages, the process has a very good confidence that the kana text is found (the brackets contains only kana characters and spaces).
    • For 15% more of the pages gives kana characters in the brackets after stripping the templates, spaces and dashes at the end (ex: '''鶴田洋久'''(つるた ひろひさ、[[1964年]] - )は).
    • For 18%, I could not parse with confidence the content of the brackets (ex: for Japan, it can contain both nihon and nippon). Maybe it would be possible to build the interface so a human selects the good parts.
    • For 3%, the title contain non-Japanese characters (like digits), and the brackets contains these characters plus kana characters, so the confidence is quite good.
    • For the remaing 28%, no kana is available.

    The program produces a CSV text file (page title;kana text) for each category above.

    From these text files, it would require to find the Wikipedia item associated to each Japanese Wipedia page title, then creating the property if it does not exist yet, and setting the attribute "imported from" to "Japanese Wikipedia".

    So, what do you think about this idea? Edit: Do you have any idea how it could be technically put in place ? (ex: upload the output of my program on a Wikimedia server, write a program/script that would run on it and perform the process described above) — Fabimaru (talk) 21:24, 8 April 2016 (UTC)

    I've considered doing something similar, so I'm in favour of it. :) - Nikki (talk) 10:42, 9 April 2016 (UTC)
    Hi. As you have probably seen, recently i have been doing exactly the same thing. However there are few things I always or never do:
    1. I import P1814 only for people.
    2. I import P1814 only for people, connected with Japan. For this I'm using a number of rules which I combine in different ways, like ?item wdt:P27 wd:Q17, ?item wdt:P19/wdt:P131* wd:Q17, ?item wdt:P106/wdt:P279* wd:Q226008, ?item wdt:P69 ?uni . ?uni wdt:P17 wd:Q17, ?item wdt:P735 ?name, ?item wdt:P734 ?sname. For name and surname I use a heuristic, where I import only items with a single japanese sitelink.
    3. I import only kana with length <= 16 and consisting of cleary identifiable name and surname, separated by space or space-like character. Other information is useless for Wikidata. There is no segmentation algorithm for kana.
    4. I autotransliterate imported names into Russian using Polivanov system (which is accepted by ruwiki community and actually works much better than human transliteration). I guess one could use Hepburn romanization system for English. Maybe @Haplology: could explain, how he created English labels for japanese people few years ago.
    The code is gone but as I recall I used a similar method of conservative regex to isolate hiragana readings of names of people on pages without enwiki sitelinks (other pages may have romanized titles, but romanization styles vary slightly on other wikis) and with the further restriction of only modern-day people, that is, people born after the Meiji Restoration plus a few years to be safe. I skipped katakana names, either in the name itself or in its reading. There was also a lot of eyeballing and checking for errors. The transliteration method was the revised Hepburn romanization in use in WP and WT. --Haplology (talk) 01:21, 11 April 2016 (UTC)
    As for me, I don't see any use for P1814 of non-japanese items. Especially I don't see reasons for importing reverse-transliterated names in katakana: these could be always autogenerated with userscript just from english labels.
    --Lockal (talk) 11:31, 9 April 2016 (UTC)
    Hi. Thank you a lot for your answer, I noticed you added P1814 to the entries of a large number of people, but I did not know your method for importing them (it could have been manual).
    In addition to the people, I think could be also useful to have the places (cities, train stations…) and institutions (schools, companies…), but I guess that it would be a challenge to be sure that the Wikidata entries/Japanese Wikipedia pages belong to these categories (ex: for the item for a temple, we may not have any property set in Wikidata so we can identify it as a temple).
    Before your answer, I thought about importing indiscriminately all the Wikipedia titles containing kanjis (starting from the Japanese Wikipedia, not from the Wikidata entries). Do you see any reason not to do so? I first did not see any reason, but now I can think about these two cases that are arguable:
    • For the Wikipedia pages titles that are not a proper noun. For example, 認知症 / dementia. I don't think it would be good for Wikidata because the Wikidata item is about the concept, not about the noun itself (if the label is changed, the kana name will be of of sync).
    • For Chinese names (which obviously contain Chinese characters). For example, is it useful for Japanese people to know that "Xi Jinping" is pronunced "しゅう きんぺい" in Japanese? (I would tend to think it is)
    So maybe for the types of pages I mentioned, some heuristics should also be used to detect the category (ex for the train stations: if the name ends with "駅"=station).
    If other types of names were to be imported, do you think it would be better to extend your tool or that I complete mine (which currently only has the Wikipedia parsing part)? — Fabimaru (talk) 18:10, 9 April 2016 (UTC)
    I agree that limiting it to proper nouns makes sense for the reason you gave. Don't forget that there are also Korean names where the Japanese name uses Chinese characters (e.g. Kim Dae-jung (Q45785), South Gyeongsang Province (Q41151)). - Nikki (talk) 14:08, 10 April 2016 (UTC)
    If the Japanese label is already in katakana, there's no point duplicating it in a P1814 statement, so that should already exclude most non-Japanese things. The main exceptions would be some Chinese and Korean names. Also, if someone does decide to try to automatically generate katakana versions, please bear in mind that the katakana version depends on the pronunciation of the name, not the spelling (e.g. ja:マイケル・ジャクソン, ja:ミカエル・ラウドルップ, ja:ミヒャエル・バラック all have an English label starting with "Michael"). - Nikki (talk) 14:08, 10 April 2016 (UTC)
    I would suggest you not to import non-segmented kana names. I'm not sure what to do with generic nouns (maybe generate kana name with MeCab/neologd and if the generated value matches the value from Wikipedia, then import it), but for humans you could at least check if kana name consists of multiple parts. Non-segmented sequences of kana characters are useless in terms of further processing. --Lockal (talk) 07:28, 11 April 2016 (UTC)
    Thanks. After trying to understand which names/nouns should be segmented, I now realize that my approach was too naive and unsatisfying. In fact the kana inside the brackets does not always match the kanjis in Wikipedia. For example, for Mitsubishi Motors, the kanji ends with " Corporation ", but the kanas do no contain it. So I guess that if I want to import something, I should restrict myself to entries for which:
    • the kanji names matches for sure the kanas.
    • the kana names don't need to be segmented. My feeling is that the trains stations, cities and temples do not need it (I think that I can start with these 3 ones).
    So I need to narrow and refine my plan, and to see how to use Widar (or any alternative). Fabimaru (talk) 20:25, 11 April 2016 (UTC)

    IEG grant submission for Wikimedia-powered e-books - feedback welcome

    Hi everyone! I have just submitted an Individual Engagement Grant request to research how it would work if one would like to create e-books with mixed content from various Wikimedia projects together. The reason for requesting this grant is that I'm working on a hybrid print and digital (Wiki)book on Dutch digital arts, for which I'd like to do such a thing, and I want to work on a more 'universal' solution from the start. You can find the grant submission here. I explicitly want to include the possibility to work with data from Wikidata in such e-books (think timelines, infoboxes, and multilinguality where possible and useful). Feedback, questions, use cases, and of course endorsements are more than welcome there. Thanks! Spinster (talk) 09:50, 12 April 2016 (UTC)

    Wikidata weekly summary #204

    Input requested for Individual Engagement Grant: creating a database for the public domain

    Hello everyone! We're presenting a proposal for building a database of argentinean authors in the public domain. Our proposal is based on a project already been carried out in Uruguay, called AutoresUy, already accepted as an unique identifier in Wikidata. We think it's necessary and desirable to have more projects of this kind. In order to achieve this, we have set up a plan that includes workshops, edit-a-thons and lots of hard work to build, improve, and expand this database and help Wikimedia projects. Please leave your input in the talk page, or directly an endorsement if you think it's worthed.

    IEG proposal for AutoresAr. --Scanno (talk) 14:55, 12 April 2016 (UTC)

    A big questionmark

    author  TomT0m / talk page Mbch331 (talk) Jobu0101 (talk) Tubezlob (🙋) Flycatchr Koxinga (talk) Fuzheado (talk) Mfchris84 (talk) Manu1400 (talk) Daniel Mietchen (talk) Nomen ad hoc (talk) 07:53, 16 November 2019 (UTC)

      Notified participants of WikiProject Award I think we need to talk about this all together.

    hi, I went bald and implemented an idea talked of in a discussion about awards (see WikiProject Award because the discussion was stunned:

    • An award is the class of all the time someone recieved it. Which means that every award like "nobel price" is a subclass of "award", not an instance.
    • Almost all awards are given every year. I created a metaclass for this kind of awards to be an instance of. This metaclass regroups all awards whose ceremony take place repeatedly every year. So far so good. I just see this edit from Abc82 (talkcontribslogs) : WTF guy, are you again putting almost nonsensical statements on your own ? What is this ? it's not the first time ... author  TomT0m / talk page 13:18, 10 April 2016 (UTC)
    This does not sound right to me. An award in an individual recognization of one or more persons or organizations. Therefore, "the Oscars", "the Oscars 2016", or "the Oscars for cinematography" are classes, whose instances are those individual recognizations. --Srittau (talk) 15:26, 10 April 2016 (UTC)
    We probably need to distinguish between those things:
    • Award type ("Oscars")
    • Award years ("Oscars 2016")
    • Award categories ("Oscar for cinematography")
    • Individual recognizations. ("Oscar for cinematography in 2016")
    I would consider the former three to be classes, while the last one is not a class. --Srittau (talk) 15:30, 10 April 2016 (UTC)
    More or less. I'd agree that the last one is an instance of Oscardization, more precisely an instance of Academy Award for Best Cinematography (Q131520)    . I'd also agree that "Oscars" is a type of award, considering it's the awards given by a single organisation (same conferred by (P1027) for all Oscars, the Academy) ) I actually wanted to implement a metaclass with
    < Award type > has quality (P1552)   < organization that offer the award >
    but I did not find actually the right item. I wonder if periodic price (Q23755142)     qualifies for being this "award type" item. In most of the case I guess. author  TomT0m / talk page 19:23, 10 April 2016 (UTC)
    For the statement that states who won the "Academy Award for Best Cinematography in 2016", instead of creating an item for the 2016 award, why not say the person was awarded with an "Academy Award for Best in Cinematography" then qualified with the year awarded? —seav (talk) 00:22, 13 April 2016 (UTC)

    Cell lines

    Wanting to know how I can contribute to the Wikidata project by sharing the data on cell lines that is actively curated in the Cellosaurus:

     – The preceding unsigned comment was added by Amb sib (talk • contribs) at 08:49, 12 April 2016‎ (UTC).

    Thank you! Check out Wikidata:Data donation. --Denny (talk) 17:10, 13 April 2016 (UTC)

    old names of a sport club ?

    Hi everyone. I would like to add old names of sport clubs. Should I use for this official name (P1448), which seems to be dedicated to places, or the more recent name (P2561), or another one I didn't find ? Thanks. --H4stings (talk) 08:58, 13 April 2016 (UTC)

    Use official name (P1448) with date qualifiers. Sjoerd de Bruin (talk) 11:16, 13 April 2016 (UTC)
    Ok thanks. --H4stings (talk) 14:22, 13 April 2016 (UTC)

    Linking a wikidata item to a subsection of a wikipedia page

    How does one do this? My example is a wikidata item: Gene Ontology Consortium (Q23809253) which I want to link to a tab on this wikipedia page:

    A) is this a bad idea and; B) if not how might it be implemented?


    --Mhaendel (talk) 15:33, 13 April 2016 (UTC)

    Mhaendel It is not possible to link a page with two items --ديفيد عادل وهبة خليل 2 (talk) 15:44, 13 April 2016 (UTC)
    this was discussed previously as a justification for linking to redirects. Currently the way you would have to do that (in this case as an example) is create a stub page in enwiki for "Gene ontology consortium", link to that, then change that stub to a redirect to where you want. This works, but some people object to it... ArthurPSmith (talk) 15:59, 13 April 2016 (UTC)
    A) Yes it's a bad idea and B) it may not be implemented. The workaround described above is a bug. There are several reasons why this is not possible. One is the fact that Wikidata has a 1-1 relationship with Wikipedia articles because of the problems with page moves on Wikipedia to improve article titles. On Wikidata only one field has alternate aliases, and that is the label field. The titles of Wikipedia articles do not change often, but their subsections change more often. This is the same reason that on Wikipedia, there is a link to "Cite this page" but there is no "Cite this paragraph". This is tracked in Phabricator . --Jane023 (talk) 16:31, 13 April 2016 (UTC)

    Go to en.wikipedia page for a Qid

    Hi - if i have a wikidata Q number for an item, is there a url I can use that will redirect me to the equivalent wikipedia page, if it exists (e.g. using the stored sitelinks information). Something like HYanWong (talk) 14:04, 11 April 2016 (UTC)

    Hello! We have this: see Special:GoToLinkedPage. Sjoerd de Bruin (talk) 14:12, 11 April 2016 (UTC)
    Perfect. Thanks! I couldn't find it documented anywhere obvious, so perhaps might be worth flagging up somehow. But it might just be my poor search strategy. HYanWong (talk) 15:01, 11 April 2016 (UTC)
    p.s. it might be helpful to give an example on that page (e.g. I can't edit the text there, otherwise I'd have added it. HYanWong (talk) 15:07, 11 April 2016 (UTC)
    You could use Reasonator. It is included and easy to use. GerardM (talk) 06:35, 14 April 2016 (UTC)

    Keeping track of things

    Keeping track of things on Wikidata is difficult. The broadness of the project's scope, granularity of its content, and diversity of its languages create a situation where normal monitoring tools like watchlists and recent changes are increasingly insufficient.

    Changes, both large and small, should be monitorable without difficulty. Users should be able to easily follow the areas they are interested in, see what changes others are making, and easily interact with the edits and editors involved. This is critical to having a functioning wiki.

    I've put together a list of the most pressing problems in this area, along with my proposals for solutions and implementations of those solutions. Please comment, and feel free to propose alternative solutions and/or implementations.

    Keeping track of mass-edits

    • Problem: Automated or semi-automated mass-edits are hard to keep track of or undo.
      • Solution: A central, trackable log of all mass-edits of over 100 edits. This includes both bot runs and large groups of edits done using tools like Autolist. Users should be able to follow the log via watchlist, understand the parameters of the edit groups, and easily undo individual groups of changes. Finding which run an edit belongs to should be simple.
        • Implementation: A log as a wiki page, with each group of edits having a listing, probably using a template. Ideally, tools like Autolist would add log entries automatically, filling in information such as precisely what source is used for determining which items to edit (SPARQL, WDQ, Category, etc), and what edits are made, a timestamp of when the run started and ended, and links to relevant discussions if applicable. If automatic updates are not an option, then the log could be filled manually using either preloaded templates or the form wizard. An undo button next to each run links to a preloaded filled-out form for User:RollBot, which simultaneously pings the changes' author. RollBot's undos should have relevant information in the edit summaries. The log is in chronological order and split by year (or maybe month?), and all edits within the group would ideally link to the log entry in the edit summary. --Yair rand (talk) 22:10, 11 April 2016 (UTC)

    Recent Changes flooding

    • Problem: The Recent Changes are frequently clogged by mass-edits.
      • Solution: Encourage more use of the Flood flag. Requesting the flag should be simpler. Users about to do mass-edits or in the process of doing them without a flood flag should be notified, and provided with a helpful link.
        • Implementation: Tools should warn users about to do mass edits if they don't have the flag set, and provide the necessary link. A talk page template for unauthorized bots and unflagged flooders should be created. Perhaps a bot could automatically warn all users doing too many edits too fast. WD:FLOOD should prominently link either to the correct noticeboard for requesting the flag, or even directly to the form. A flood flag request template should include a direct link for bureaucrats to enable the flag, and for the requester to remove it afterwards. Perhaps the template should also ping all current bureaucrats, in order to achieve faster response times.
          (It may be worth it to have a [default|rights=flood] gadget add a button to the sidebar or the template to simultaneously remove the flood flag and close the current mass-edit run log entry, if both proposals are implemented.) --Yair rand (talk) 22:10, 11 April 2016 (UTC)
          • I think if we really want to encourage people to use the flood flag, it would need to be more automated, e.g. QuickStatements is really useful and there's plenty of people using it. It would be easier for QuickStatements to set/remove the flag itself instead of trying to convince every single person who uses it to request the flag manually every single time they want to use it. - Nikki (talk) 06:46, 12 April 2016 (UTC)
            • @Nikki: Non-admins can't add the flag to their own account. --Yair rand (talk) 15:32, 12 April 2016 (UTC)
              • I know. We're talking about ideas of what we could do/change to improve things though, right? :) Another idea: Improve tag filtering. It seems to be possible to show things with a certain tag, but not to filter out things with one of a list of tags. The vast majority of flooding I see comes from QuickStatements which are already tagged (although not uniquely since they're lumped under Widar, but perhaps that could be changed). - Nikki (talk) 17:59, 12 April 2016 (UTC)
                @Nikki: Below I introduce the idea of a group of "Self-propmoters". -- Innocent bystander (talk) 14:57, 13 April 2016 (UTC)

    Never heard of the flood flag, and I am a heavy user of quick statements. It might help to inform the people what you are proposing exactly, because I still don't understand after reading this. --Jane023 (talk) 07:37, 12 April 2016 (UTC)

    The Flood flag is a kind of "Bot flag light". Your edits are hidden in RC, but you do not have the other rights bots have. It can be granted in a more simple way and you can remove the flag yourself. -- Innocent bystander (talk) 10:00, 12 April 2016 (UTC)
    Thanks for the explanation. I can see that it is something that has no benefit to the task at hand, so the motivation to use it would be low. I asked for a bot flag so I could use quick statements in a way that was faster for me, but so far I have the difficulties of logging in and out with a different user account and no significant time gains. Your idea of building this in to quick statements is probably the best way forward. I think we had that in the beginning but people were confused about it and didn't trust those edits. I am not sure this will help that problem. --Jane023 (talk) 18:06, 12 April 2016 (UTC)
    @Jane023: I applied as sysop to be able to set and reset this flag myself, without asking our crats every time. -- Innocent bystander (talk) 18:24, 12 April 2016 (UTC)


    @All: Starting a new user group of "Self-promoters" who like sysops are trusted to set and unset their own flood flag would maybe be an alternative? It is maybe easy to get trust enough to become a sysop here, but not all of us feel comfortable with the extra tools associated with this group. I have myself, by the help of some css-magic, removed the rollback-links which are easy to hit by accident. Without that option I would probably never have applied as sysop (again). -- Innocent bystander (talk) 18:24, 12 April 2016 (UTC)
    I like this idea. --Yair rand (talk) 20:30, 12 April 2016 (UTC)
    +1. Yellowcard (talk) 15:15, 13 April 2016 (UTC)
    +1. This seems like a very good use of our permissions system. —seav (talk) 03:00, 14 April 2016 (UTC)
    I have no opinion on its desirability, but please call it something else: self-flaggers, auto-flaggers, floodflaggers, etc. A self-promoter is something to be combatted. - Brya (talk) 05:45, 14 April 2016 (UTC)
    @Brya: "Self-promoters" is a description of a potential new group, not a proposed name of it. I think it looks like the idea already has enough support to start a RFC about it. Decisions like this, who affects the whole project cannot be taken in a <h4/>-sub-thread in the English Project Chat. -- Innocent bystander (talk) 05:56, 14 April 2016 (UTC)

    Change list difficulties

    • Problem: Change lists, such as Recent Changes and watchlists, are too hard to read through and contain too many edits clearly outside of one's area of interest. As a result, relevant changes are missed. Furthermore, on client-wiki watchlists (Wikipedia watchlists, etc), the lists are so full of irrelevant changes, and the changes listings contain so little information, that users tend to either disable showing Wikidata changes or ignore them.
      • Solution: Change lists should allow for filtering, and users should be able to see all the details of an edit without clicking "diff" on every one of them.
        • Implementation: DiffLists.js offers a basic version of this, with limited by-language and by-property filtering, and shows a smaller version of the edit diff instead of the regular autogenerated edit summaries. However, as a result of being entirely on the front-end, the script has substantial limitations. A proper back-end version would have many advantages, such as more reasonable performance, potential for widespread use without smashing the servers, and the ability to display a stable number of results, allowing for more usable property and language filtering.
          (Other useful potential features: A better property-selecting mechanism and language selector, support for history and contributions pages, and defaulting language settings to babel box contents.)
          Client-wiki watchlists, by default, should hide all identifier statements and foreign-language labels/descriptions/aliases, and possibly even interwikis to languages the user does not speak.
          phab:T121361 is the task for showing diffs on the change lists. Afaik, there isn't one open yet for filtering options. --Yair rand (talk) 22:10, 11 April 2016 (UTC)
    +1, I love the DiffLists tool. I don't remember how I could be able to use the recent changes list before, and I also think it would benefit greatly from being implemented within Wikibase.
    Additionally, I'd like to see more people using the patrol feature (which is totally independent from DiffLists, just to clarify this). When looking for vandalism, it is plainly impossible to watch all changes on Wikidata, you can always only look at a very small excerpt. At the moment it's not even really possible to check all unpatrolled edits for a prolonged period of time, even with the DiffLists tool, which significantly reduces the effort you have to take. So if most people would mark most edits they have checked (and kept or reverted) as patrolled, we would win a lot already. --YMS (talk) 10:37, 13 April 2016 (UTC)

    Keeping track of topic areas

    • Problem: Topics spread across many items are hard to follow. Certain areas are divided into enough items that it's difficult to watch them all, or keep a watchlist updated with new "sub-items". The recent changes feed is too broad, watchlists not broad enough, and normal use of "Related pages" too limited.
      • Solution: Individual Wikiprojects could have separate change feeds, based on lists of relevant items and properties.
        • Implementation: Much of this would be best based off of lists automatically bot-updated from SPARQL queries. The feed could come from Related Changes of the list, and be prominently linked to from Wikiproject pages. Some more filtering of the feed could be done by a server-side version of DiffLists, as described above. (Show only relevant statements and such.) The details of the filtering could be included within the link ("?rc-filters=", etc), and filled in automatically. Some manual maintaining of the list of relevant items may also be necessary. For the query-generated parts, there could be an "expiry" system for removal, as opposed to immediate removal, so that statement removals/changes that would ordinarily remove the item from the list would still show up in the feed. --Yair rand (talk) 22:10, 11 April 2016 (UTC)

    Discussion dispersion

    • Problem: Discussion areas are far too scattered, causing people to be unaware of discussions they would be interested in. Every individual item has a talk page, and it's often hard to find a more general discussion page for a topic, or otherwise notify everyone who would be interested in participating.

    General discussion

    Illegal language error

    I am trying to add title (P1476): "Binny und der Geist" to Binny and the Ghost (Q17457631). For the language, I typed German and chose "German (de)" from the dropdown. When saving, it says "Illegal language: German". nyuszika7h (talk) 15:28, 11 April 2016 (UTC)

    It's as if it ignores what I choose from the dropdown. Manually typing "de" worked. nyuszika7h (talk) 15:31, 11 April 2016 (UTC)
    Firefox 45 on Windows 10, if that matters. nyuszika7h (talk) 15:31, 11 April 2016 (UTC)
    @Lydia Pintscher (WMDE): This seems like something you might want to look into :) It sounds quite similar to what I wrote about in my last comment on phab:T110043 - Nikki (talk) 08:47, 13 April 2016 (UTC)
    It seems like selecting a suggestion doesn't update the input, so "German" stays "German" instead of "de". Sjoerd de Bruin (talk) 08:53, 13 April 2016 (UTC)
    Thanks! Looking into it. --Lydia Pintscher (WMDE) (talk) 14:01, 14 April 2016 (UTC)

    How could we ...

    Following the previous thread by Yair Rand I'll try a mini-rfc.

    First round on questions for now (both choices can be "yesed" if you want that all three stuffs are aligned :

    See Wikidata:Requests_for_comment/Reforming_the_property_creation_process#Reorganization.3F for some earlier discussion and related proposals on this. How does an RFC get concluded anyway? TomT0m it might be good to list all the specific reorganizations that have been proposed so far here? ArthurPSmith (talk) 20:21, 12 April 2016 (UTC)

    Alignement of proposal categories on class hierarchies

    1. Should we try to align the property proposal categories to the class hierarchy ? This would concretely mean that each section of WD:PP would map to a top class of our class hierarchy.


      Support --Jane023 (talk) 12:45, 13 April 2016 (UTC)




    Alignment of property proposal structure to wikiprojects

    2. Should we try to align the property proposal categories to the wikiproject structure ? I have no concrete idea on how we could implement that. A property proposal section in each wikiproject ? And wd:pp would link/transclude these pages ?





    You lost me on this one. No idea what you mean at all. --Jane023 (talk) 12:46, 13 April 2016 (UTC)

    Documentation of classes

    1. Should we focus more on documenting classes instead of properties ? Every class has a set of natural properties that can be applied to her (and its parent property), can have a showcase instance, ... This would mean that we could be able to document several properties on the same page instead of on all the property pages, which leads to discussion and information dispertion.

    I think we should probably have documentation on both classes and properties. There would be some duplication, but I think it would be worth it. --Yair rand (talk) 20:28, 12 April 2016 (UTC)





    Property proposals organized by class

    2. Should we try to associate a property proposal to a class ? This might be implemented as such : one property proposal page for each class, all this pages indexed in wd:pp or one of its subpages. There could be a path for example "wd:pp" -> "abstract object" -> "creative work class" -> "live performance class" -> "opera" -> existing properties for an opera (or its parent class). Then the user could find what he's looking for (or not) and propose something on the property proposal of a page. There is a "all proposals" page where all proposals have to be listed, whith just a minimal description and not all the details of the proposal.





    The above section was added by TomT0m, 14:00, April 12, 2016‎

    I would really like it if each proposal were its own subpage. Those could be included on whichever pages we want and easily moved between pages without losing any history. The current pages are too large for me to keep track of and since I can't add individual proposals to my watchlist, I often end up missing discussions that I'd like to follow. - Nikki (talk) 20:56, 12 April 2016 (UTC)
    +1 to Nikki's suggestion. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 18:23, 13 April 2016 (UTC)
    I wanted to propose this at last RfC. One thing to think about is naming. At enwiki's AfD this is easy - simply page title. Not sure, what could be used here at PP. Random identifier - not good, property name - they can change very differently, and not everybody is adding titles in English (they should, but as this is not "English Wikidata", this can't be obligatory). But OK, naming is pretty minor thing to come up with. --Edgars2007 (talk) 07:14, 14 April 2016 (UTC)
    +1. Another advantage of this is that we could transclude a proposal to multiple overview pages. That means we could have pages sorting the proposal by classes, and other pages sorting the proposal by age or status. --Pasleim (talk) 13:57, 14 April 2016 (UTC)
    +1 from me too. This is how it's done in the Requests for Permissions pages here, for example Wikidata:Requests for permissions/Bot. It seems to work well there. I would just use the initial property name (in any convenient language) as the page name; if that is a duplicate of a previous entry then add a number. If the proposed property name changes during the course of discussion I think it would be fine to still have the discussion at the page pointed to by the original name. It would never need to move to another page; after closing the discussion it should be auto-archived the way those bot requests are. This probably should be a full RFC though, not just a chat topic. ArthurPSmith (talk) 18:20, 14 April 2016 (UTC)

    Help translate Listeria bot on local Wikipedia projects

    Hi all, I am working with User:Pigsonthewing on the Wikidata:TED data donation. I have built lists of speakers and talks and would like to build these lists on those local Wikipedia projects for which we have translated titles. See the overview here: outreach:TED conferences. I am missing Listeria bot however for the Wikipedia languages ja/ko/zh/tr/vi/fa. Is there anyone here that can help out? A local template needs to be created and then bot access for ListeriaBot needs to be applied for and granted. Andy, Magnus and I don't speak any of those languages, so it would be great if someone here could help out in those projects who 1) understand what Listeria bot is and 2) understand how to apply for a bot flag on their local wiki. Thanks! --Jane023 (talk) 12:22, 13 April 2016 (UTC)

    To clarify, the list templates Template:Wikidata list and Template:Wikidata list end must be created on the local project (preferably with redirects so non-local users like myself can still navigate with them). You can see the template in use for many things if you click "What links here" from any of the template, but for example I also just used it here: Talk:Q19507487. --Jane023 (talk) 08:37, 14 April 2016 (UTC)
    And now you can see an overview of what User:ListeriaBot does here Botstatus. --Jane023 (talk) 17:30, 14 April 2016 (UTC)

    Please merge

    Q9483060 and Q8475387 are now one and the same, following the outcome of Wikipedia:Categories_for_discussion/Log/2016_February_9#Category:Functionaries_of_the_Stalinist_regime_in_Poland. I don't know how to fix the issue of making sure the interwiki links work (i.e. pl:Kategoria:Funkcjonariusze Ministerstwa Bezpieczeństwa Publicznego should now be part of en:Category:Ministry of Public Security (Poland) officials level. --Piotrus (talk) 08:15, 14 April 2016 (UTC)

    I have moved the link for you. I'm pretty sure we have a help page for this...? --Izno (talk) 13:24, 14 April 2016 (UTC)

    Copy reference

    Hi. Is copy reference tool working or not? I can only copy a reference, not add it. Xaris333 (talk) 17:46, 25 March 2016 (UTC)

    Can confirm, the following JavaScript error is given: TypeError: null is not an object (evaluating 'this._qualifiers.value'). @Bene*: can you take a look? Sjoerd de Bruin (talk) 17:49, 25 March 2016 (UTC)

    Can anyone help? Xaris333 (talk) 12:25, 31 March 2016 (UTC)

    I tried to fix it; Does the gadget work again for you? Adrian Heine (WMDE) (talk) 06:39, 7 April 2016 (UTC)

    Adrian Heine (WMDE) Just try it. Still not working. Xaris333 (talk) 18:39, 7 April 2016 (UTC)

    I posted a hopefully correct fix on the gadget's talk page, but no sysop responded yet. Adrian Heine (WMDE) (talk) 11:49, 12 April 2016 (UTC)

    Lets hope that it will fix it. We really need that gadget. Xaris333 (talk) 12:01, 12 April 2016 (UTC)

    @Xaris333: It should be fixed now if you want to test it. - Nikki (talk) 20:00, 14 April 2016 (UTC)
    Nikki still not working. Xaris333 (talk) 22:25, 14 April 2016 (UTC)
    @Xaris333: It's working fine for me - could you provide some more information? e.g. which browser? what happens? if you know how to get to the error console, are there any errors there? have you tried a forced refresh? - Nikki (talk) 12:25, 15 April 2016 (UTC)
    Nikki Pls try to use it to one page and then to one other page. Yesterday it worked for me when I try it (after you said me to test it). But only for a page. Then stopped again. I have tried refresh. Chrome. I don't know about error console. Xaris333 (talk) 13:55, 15 April 2016 (UTC)
    I have used it on multiple pages and it still works for me :/ I'm not sure what to suggest... @Adrian Heine (WMDE): any ideas? - Nikki (talk) 14:05, 15 April 2016 (UTC)

    Its working now! Xaris333 (talk) 17:40, 15 April 2016 (UTC)

    Emoji 😀

    Emojis are now avaible on Wikidata   --Tubezlob (🙋) 20:59, 12 April 2016 (UTC)

    Emoji palette