Shortcuts: WD:PC, WD:CHAT, WD:?

Wikidata:Project chat

From Wikidata
(Redirected from Wikidata:Project Chat)
Jump to: navigation, search
Wikidata project chat
Place used to discuss any and all aspects of Wikidata: the project itself, policy and proposals, individual data items, technical issues, etc.
Please take a look at the frequently asked questions to see if your question has already been answered.
Please use {{Q}} or {{P}}, the first time you mention an item, or property, respectively.
Also see status updates to keep up-to-date on important things around Wikidata.
Requests for deletions can be made here.
Merging instructions can be found here.

IRC channel: #wikidata connect
On this page, old discussions are archived. An overview of all archives can be found at this page's archive index. The current archive is located at 2017/06.





for permissions


for deletions


for deletion

for comment

and imports

a query



"Single value" constraints and deprecated values[edit]

We could do with either changing our "single value" constraint so that deprecated values are (optionally) ignored, or having an alternative constraint that could be applied to that end (without the need to write complex constraint code). Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 16:43, 12 June 2017 (UTC)

Not just for deprecated please. E.g. for Thailand central administrative unit code (P1067), there is one current single value, but there may be several older ones - which then have normal rank, not deprecated, as they were correct in past, whereas the current value has rank high. So single value should just check for single value in the highest used rank. Ahoerstemeier (talk) 21:43, 12 June 2017 (UTC)
I'd also suggest that it pay attention to whether there's a start/end date on entries. There could be two valid historic values for something, both marked 'normal', and with suitable start/end dates, but no current version to mark as preferred. --Oravrattas (talk) 06:52, 13 June 2017 (UTC)
In that case`, why not add a "no value" with high rank to mark that there is not current value? Then it'd work without problems by checking on the highest used rank. Ahoerstemeier (talk) 07:52, 13 June 2017 (UTC)
That is fine in the case where it's definitively known that there is no valid value at the moment, but it's possible to have a case where there are historic versions that are known to no longer be valid, and where people have correctly added the information for those. Forcing them to then either find a new value, or explicitly choose between 'no value' and 'unknown value', when they may not have that knowledge, simply to satisfy a constraint check, violates the "missing is not broken" principle. --Oravrattas (talk) 13:51, 13 June 2017 (UTC)
The constraints are not enforced, so they don't force anyone to do anything. - Nikki (talk) 14:07, 16 June 2017 (UTC)
I think we can go with this: "single value" now means "single best value", ie. the highest rank in a statement group should only occur once. Matěj Suchánek (talk) 14:16, 16 June 2017 (UTC)
When it comes to authority control, where there should be only one correct value, I don't think 2 statements with rank normal and 1 statement with the preferred rank should pass the contraint.
I think that the default "single value" constraint should allow depricated values but limit the amount of other values to one. We can have an additional "single best value constraint that also allows multiple statements with the normal rank along one statement with the preferred rank. ChristianKl (talk) 13:32, 18 June 2017 (UTC)

@Lydia Pintscher (WMDE): FYI. I suspect you'll want a Phabricator ticket, but this is you opportunity to comment first. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 21:08, 23 June 2017 (UTC)

We have phabricator:T167653 already it seems. My current thinking is that "single value" means there should only be one best ranked statement. Are there cases where this leads to data we do want to flag as not being flagged? --Lydia Pintscher (WMDE) (talk) 09:23, 26 June 2017 (UTC)

GeoNames and, the first to throw a stone...[edit]

I exchanged e-mails with Marc Wick, the founder of GeoNames. It turns out that he is looking into connecting GeoNames to Wikidata. He finds problems that need to be resolved because of errors on our end. Errors are to be expected.

I asked permission and this is where you can find the e-mails so far. As I have asserted before, when there is cooperation any complementary data is welcome because it improves an 100% error rate to something that is more manageable. The whole history with Wikipedias importing data from GeoNames is imho our own problem. When we welcome the data they want to import and provide the platform to generate texts, texts that are cached and not saved, our efforts to curate data will have the biggest effect.

What is quite clear to me is that only the first to throw a stone at GeoNames should be the one that makes no mistakes. As we all make mistakes collaborating is the only sane way forward. Thanks, GerardM (talk) 08:31, 17 June 2017 (UTC)

Since I already have tried to work with this dataset, I would like to start to pinpoint some weak points. The links to Wikipedia in GeoNames can a little to often not be trusted. Unfortunately we some years ago started to add GeoNames ID (P1566) here based on those WP-links. A little to often, the only thing the WP-articles and the GeoNames-items have in common is the label. A second problem is that the coordinates are rounded to minutes. That is not very useful when it is used to pinpoint a specific summit (Q207326) among others on a mountain. It also makes it difficult to find where a specific reef is IRL. A large problem is how the dataset in GeoNames has been collected. It looks like any name on a map have been harvested. But what those names describes has not always been identified in a good way. Beyne-Heusay (Q681039) is a commune in Belgium. There is no populated area with this name. I live fairly close to this point. I can tell you that there is no populated place with the name "Ytterlännäs". Ytterlännäs is the name of several existing and former administrative units, but there has never existed a populated place with this name. The namesake of these administrative units is Näs, located next to the old church of Ytterlännäs.
The problems on svwiki has not only been the quality of this source. How the project was implemented and what complementary sources was added together with the algoritms used by the bot also have affected the project. I do not in any way oppose any cooperation between GeoNames and Wikidata. In fact, I would have preferred if Lsjbot would have started here instead of on Wikipedia. That would have included more experienced users in the discussions about how the project should have been implemented. -- Innocent bystander (talk) 10:25, 17 June 2017 (UTC)
Thank you for your thoughtful reply. Yes, there are issues and these issues are found at both Wikidata and GeoNames (see blogpost). What we really need is a way to communicate issues in both directions. We are good at the things we do but we communicate badly. I am not a Wikipedian and I find many issues with awards that are fine in Wikidata but still problematic at the Wikipedias I frequent.
If there is one thing I wish for the strategic process is that communication of issues gets a better facilitation. Thanks, GerardM (talk) 11:15, 17 June 2017 (UTC)
Yes, the definition of "city" is one of the largest problems we have here. The Swedish article Stad mentions several different meanings of the word in Swedish alone. It could for example mean both an administrative entity or a larger settlement who is identified with one single name. The original meaning is "place".
Another problem worth talking about is the difficulties to identify what a "University" or "School" is. Both Wikidata and GeoNames have this problem. Universitys and Schools are something else than a building. The Lsjbot-articles about Universitys became something to laugh about. A University is located in B, it has climate C and is surrounded by D. I worked some time in Linköping University (Q782600). Telling that it is located at a specific place and describing the nature around it does not make sense. It has activity in several places in Sweden. The same thing could also be said about many things. Is a Municipality an organisation or a geographic feature? I work with a lot of people in the local municipality, and we mainly discuss such things as social care, economy and health care. We never discuss geography. -- Innocent bystander (talk) 15:04, 17 June 2017 (UTC)
In practice I think we should subclass the concept of city more frequently. When in German a city is a human settlement with at least X people and the French concept is one with at least Y people it makes sense to have a concept of "German city" and one of "French city". ChristianKl (talk) 17:31, 17 June 2017 (UTC)
Subclassing based on a local understanding? It should suffice to have the number of inhabitants for places. It is the user who understands the concept of city in a particular way. Having a subclass of French city only means city in France to me. Thanks, GerardM (talk) 18:40, 17 June 2017 (UTC)
Well, it is not as simple as body counting. If a place has a population of 200,000 but still has the infrastructure of a village, it is still today called a village ("by") in Swedish. A place here who do not have a liquor store and a pharmacy can here hardly be called a city. -- Innocent bystander (talk) 10:05, 18 June 2017 (UTC)
We do normally do this. I'm not sure why we don't yet for Germany. - Nikki (talk) 09:55, 18 June 2017 (UTC)

While working on the Thai localities, I also noticed several wrong entries in geonames when it comes to the populated place category - luckily I could stop the ceb-wikipedia bot from importing the whole of geonames in Thailand before it was too late. However, it seems most of these were imported from yet another database, which we have as GNS Unique Feature ID (P2326), so the blame for wrong entries goes one step further. Also, the above mentioned "Ytterlännäs" originates from GNS-UFI -2536857. Sadly, it seems the original ID from GNS isn't found in the geonames data anymore. That said, it is still a good idea to closer link geonames with wikidata, but don't blindly import from geonames. Ahoerstemeier (talk) 21:03, 17 June 2017 (UTC)

When we do not have data, we have a 100% failure. When we are not to blindly import, what scenario do you propose? Thanks, GerardM (talk) 10:36, 18 June 2017 (UTC)
One common mistake in GeoNames is that they mix up items with the same label. This item has the name "Viksäter" and GeoNames tells it has a population of 309. But this is a village with only a handful of houses. If they do not hide 300 aliens in the barns, that data is completely wrong. The Viksäter who had a population of 309 is this as of 2005. Note that the data was removed in the history of the item. But it was rollbacked. There are no official records of how many people lives in villages like this, at least we have not had any for 100 years. Wrong data is only useful as fiction, and that is not what I am interested in here! -- Innocent bystander (talk) 11:59, 18 June 2017 (UTC)
We don't have "no data". Wikidata has some data. If we take data about cities there are use-cases where it's more important that the data we have is trustworthy, that there's a lot of data. False positives can be more harmful than false negatives and it's usually more work to remove false positives as it takes more human analysis. ChristianKl (talk) 18:40, 18 June 2017 (UTC)

I have no idea about the quality of the GeoNames database but it seems that the GeoNames ID (P1566) we already have in Wikidata are also to blame. Following up on User:ArthurPSmith's import of GRID I have set out to add headquarters location (P159) to the company items he created, using GRID's data. GRID provides GeoNames ids, but it turns out that I get more accurate reconciliation results by ignoring them, because of the poor quality of the claims on Wikidata. This is a shame, because in many cases GeoNames ids bring a real value by successfully disambiguating between cities. For instance, we have

< Seoul (Q8684) View with Reasonator View with SQID > GeoNames ID (P1566) View with SQID < 1835847 >

but GRID uses 1862415, which seems to be the correct value. I observe a disagreement in 20% of the reconciled cities.

It seems that many disagreements are of this king (the confusion between a big city and an administrative territorial entity that contains it and has the same name). That could probably be fixed automatically, but I do not have the time to look into that. − Pintoch (talk) 17:48, 18 June 2017 (UTC)

Some facts about GeoNames ID (P1566) and Lsjbot (Q17430942)[edit]

The mass creation of bot generated local Wikipedia articles only based on GeoNames (Q830106) is similar to that ones only based on Catalogue of Life (Q38840) by Lsjbot (Q17430942). In both cases we are enforced to have a (sometimes dubious) labeled item and to get in (by bots) more questionable property values or relations. Obviously the creation was done without sanity checks. Collaborating with external data providers is a good thing, but first of all I'm missing a way to a give straight feedback to the Wikimedia communities that are responsible. --Succu (talk) 21:55, 18 June 2017 (UTC)

It is now June 2017. This happened and as I said before, I prefer to have the data first in Wikidata and use caching to provide the texts into the Wikipedias. At GeoNames they have noticed that our data is wrong in places, they can collaborate with us. My assertion is that when there is collaboration, we can do better for both our projects.
This is not about feedback to the Wikipedias involved. This is about collaborating with GeoNames. Thanks, GerardM (talk) 01:08, 19 June 2017 (UTC)
I think this collaboration would be most welcome! − Pintoch (talk) 23:37, 19 June 2017 (UTC)
@Succu: The articles are not only based on GeoNames. They are based on data from Nasa and some other sources together with some algorithms who interpret in what range a mountain is located and the size of lakes etc. A deep review of how it handled Finland shows that the algorithms had severe flaws, so also the complementary sources (which tells Finland has savanna). Add to that much of the data is based on coordinates that sometimes are wrong. I stopped supporting it when we reached Faroe Islands (Q4628) and I discovered that big parts of the poor nation was located on the bottom of the Atlantic Ocean.
I love to see a collaboration with GeoNames. But if it means that much of the data from that database is imported here without review of the quality of each part, I recommend against it. The flaws are not only in our end. Au contraire, they are much worse at GeoNames! -- Innocent bystander (talk) 19:19, 20 June 2017 (UTC)

Integrate the merge function into the default UI[edit]

Currently, a user has to activity a gadget to be shown the merge function. This means that new users are discouraged from merging items. Giving that we have a huge merge backlog, it might be worthwhile to integrate the merge function by default. What do you think? ChristianKl (talk) 07:30, 18 June 2017 (UTC)

I agree this would be very useful. We can enable the gadget by default, that's by far the easiest way to do it. Opinions please! :-) Multichill (talk) 10:55, 18 June 2017 (UTC)
I think the gadget lacks some quickly accessible explanations how to use it, what it does and some warnings when not to use it. This is necessary to prevent accidental misuse. Matěj Suchánek (talk) 15:28, 18 June 2017 (UTC)
@Matěj Suchánek: Do you think there's significant accidental misuse in the status quo? ChristianKl (talk) 20:33, 18 June 2017 (UTC)
@ChristianKl: Have a look at these contributions. In the last month there were 50 unpatrolled redirects created by IP's, where in the last 2,5 hours 50 redirects were created by users who were logged in. Q.Zanden questions? 21:33, 18 June 2017 (UTC)
(edit conflict) Just a question (unclear to me): Do you propose enabling for all logged in users or just everybody?
There will always be some misuse since mistakes always happen. But the new audience to get this gadget enabled (ie. newbies) is more likely to make mistakes. This is my concern, hence my proposal. Matěj Suchánek (talk) 20:57, 18 June 2017 (UTC)
My preference would be to enable it for everybody. Sometimes this will indeed lead to mistakes but I think that the amount of increased engagement is more valuable than the mistakes that will be made. I would also be okay with enabling it for logged in users. ChristianKl (talk) 22:44, 18 June 2017 (UTC)
Symbol support vote.svg Support. − Pintoch (talk) 20:47, 18 June 2017 (UTC)
ChristianKl - while I think this would ultimately be the right thing to do, there are several features that I believe are not now present in the merge gadget that I think should be addressed before it receives more widespread use:
  • I don't believe it recognizes different from (P1889) which should prevent merges between the respective items if it is present on either of them. There are perhaps some other properties that should also prevent or at least raise a question about merges (for example if country (P17) is set for both and differs, or if coordinate location (P625) differs by more than, say 10 km (?)).
  • the description in a given language on the merged item is completely lost if the item it is being merged into already has any sort of description in that language (and many old items have bare-bones descriptions auto-generated from P31's). There should be a way to pick the better description or merge descriptions in some way if appropriate.
  • merging should sensibly handle the case where one item has a wikilink to a redirect and the other has the direct link (i.e. allow the merge in this case even though both pages have links in the same language).
otherwise I think merge really needs to be limited to more experienced wikidata users who can recognize and know how to deal with these issues. ArthurPSmith (talk) 23:04, 18 June 2017 (UTC)
  • Pictogram voting comment.svg Comment ... "we have a huge merge backlog ..." citation needed! Or maybe it is language specific. If we think it is true, can we demonstrate that?
  • +1 to Matěj Suchánek and ArthurPSmith's statements. I would prefer that there was a new right developed for that could be auto-assigned after reaching an edit count criteria, rather than all users at any time. There definitely needs to be some better supporting information. Numbers of us more experienced users have fallen over with categories to main; taxonomy; settlement to municipal where there has been errors in the existing links, or other errors that make them look the same.
  • We need an improved "danger" list or better builds to prevent merges apart from "different from". I have seen "book/literary work/..." (or similar) and "edition (Q3331189) merged which is problematic.  — billinghurst sDrewth 01:47, 19 June 2017 (UTC)
  • The fact that you see books being merged with edition (Q3331189) doesn't necessarily mean that someone used the merge tool to do it. Magnus merge game suggests items to be merged when they share labels or aliases. When the book and the edition share a name it would suggest them to be merged. ChristianKl (talk) 07:53, 19 June 2017 (UTC)
    Pictogram voting comment.svg Comment @ChristianKl: then please have those instructions updated or removed as it is simply inappropriate, and we need a stronger means to prevent such. Further, I would suggest that edition (Q3331189) should basically not be presented with the suggestion to be merged, it is very unlikely we will need to merge editions  — billinghurst sDrewth 13:05, 24 June 2017 (UTC)
  • I think we have a merge backlog, because in the SPARQL query that's supposed to look for items that are supposed to be deleted that's shared on the German Project Chat the event that items should be merged happens more frequently than that they should be deleted. In the recent RFC about allowing redirects there was also a concern that it increases an already existing merge backlog. ChristianKl (talk) 07:53, 19 June 2017 (UTC)
    Please provide a link to the query so we can all see the evidence. PLbot produces lists of works to merge, and from my observations of English language components it would appear to be in a managed state.  — billinghurst sDrewth 13:12, 24 June 2017 (UTC)
  • PLbot can only find items that have actually the same name. When working through I found a bunch of items where merging made more sense. At the moment, for some reason that query seems to time out. ChristianKl (talk) 11:32, 27 June 2017 (UTC)
  • If we want to couple merging to a user right, it might make sense to use the already existing auto-patrol flag. It could work for both the merge tool and for the merging game. ChristianKl (talk) 13:03, 19 June 2017 (UTC)
    Yes you could merge it if you thought that they should be co-assigned. That said, that is an applied right either by application or grant of administrator, so it will not necessarily meet your needs unless administrators are pro-actively assigning the right. And to note that you would need to do a little magic through your gadgets as people should still be able to turn it off if they do not wish to see/have the merge functionality. As it is already a gadget, I don't think that you will get best value to co-assign it that way.  — billinghurst sDrewth 22:53, 26 June 2017 (UTC)
    Another thought. If we have some editing criteria, eg. 1000 edits, as a point when we think that someone knows enough to merge, we could just run a bot through finding users who pass a milestone value during the past <name your time period> and leave a note about the gadget.  — billinghurst sDrewth 22:58, 26 June 2017 (UTC)
  • Given that the user has to click on "More" before they see the possibility to merge, I can't imagine why someone wouldn't want to have the merge function enabled. ChristianKl (talk) 11:32, 27 June 2017 (UTC)

Release date of computers[edit]

Hi. I can't find a common criterion for a property to mark the release date of computers:

  1. ZX Spectrum (Q23882)
  2. Amiga (Q100047)
  3. Apple II series (Q201652)
  4. Didaktik (Q576901)

Which one would be preferred one? {{subst:Usuario:Roberpl/Firma}} 07:33, 19 June 2017 (UTC)

inception (P571). Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 07:54, 19 June 2017 (UTC)
The start of production isn't the release date. Apart from that inception works as per Pigsonthewing. ChristianKl (talk) 11:01, 21 June 2017 (UTC)

Phone, tablet, laptops technical specifications[edit]

Hi, I'm planning to buy a new tablet and decided to use service to find the best option. Unfortunately I found, that there is not much data about tablets even for the popular ones. For example look at Samsung Galaxy Tab S3, at the time I'm looked at it, it had zero statements. While Wikipedia info box has all the data I need, but not querieble. Then I decided to fill in missing data manually for a few devices I'm interested in. But unfortunately there is no properties for that kind of things either. So my question is, is it possible to create and use needed properties for phones, tablets and laptops without long approval period. I don't know how long it takes to get a proposed property accepted? And I need a bunch of properties. Also I have good Python skills, and could write a script or a bot, to import all this data from Wikipedia info boxes if someone gives me directions how to do that. But of course, without properties, it will not be possible anyway. Sirexo (talk) 18:51, 20 June 2017 (UTC)

Properties are given a minimum of one week for review. If you have a good case, good examples, and fill out the proposal template correctly (see some of the existing proposals at WD:PP) then it should be straightforward. Do search the existing properties first to see if there's already something that meets your needs though. ArthurPSmith (talk) 20:54, 20 June 2017 (UTC)
Sirexo - you might also want to check out meta:WikiObject proposed by Qupro. It has a list of useful wikidata properties for specific products - however we have so far not had a lot of detailed product information added to wikidata in general. ArthurPSmith (talk) 21:01, 20 June 2017 (UTC)
As ArthurPSmith said accepting a property takes at least a week.
When Wikipedia writes a template it's only important to keep the specific items for which the template is created in mind. On the other hand when we create properties for Wikidata it's important to also keep in mind how the property is used in other items as well. It's also important to find names that make it unlikely that the property will get misused.
You could start by modeling one example in . ChristianKl (talk) 11:17, 21 June 2017 (UTC)

A visual way to query Wikidata[edit]

Hey everyone, i've made a tool that allows you to query Wikidata in a visual waywithout using SPARQL. It's called VizQuery.

The possibilities of using Wikidata to do interesting queries are endless, and the current query service allows for very powerful queries indeed. However, i feel that for the general public, especially those who are not that technical, it might be a bit overwhelming and difficult for them to learn a complex language such as SPARQL. To make people familiar with the concept of queries i believe a somewhat less intimidating approach might be useful, hence this tool.

VizQuery is only capable of doing a subset of possible queries. It's basically simple triples, variables (prefixed with '?') and literals (between "quotes"). You can do pretty powerful queries with only those things though. For example, here's a query with vegetarians who are married to a vegetarian.

Under the hood VizQuery uses Ruben Verborgh's SPARQL.js library to convert between JSON and SPARQL, so theoretically every SPARQL query you could do in the regular query service can be done in VizQuery. However, many queries won't work because the visual interface only supports a subset of options: it's pretty hard to create user-friendly GUI representations of many of the complex SPARQL features. :)

Anyway, i'd like to hear what you think. Bugs, feature request and pull requests are also welcome on my Github page.

Husky (talk) 23:47, 20 June 2017 (UTC)

Thanks, Husky! Pretty cool! It's useful for showing people the most basic stuff of queries in Wikidata. Strakhov (talk) 14:38, 21 June 2017 (UTC)
This is really great! I've bookmarked it. - PKM (talk) 18:53, 21 June 2017 (UTC)
Nice tool! Consider to add support translation of of interface and make it work in more browsers (at Internet Explorer (Q1575) or Mozilla Firefox (Q698) does not work for me).

An Idea - User statistics based on wikidata[edit]

I'm thinking about a possibility to use the query service for statistical issues related to Users of Wikimedia & partner projects. For example, it would be interesting to view the activity of users based on their projects, activity type (edits), locations, etc. I know some statistical tools exist already but with limited abilities. I think that involving wikidata in this task could bring encouraging results to present to the different wiki communities. My question is: has this idea been discussed before? and if yes, what is the actual state of this issue? If no, I'd like to listen to your opinions. Greeting! --Sky xe (talk) 01:36, 21 June 2017 (UTC)

Wikimedia cares strongly about privacy. If you propose a tool to analyze people's locations that's unlikely to get very far. Fleshing out a proposal that produces an added benefit, doesn't take too many technical resources and that doesn't violate privacy is non-trivial. ChristianKl (talk) 10:36, 21 June 2017 (UTC)
I guess Sky xe (talkcontribslogs) means locations of the subjects edited. Such things like coordinate location (P625) of articles/items edited by a specific user. A user like Salgo60 who edits a lot about people buried on a burial place in Stockholm will then be very sharply located. I, who in my early years here at WMF-wikis, edited a lot in the Asteroid belt and in the realms of the (ex)planet Pluto will have a very interesting pattern. Face-smile.svg
-- Innocent bystander (talk) 10:46, 21 June 2017 (UTC)
I agree with ChristianKl here. There are already plenty of tools around which allow to stalk users to an uncomfortable extent. While those tool results base on a lot of publicly available information, the accumulation and evaluation of this information often goes too far to my opinion. For me this is in fact the main reason not to reveal my real life identity, and not to attend real life community gatherings such as the WikidataCon. —MisterSynergy (talk) 10:46, 21 June 2017 (UTC)
Agreeing with ChristianKI and MisterSynergy if this is related to contributor's identity, IP's, edition patterns etc. And still waiting for the arrival of structured data on Commons. Let's see how that stuff handles dates, places, coordinates and let stalkers picture a "chronomap" of a contributor with a handy app. Strakhov (talk) 14:27, 21 June 2017 (UTC)

Thank you all for the answers. I agree that privacy indeed is a critical point. But imagine we can view and evaluate statistics -by non violating the privacy of all user identities- such as:

  • Which wiki project has been most active in the last period (week, month, year) relative to the number of active users and which languages are basically involved.
  • Which fields have been edited most, e.g. in the English Wikipedia.
  • What is relation between users and their activity on the different wiki projects (to bypass the privacy challenge, e.g. identification only by countries instead of geographic coordinates based on IP protocols)
  • etc.. a lot more could be possible.

I think that the idea deserves to be discussed and I'm sure, more aspects will be enlighten thereafter.. Greetings! --Sky xe (talk) 16:47, 21 June 2017 (UTC)

Given current practice, I think the country in which a user resides is seen as private. Once you allow queries, they can also used for things besides your examples. ChristianKl (talk) 10:15, 22 June 2017 (UTC)

Wiktionary sitelinks enabled in Wikidata[edit]

Hello all,

As mentioned previously, we are now able to store the interwiki links of all the Wiktionaries namespaces (except main, citations, user and talk) in Wikidata.

Important: even if it is technically possible, you should not link Wiktionary main namespace pages from Wikidata. The interwiki links for them are already provided by Cognate, and in the future, Wikidata will also have special entity types for lexemes (see Wikidata:Wiktionary and mw:Extension:WikibaseLexeme/Data Model).

How can you help?

  • First of all, you can help us translating this documentation page in the languages you know.
  • If you know tools, scripts, bots, that could be useful for the migration process and removing the manual sitelinks, please share your informations on the page and offer help to people who would need to use them.
  • You may want to pay special attention to the new created items and all recent changes that will result from this new feature available for Wiktionaries.
  • Be friendly and welcoming with the Wiktionary editors :) Help them if necessary, make them feel part of the great Wikidata community.

Thank you very much! Lea Lacroix (WMDE) (talk) 08:36, 21 June 2017 (UTC)

What's the motivation to enable the future before the lexeme data type exists? ChristianKl (talk) 10:25, 21 June 2017 (UTC)
@ChristianKl: This is for sitelinks for everything that's not in the main namespace. So categories, templates, modules and project pages basically. Jon Harald Søby (talk) 10:43, 21 June 2017 (UTC)
Are you aware of any bot that could transfer automatically all the interwiki links to wikidata? I guess that bots which did the job for other projects can be adapted for this task. Pamputt (talk) 00:59, 22 June 2017 (UTC)
@Jon Harald Søby: But it's still really possible to add Wiktionary main ns pages, phab:T158323 said that an AbuseFilter should prevent such edits but where's that AF? --Liuxinyu970226 (talk) 02:24, 22 June 2017 (UTC)
@Liuxinyu970226: seems that such an abuse filter wasn't created yet. You can ask at WD:AN and someone with needed skills and rights will do it, probably. XXN, 11:53, 26 June 2017 (UTC)

Item/property descriptions on mouseover?[edit]

I noticed today that the watchlist has a feature where hovering over an item or property link gives the label and description in the user's default language - so, for example, Cédric Villani (Q334065) has a little label saying "Cédric Villani | French mathematician". RecentChanges and Contributions both have the same feature.

Elsewhere on Wikidata (such as here, or on an item page), hovering over an item/property link just gives the Q-number or P-number. It feels like it would be useful to turn the detailed version on everywhere, particularly when something links to an ambiguously titled item. Any thoughts? If there's interest I'll file a bug request. Andrew Gray (talk) 15:58, 21 June 2017 (UTC)

I would prefer disambiguation without hovers, but (in curly braces).
description, if absent then P31, if absent then P279 d1g (talk) 16:25, 21 June 2017 (UTC)
I don't think your description is accurate. Sometimes on the item page I see the description displayed. I think there's a script that tries to load the description the moment you however over it. If the loading takes to long the ID get's displayed. At least that's the best hypothesis for why it's behavior. ChristianKl (talk) 20:02, 21 June 2017 (UTC)
Interesting - I've never spotted this on an item page. Andrew Gray (talk) 12:20, 23 June 2017 (UTC)

Best practice etiquette related to choosing image (P18) property[edit]

I am working on importing some of the data stored in Commons Creator templates to Wikidata, and I am looking now at images used in creator infoboxes. I imported all the images that were missing on Wikidata, but now I am looking at cases where both Creator template and Wikidata item has an image and which are different. You can see Creator templates in question in here and the images at c:User:Jarekt/f. I think I have a good idea of what is superior image to depict a person, but would like to double check if others feel the same way. My current thinking is to replace Wikidata image if:

  • new image depicts person and current image depicts work created by the person
  • new image is a better quality version of the old image
  • new image shows a single person and old image multiple people
  • favor images which show face.

Any other suggestions, to help pick "better" image (P18)?

  • Shall we favor "portrait" vs. "landscape" aspect ratio?
  • In case of painters and photographers I prefer realistic self-portraits to depict a person. How others feel about it?
  • Is it a good idea to have multiple good images stored in P18? I think not, but just checking.

--Jarekt (talk) 16:56, 21 June 2017 (UTC)

I completely support your first four points. Additionally, I think a color portrait is better than an engraving of a portrait in most cases. And if we have a mediocre "head shot" and a good full-length portrait, we should consider using a head-and-shoulders crop from the full-length portrait. I also prefer "portrait" rather than "landscape" orientation. On self-portraits, I generally agree; but I'd make an exception for Dante Gabriel Rossetti and use the Watts portrait rather than the self-portrait of DGR as a very young man. - PKM (talk) 01:29, 22 June 2017 (UTC)
I agree with all four suggestions. Also portrait ratio should be preferred, because of wide use image (P18) in infoboxes. No preference about self-portraits. Single image is preferred.
Watch out for pictures not representing the person like images of signature or grave (without picture of bust of person).--Jklamo (talk) 11:25, 22 June 2017 (UTC)

Thank you both for confirming my rules of thumb. I am replacing hundreds of images with (hopefully) better ones. I thought I will share a little shortcut I find very useful which I use in c:User:Jarekt/f. That page was created by a excel spreadsheet pairing up images from commons with images from wikidata for the same person. A little (+) can be used to perform 2-click replacement of the wikidata image with Creator one (I am not suggesting the actual replacement in case of those 2 images, so only do one click). That technique can be probably used for other mass tasks with human in the loop. --Jarekt (talk) 15:33, 22 June 2017 (UTC)

For Classical Greek and Roman authors, a contemporary bust (Q241045) (sculpture) is usually the best option over a mosaic, painting, etching, or later form of depiction. The best photographs of busts have the face pointed towards the viewer, and the best choice of bust will have the nose intact. When a bust is not available, a depiction on a coin may be useful. Care has to be exercised though, because some sculptures have been incorrectly identified, and later found to be so. The current data accompanying some of these sculptures is not always present yet in Commons or Wikidata. --EncycloPetey (talk) 21:02, 22 June 2017 (UTC)

Agree, I will keep that in mind when picking images. --Jarekt (talk) 12:11, 23 June 2017 (UTC)

To start with, do we all agree that there should be only one file in P18 for every item? I can easily see that if not, Wikidata can become a trash can similar to Commons; on the other hand, I see added value in say having an exterion and an interior of a building. Has this actually ever been discussed?--Ymblanter (talk) 08:49, 23 June 2017 (UTC)

Well, I would prefer to see both a male and a female file in lion (Q140) like it already has.
Replacing an image which already has a caption in a language you cannot interpret and change could be very troublesome.
A related topic: I would like to allow such things as depicts (P180) as a qualifier to P18. Often we have pictures of churches to illustrate a village. But how to tell which church? The discussions this far has denied such usage, but I think it could be very useful since you cannot add links into the caption-property. And what a picture illustrates depends on in which item/article you use the file in. File:Carlito syrichta on the shoulder of a human.jpg has for example been used to illustrate everything from Lsjbot to Primates and everything in between. -- Innocent bystander (talk) 11:49, 23 June 2017 (UTC)
I agree that male and female images (if different) for living organisms is a great idea. I guess that is an example of a case where multiple images stored in image (P18) are a good idea. In context of Commons Creator templates images will be mostly of people, and for such I find multiple images troublesome. I did not pay much attention to captions, as so few images have them I did not notice you could add them. To me a new better image outweighs the loss of a caption, but it should be considered in case of minor improvements. One thing I do not like in case of P18 images of people is to see images depicting their artworks instead of depicting the artists. When the artwork shows someone else that is hard to spot. --Jarekt (talk) 12:11, 23 June 2017 (UTC)
Agree, if it isn't a good self portrait, an art made of the subject is not a good file at all in P18. I guess they have been imported from WP-articles and things like that. No big problem in WP if the caption tells it isn't the subject, but such things goes unnoticed by these rough tools. . -- Innocent bystander (talk) 12:22, 23 June 2017 (UTC)
I like your rules of thumb and I agree that "image of the artist's work" should never be in P18. However, would it be worth creating a new property for "image of an example of this person's work", to go alongside images of plaques, graves, signatures, etc? This would give us soemwhere to put these, which might be useful for the cases when we have no other appropriate image but still want to have some kind of illustration (eg for an infobox). Andrew Gray (talk) 12:19, 23 June 2017 (UTC)
Well, don't we already have a property for "Notable work"? I guess P18 could be used as qualifier to those. Or even better, be put inside those "work-items". -- Innocent bystander 12:25, 23 June 2017 (UTC)~
Agree, Works by the creators should be added to notable work (P800). In case of paintings user:Multichill and other Wikidata:WikiProject sum of all paintings participants are adding a ton of items for those. I do not know if we have other projects for sculptures, drawings, books, photographs, etc., but we should. --Jarekt (talk) 13:06, 23 June 2017 (UTC)
  • {comment}} @Jarekt: I don't see that we are exactly limited to a single image, though where adding more than one it is important that one is marked as preferred.  — billinghurst sDrewth 13:00, 24 June 2017 (UTC)

Identifier for FamilySearch[edit]

I may have asked this before, but I can't find the results of the discussion. Should we have a Identifier for FamilySearch even if there is no landing page for the link unless you are registered for free. For biographies it would provide the data and actual image of the record for people in the census, and other primary records for confirming birth, marriage, and death. Here is the page for Elizabeth Coleman White at To see it you would need to register, but the site is a cornucopia of biographical information based on primary documents. We link to, but that does not have the original documents for free. --Richard Arthur Norton (1958- ) (talk) 23:41, 21 June 2017 (UTC)

Properties are created after a property proposal in Wikidata:Property proposal. Feel free to create a proporty proposal for FamilySearch. ChristianKl (talk) 09:01, 22 June 2017 (UTC)
We already have FamilySearch person ID (P2889). —MisterSynergy (talk) 09:05, 22 June 2017 (UTC)
Thanks. --Richard Arthur Norton (1958- ) (talk) 17:52, 22 June 2017 (UTC)

Nutritional Information[edit]

We currently don't have any nutritional information in Wikidata. It seems USDA Food Composition Databases contains a great database by the US government. US government data is in the public domain, so we could import it. Do we currently have the necessary properties? ChristianKl (talk) 12:59, 22 June 2017 (UTC)

you might want to check out properties that are instances of Wikidata property related to food and eating (Q23038310) - there don't seem to be many yet though. ArthurPSmith (talk) 17:54, 22 June 2017 (UTC)
Pierre from Wikidata & Open Food Facts, reporting for duty :-)
- We recently imported all the USDA db into Open Food Facts. We're looking for contributors who can scan their food, and add pictures to augment the data. Millions of photographic evidence to take. There are also some generic profiles (without barcodes) in that database, which we haven't imported into OFF, but that I've put up for matching on Mix N'Match. More info at
- Also, feel free to help Open Food Facts: we need contributors, translators, coders...
>> Web:
>> Android: (also on FDroid)/ iOS:
>> Translators needed @
--Teolemon (talk) 15:32, 23 June 2017 (UTC)
Unfortunately, Open Food Facts seems to have a license that's not compatible with Wikidata. I think having data under CC-0 so that everybody can use it is a more worthy project. ChristianKl (talk) 17:42, 23 June 2017 (UTC)
The USDA db is public domain, as all works by US Civil Servants, although derived (for the part with barcodes) from producer data. So you can import it into Wikidata. For Open Food Facts, I do get your point: We have chosen the ODbL to ensure the Db keeps growing, and that commercial apps play the game of sending pictures back. --Teolemon (talk) 18:45, 23 June 2017 (UTC)
Yes, importing from USDA would work for Wikidata.
I think you are wrong about the effects of licensing. Any company that uses a data set has an interest in that data being high quality. We likely wouldn't be able to exchange data with Songkick if we would be ODL-licensed.
In many cases, I would estimate that a serious company will build their own database instead of using OpenFoodFacts. The will use BSD/Apache-licensed libraries but they won't use GPL licensed ones. ChristianKl (talk) 21:20, 23 June 2017 (UTC)

Link to SQID tool in the left hand bar[edit]

I really like the gadget that displays the Resonator link in the left hand bar. I think it would be good to have the same for the tool at as well. Is there a reason against having such a gadget? ChristianKl (talk) 18:08, 22 June 2017 (UTC)

I think that SQID (Q24298088) should be default view, when non-editors view site (to explore data and "like this" buttons)
What we have now should be "edit" mode (with many edit buttons) d1g (talk) 07:18, 24 June 2017 (UTC)

Good and bad news about cross-wiki search results in Wikipedia[edit]

Good news: The cross-wiki search results from other projects are now live in Wikipedia. Bad news: The search results from Wikidata are not part of the plan, meaning users won't see those results in any Wikipedia site. Feel free to share your thoughts here. --George Ho (talk) 20:36, 22 June 2017 (UTC)

George Ho does the "Discovery Team" know that wikidata exists? But I think this is somewhat moot - or at least up to the language wikipedias how to deal with. In, for example, svwiki if your search turns up few or no results it also does a wikidata search - for example [1]. ArthurPSmith (talk) 13:18, 23 June 2017 (UTC)..
They have enabled a gadget on the Swedish Wikipedia to show results from Wikidata. And seems the Discovery team is aware of Wikidata per this. Stryn (talk) 14:55, 23 June 2017 (UTC)

date of death of Honoré de Balzac[edit]

I was just looking at date of death (P570) of Honoré de Balzac and we seem to have two values 18 August that all Wikipedia articles state and 19 August "stated in" "Integrated Authority File" 3 years ago. I assume "Integrated Authority File" means this link, but I could not find any mention of date of death there. Is it OK to delete that date if the source no longer states it? If there is a controversy over his date of death, I then we need to keep both dates but otherwise this looks like typo that someone corrected long time ago. What is proper procedure of removing the data? --Jarekt (talk) 12:50, 23 June 2017 (UTC)

I think it actually references this link (see last triple) --Nono314 (talk) 17:08, 23 June 2017 (UTC)
Ok so that date is still returned by I wonder if it is a typo or there is a real controversy about date of his death. --Jarekt (talk) 18:37, 23 June 2017 (UTC)
He passed during the night. According to frwikisource, it would have occurred between 22:30 and 22:40 on 18th. --Nono314 (talk) 19:56, 23 June 2017 (UTC)

Members of the European Parliament[edit]

I just saw this query on Twitter:

Does anyone have a spreadsheet of MEPs sorted by committee? [...] everything on EP site is so unhelpful.

How far are we from being able to answer that, and who is working on MEP data? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 21:00, 23 June 2017 (UTC)

I haven't looked into European Parliament information in Wikidata in any great depth yet, but my impression is that we're quite far away from being able to answer something like this. Most of the position held (P39) entries that I've seen for member of the European Parliament (Q27169) are completely bare — i.e. not even qualifiers that will tell us when this was true, or which constituencies they were elected from — so we have no way to even produce a list of who we think the current MEPs are to see how complete it looks. And committee memberships is something that we don't really have in much depth for any legislature yet (and not at all for the vast majority of them). My impression is that some people have done work tidying up data on the MEPs from individual countries, but no-one is actively working on the European Parliament as a whole. --Oravrattas (talk) 06:30, 24 June 2017 (UTC)
We even have a few hundred with no country, as well. There was a push to get them all in based on matching to a database produced by the EP - so we do have them all, give or take some errors or duplication, but beyond that the data is pretty light. Andrew Gray (talk) 10:32, 24 June 2017 (UTC)

3600 via P1186 / MEP directory ID. Links at Wikidata:EveryPolitician need updating. --Atlasowa (talk) 13:09, 26 June 2017 (UTC)

Property proposal - template transclusion failure?[edit]

I created my first property proposal here and attempted to transclude it into Wikidata:Property_proposal/Creative_work with this edit. However, the property proposal page is still showing the following message near the top of the page:

You have not transcluded your proposal on Wikidata:Property proposal/Creative work yet. Please do it.

How can I fix this? Zazpot (talk) 10:17, 24 June 2017 (UTC)

I such cases, it is always worth a try to make a null-edit to the page. (Press the edit button, and save again without changing anything.) Whatever caused the problem, looks like it is solved now. -- Innocent bystander (talk) 10:21, 24 June 2017 (UTC)
I think that this discussion is resolved and can be archived. If you disagree, don't hesitate to replace this template with your comment. Matěj Suchánek (talk) 19:09, 26 June 2017 (UTC)

Annoyed with the amount of merges that can be done[edit]

I really hate it when people request an item to be deleted when all they could do is merge it. Waste of bytes for system, waste of my time, and just a waste of time for the person submitting because they have to find the duplicate and write that in reason. PokestarFan • Drink some tea and talk with me • Stalk my edits 12:59, 24 June 2017 (UTC)

Bot reversion of vandalism[edit]

On the English Wikipedia and other large wikis, there are a multitude of bots such as ClueBot NG which revert obvious vandalism (e.g. nonsense, rude words, page blanking) quickly after its insertion into articles. Is there a bot currently doing this on Wikidata; and if not, is it possible for this to be done (for statements, as well as labels and descriptions) given the obvious differences in data format? Jc86035 (talk) 14:37, 24 June 2017 (UTC)

That's a very interesting topic! I guess a machine learning approach could go a long way. There are many interesting features to take into account the constraint violations (not just format), the editor's experience, the tags, the edit type, size… − Pintoch (talk) 18:18, 24 June 2017 (UTC)
ORES already does the machine learning work and is supposed to have an interface that communicates the likelihood that an edit is vandalism. ChristianKl (talk) 09:04, 25 June 2017 (UTC)
Oh right, I forgot about this one. Is there a bot that uses ORES to actually revert things? − Pintoch (talk) 09:11, 25 June 2017 (UTC)
At the moment there isn't a bot, but if I understand the ORES mission the right way, they see it as their role to provide the necessary information for a potential bot to do a task like this. ChristianKl (talk) 10:03, 25 June 2017 (UTC)
@Ladsgroup: You might want to chime in. --Lydia Pintscher (WMDE) (talk) 09:28, 26 June 2017 (UTC)
I've found ORES marking a large number of my recent merge edits in red - I think it needs to be adjusted for how wikidata works, because it is badly wrong about vandalism in a lot of cases here. ArthurPSmith (talk) 14:21, 26 June 2017 (UTC)

The trouble with vandalism has always been that overt vandalism is easy but covert vandalism is hard to detect. I suspect that there is less overt vandalism on Wikidata. That's not to say I am not interested in the problem. Face-smile.svg All the best: Rich Farmbrough12:31, 26 June 2017 (UTC).

Yes, that's a correct assessment. I will check what threshold can be used to run such a bot. Amir (talk) 12:33, 26 June 2017 (UTC)

Mapping languages on Wikidata[edit]

Hi all

I'm trying to map all the languages from the UNESCO Atlas of World Languages in Danger on Wikidata before 6th July for Celtic Knot: Wikipedia Language conference. There are around 400 items left, all help would be appreciated.

There's also a matching taking place for Glottolog which is a more comprehensive database but it has over 15,000 items still to match, if anyone knows any clever ways to match them that would be amazing.

Thanks very much

--John Cummings (talk) 15:14, 24 June 2017 (UTC)

@John Cummings: You might want to check out OpenRefine - latest release has wikidata matching allowing you to correlate metadata with specific wikidata properties. I'm not sure how workable that would be for languages but it might be worth trying. Once you have a list of matches you can export a CSV file from OpenRefine that can be tweaked to run in QuickStatements. ArthurPSmith (talk) 14:25, 26 June 2017 (UTC)
Many thanks ArthurPSmith (talkcontribslogs), this is quite a lot beyond my technical ability but is anyone else knows how to do it I'm very happy to do any grunt work needed. Also you may be interested in this --John Cummings (talk) 14:35, 26 June 2017 (UTC)

Use of Q7066 as a religion[edit]

Is religion (P140)atheism (Q7066) correct? Atheism isn't really a religion, so it might be a misapplication of the property. There was an enwiki RFC a while ago which resulted in the consensus that "religion: atheism" and "religion: agnosticism" shouldn't be included in infoboxes. (The consensus was also that "religion: none" shouldn't be included, but we can still arbitrarily use no value for everything so I'm unsure of the implications of that here.) Jc86035 (talk) 15:45, 24 June 2017 (UTC)

The general principle of Wikidata is plurality. If sources say that a person's religion is "atheism" we can report this in Wikidata. ChristianKl (talk) 16:26, 24 June 2017 (UTC)
We have lifestyle (P1576) for the other situations though. There is still some discussion happening on the talk page there though. Sjoerd de Bruin (talk) 16:29, 24 June 2017 (UTC)
This is fine when religion is a subclass of lifestyle. Thanks, GerardM (talk) 17:17, 24 June 2017 (UTC)
No it is not correct; just as "not collecting stamps" is not a hobby. You may though use "religion -> no value", and give a source if you have one. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 16:40, 24 June 2017 (UTC)
Few people make a point to publically identify as a person who's "not collecting stamps" but a decent amount of people does identify as "atheist". ChristianKl (talk) 19:39, 24 June 2017 (UTC)
@ChristianKl: But atheism still isn't a religion ("sacred belief system"), so it's strictly incorrect to say with the current definition of the property that someone's religion is atheism. Jc86035 (talk) 04:06, 25 June 2017 (UTC)
@Jc86035: You make more or less the same point as The Bible: "A god who cannot walk, cannot hear, cannot see and cannot talk is not a god." If I make that my postulate I could remove such statements from Zeus (Q34201) too. No! Let our sources guide us here, not our private opinions. -- Innocent bystander (talk) 06:51, 25 June 2017 (UTC)
@Jc86035: If you use the term "sacred" in the way it's used in modern decision science, I think there are plenty of beliefs in the cluster that the average atheists holds that are sacred for them. ChristianKl (talk) 08:39, 25 June 2017 (UTC)
@ChristianKl: Regardless of what the definition of religion (Q9174) is, atheism (Q7066) is not an instance of (P31) religion (Q9174), so using religion (P140)atheism (Q7066) is incorrect and would be better as religion (P140)no value and [some other property] → atheism (Q7066). Alternately items which are instance of (P31) world view (Q49447) could be accepted as values for religion (P140) and the definition of the property could be changed. Jc86035 (talk) 08:48, 25 June 2017 (UTC)
My opinion is that novalue cannot be used to imply Atheism. That better fit those who tell they have no opinions about religion. -- Innocent bystander (talk) 14:05, 25 June 2017 (UTC)
The indication of religion identifies association more than a firm belief. The level of religiosity is not determined in any way. Given that it is about association, it follows that atheism can be seen in this light. They consider that there is no obvious choice to make between any and all of the religions we can believe in. Thanks, GerardM (talk) 15:39, 25 June 2017 (UTC)

@Jc86035: English label could be changed, Russian states "attitude towards religion" I also added an exception d1g (talk) 09:44, 25 June 2017 (UTC)

  • Statements using religion (P140)atheism (Q7066) should be removed, because atheism is not a subclass of religion (as not collecting st[...]). Here it's even simpler than in en.wikipedia. We have constraints! We are a semantic database! Feel free to enact a new property called "attitude towards religion" (including values such as "religious", "atheist", "agnostic",...) (Being honest, I don't see Shia Islam as an "attitude towards religion"). Strakhov (talk) 17:28, 25 June 2017 (UTC)
1. Not before we have a separate property for world view (Q49447) or existence of God (Q1699404)
2. religion (P140)atheism (Q7066) claims break nothing, because other claims in religion (P140) are impossible for atheists. d1g (talk) 17:52, 25 June 2017 (UTC)
It breaks the very same thing as storing in child (P40) a hypothetic "no label (Q1200000000000)" labeled in English "no child". It breaks logic. We have around 4000 properties and growing, what's the problem in creating an extra one for this stuff? We can call it "post-world view", "view of very deep things beyond human comprehension", "attitude towards religion", "position towards the idea of God". It doesn't matter at all. If not, put it in a generic "lifestyle (P1576)", but not in religion because is not a religion. Strakhov (talk) 18:25, 25 June 2017 (UTC)
"life stance" was used as English alias, "отношение к религии" is used as main Label in Russian
We use abstract items for some properties anonymous (Q4233718) it isn't something clearly harmful.
Additional properties like "no author" complicate things similar how anonymous (Q4233718) complicates one property. d1g (talk) 18:38, 25 June 2017 (UTC)
Yeah, we use abstract items, it's not clearly harmful... but, I insist, what's the problem in storing different things as "religions" and "positions towards religions" in different properties? Where's the need, the reason after that mix here? We are not paper. After that, different projects will be able to display that data whatever way they want, as a religion or as a view of something or ...not showing it at all. Strakhov (talk) 19:00, 25 June 2017 (UTC)
See number 1, theism (Q620805) aren't different from agnosticism (Q288928) or atheism (Q7066), they represent existence of God (Q1699404) or world view (Q49447)
d1g (talk) 19:13, 25 June 2017 (UTC)
I see 1. And? Of course atheism and agnosticism are not the only isms that shouldn't be there. theism (Q620805) should not stored as a religion either. So what? Create a property for that and call it existence of God (Q1699404) or world view (Q49447) or "position" with qualifier "towards" -> "divine existence". Or rename "religion", call it "world view (Q49447)" and get the scope wider. But there's no reason for not doing things right. Strakhov (talk) 19:27, 25 June 2017 (UTC)
I added instance of (P31) religion to atheism, with an authoritative court case as source. ChristianKl (talk) 16:17, 26 June 2017 (UTC)
@ChristianKl: Then you should change this statement and create another class « ideology considered a religion in the court juriction » because a court is authoritative in law, but definitely not in theological philosophy. This jurisdiction is probably not the whole world as well and uses its own arbitrary notion of a religion. Let’s make clear which one the definition is. author  TomT0m / talk page 19:19, 26 June 2017 (UTC)
@TomT0m: Wikidata is designed to allow for different parties having different opinions. statement disputed by (P1310)-qualifiers can be used to indicate that different sources have different opinions. The opinion of the Seventh Circuit of the United States Court of Appeals is certainly not a more arbitrary notion than that of individual Wikidatians. ChristianKl (talk) 19:52, 26 June 2017 (UTC)
@ChristianKl: You misses the point entirely. It’s all about the definition of a religion in the USA. And this is where the problem should be solved : at the definition (hence item or property) level. 1 definition = 1 item. Otherwise you can add the dispute qualifier on every claim in which the item is used in that sense. This is useless data duplication. Worse, this is data loss because we for sure know this is the US notion of a religion that is used and not anyone else. This is not made explicit and you choose to use an ambiguous item instead. This is bad practice. For example, I think a correct use of disputed by would be another US court that had another opinion. Think that, as is, the statement and all the like could be disputed by virtually anyone … this is wayyyy too open. author  TomT0m / talk page 06:33, 27 June 2017 (UTC)
I found this source by looking in Google Scholar for answers to the question "Is Atheism a religion?". I consider that a good general heuristic for answering questions. Wikidata doesn't have any "definition" property. We don't have a Israel according to the Israeli and a Israel according to the Palestianians item even through both define it differently (and with different borders). Since the start, Wikidata has used "disputed by" for cases like that. ChristianKl (talk) 08:32, 27 June 2017 (UTC)
@ChristianKl: Uncompelling at all. They consciously chose a definition : see for example. They made a choice, they had to. But … this is a choice and has no pretention to universality. Which Wikidata has. By NPOV. The best way so do this is to avoid as much as possible item ambiguity. Hence relying on definitions instead of labels. Wikidata can have as many items with the same label as we like. There is other definition of religion. Certain philosopher even consider (see cites Daniel Dubuisson (Q3013975) View with Reasonator View with SQID thinks the concept of a religion is unusable because too vague or ill defined to actually be used … Which emphasize the fact we should just use more precise concepts. Lets not enter the troll, be specific and objective. What is considered a religion by the USA is objective. «The» definition of «religion» is not. It’s a trap. On the other hand, stuff like « monotheistic religion» are easier and more precise. Better use them. author  TomT0m / talk page 19:19, 27 June 2017 (UTC)
Wikidata NPOV means to report what different sources are saying and use "disputed by" when warrented. It's not a principle that's about excluding viewpoints. It might very well be that there's some sociologist who researched what's called a religion in the US or for that matter in other countries. Currently, I don't see that you linked or anybody else linked to any such sources. Wikidata does have the more precise concept of monotheistic religion (Q19842652) and I do think that it makes sense that abrahamic religion (Q47280) subclasses it instead of subclassing religion (Q9174) directly.
The problem with definitions is that they are in a specific language. In Wikidata we don't want to depend on a specific language and thus have no definition property. If you think we should have definitions in Wikidata, feel free to propose a property for definitions.
If I look at the larousse definition Google Translate translates it into "A determined set of beliefs and dogmas defining the relationship of man with the sacred." I do think that "I don't believe in God" is a statement that verbalizes my relationship with the sacred.
For many use-case the concept of religion might be unusable because it's to vague. For the use-case that the court had, it doesn't seem to vague. We have use-cases for which it can be useful. A person might ask themselves about the religion of the politicians that sit in congress and how that shifts over time. Having that information in Wikidata is useful. For that use-case, it's useful to have a broad concept of religion that includes atheism. ChristianKl (talk) 19:57, 27 June 2017 (UTC)
Is agnosticism (Q288928) a religion and should be modeled via religion (P140)? --Succu (talk) 20:13, 27 June 2017 (UTC)

Data donation: Giphy[edit]

I'm pleased to report that I have secured another data donation: over 4.7K of Giphy user account IDs, for brands and artists, to be matched to Giphy username (P4013). These are already in Mix'n'Match, where over 20% are already matched and uploaded to Wikidata. The rest await attention! Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 16:38, 24 June 2017 (UTC)

Bulk create items without statements, but only a single sitelink ?[edit]

In the past, occasionally, bots create thousands of items merely with sitelinks, but no further statements.

The situation has improved. Items without any statements for enwiki are at 4.6% (see Wikidata:Database_reports/without_claims_by_site). Still, we have items created years ago without any statements.

The question is if we should continue to create items without any statements (and wait till some day someone adds statements to these). Alternatively, we could decide that bulk created items should include at least one statement, if if a sitelink is present.
--- Jura 22:01, 25 June 2017 (UTC)

For Wikipedia, the answer may be "yes". For Wikisource, the answer is a resounding "no". Wikisource items without any statements are valueless, and many, many, many items at Wikisource sites are chapters of books, or acts within plays, and really should not have data items. --EncycloPetey (talk) 02:50, 26 June 2017 (UTC)
Clearly the tool you mentioned, Wikidata:Database reports/without claims by site, does not work for unlinked pages. In fact, only few tools (e.g. Duplicity) work for unlinked pages, but it's a huge backlog (multiplied by hundreds of wikis), and it's not practicial unless someone periodically check them (like nlwiki). Also, another way to kill unlinked items is QuickStatements, which required both the subject and the object are connected.--GZWDer (talk) 04:25, 26 June 2017 (UTC)
@Jura1: I think that every item in Wikidata without any statement is useless (even when it has multiple sitelinks, but the more sitelinks, the more easy it is to add any statement), and therefore can be deleted. Maybe a way to decrease the number of un-statemented items is to make a list with 50 items on it and give it a week to be improved. If the week has expired, and there are still items with zero statements they can be deleted without processing through RfD. If there are others with any other idea to clean the list of empty items, feel free to add it!
@EncycloPetey:, why should chapters or acts not have any statements? They can have part of (P361), instance of (P31) chapter (Q1980247)/act (Q421744) and with series ordinal (P1545). I don't see a problem why they shouldn't have any statements. Q.Zanden questions? 10:22, 26 June 2017 (UTC)
"I think that every item in Wikidata without any statement is useless (even when it has multiple sitelinks...), and therefore can be deleted" So you'd be happy to break the interwiki linking between two Wikipedias, or other sister projects? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 10:33, 26 June 2017 (UTC)
@Pigsonthewing: No, not at all, I see how much Wikidata can do to connect all the different languages. But if there are no statements, I think that the items are quite useless. And as I also said above, the more sitelinks an item has, the easier it should be to create some statements about it because there are multiple articles about that topic. So I think we should put more work in adding statements to (almost) empty items. Q.Zanden questions? 10:42, 26 June 2017 (UTC)
@QZanden: Is your position that you see what Wikidata can do to connect different languages but you still want to remove items that provide sitelinks between different articles in different languages?
Given that descriptions of items can be displayed in the Android App, items with a single sitelink and no statements can provide benefits to users. ChristianKl (talk) 11:06, 26 June 2017 (UTC)
@Q.Zanden: Your two statements are mutually exclusive. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 11:52, 26 June 2017 (UTC)
The number of items without any sitelinks are growing steadily. It is impossible to guess what articles in any and all of the 285+ Wikipedias. When there are items, they are found when you search for them. We should more aggressively add items for articles that are not linked to an item. At that the English Wikipedia is not of a special relevance, the German Wikipedia for instance is in many fields more complete. Thanks, GerardM (talk) 11:20, 26 June 2017 (UTC)
@Q.Zanden: Because for any given chapter of a book, there must be (a) a general data item for the chapter as a work, and then (b,c,...) separate data items for each edition of that chapter. So, if we have a 1910 UK edition of The Time Machine, with 12 chapters, then we need to create 12 data items for the 12 chapters as works, then another 12 data items for the actual chapters of the 1910 UK edition. The 12 work chapters will be part of the book as a work, but the 12 edition data items for chapters will be part of the specific edition, and also editions of the work chapters. This runs into further problems when you realize that there is more than one way that The Time Machine is divided into chapters, namely editions with 12, 14, or 16 chapters, which are not analogous to each other. So we need one data item for the novel as a whole, one data item for the serialized version as a work, one data item for the Heinemann text as a work, one for the Holt text as a work, one for the Atlantic text as a work, then each individual edition can be added. But then that whole discussion about chapters as works and editions? That gets multiplied by 4 because of the three novel and one serialized edition, and each set of chapter data items needs to clearly identify to which edition of the text it belongs. And that just deals with English editions. For translations into other languages, there's no guarantee that it will fit any of those editions. And because translations are separate data items, and editions are separate data items, there will never be any interwiki links to reduce the number of data items through merging. Create chapter data items? The creation of those for a single novel can run into the hundreds. And if a text is misidentified and all its item have to be located and edited? The work and maintenance are prohibitive beyond belief with no real benefit. What is the advantage to having a data item for a chapter 3 from the 1910 UK publication of the Heinemann text of H. G. Wells' The Time MAchine? It will have no interwiki links, because only the English Wikisource will have English copies of chapter 3 of the 1910, etc., and neither will it have external links to any data items because libraries don't do that sort of thing. --EncycloPetey (talk) 11:19, 26 June 2017 (UTC)
I think the situation for WikiSource is somewhat different and not my primary concern when I started this thread.
Numbers for enwiki are actually going down (333789 / 4.6%) yesterday compared to ( 411047 / 6.2%) a year ago. Still, I don't think the current level is necessary a good base level. There are benefits of having a single sitelink items, but I'm less convinced that it outweighs disadvantages ..
--- Jura 13:41, 26 June 2017 (UTC)
  • I strongly find even items with nothing but a label and a single sitelink to in themselves be useful in wikidata. Yes they require some human attention to become more useful, but in themselves they ensure that (if they are complete) everything that any wikipedia has written about has some record in wikidata, and can therefore be found and enhanced. If otherwise one has to search through every language wikipedia to see if there is a matching entity for the external identifier you are looking at, I would find that nearly hopeless. So I would strongly oppose any effort to stop adding these items. ArthurPSmith (talk) 14:36, 26 June 2017 (UTC)
  • I find these useful as well. As long as our policy is that every article in every Wikipedia should link to one and only one Wikdata item, we should create these as we go. - PKM (talk) 19:06, 26 June 2017 (UTC)
For the Dutch Wikipedia we have an approach:
  • Monitor new article creations and try to link the to an existing item or create a new item with some statements. This is all done by humans with tools like duplicity
  • When an article is at least 28 days old and hasn't been edited for 21 days a bot comes along and creates the missing item. In practice this rarely happens, but this prevents a backlog
  • People work on Wikidata:Database reports/without claims by site/nlwiki, this is just a couple of new items and mostly old. The oldest is currently from February 2013
User:NoclaimsBot also works on the Dutch Wikipedia, but rarely has any hits these days. Would probably be good to run the newitem bot for other language Wikipedia's too. This way the new items come in a steady flow instead of a huge pile of items every once in a while. Multichill (talk) 19:36, 26 June 2017 (UTC)
  • Pictogram voting comment.svg Comment @QZanden: Chapters in a novel don't have items as they are predominantly not individually notable. How many "Chapter 1" labels would you like to see through Wikidata? Where an individual component of a work is notable, eg. a biographical entry, a paper, a poem, then WSes does look to list them individually.  — billinghurst sDrewth 21:38, 26 June 2017 (UTC)
  • Maybe we can try to adopt some of the exemplary practices of nlwiki for other Wikipedias. Obviously, it works best if contributors from a given Wikipedia regularly check duplicity. On the Wikidata side, maybe for some categories, items could be created regularly by bot after a fairly short time (include a P31 value). For others, the bot could create them after a wait time (even without P31/P279).
    --- Jura 11:47, 27 June 2017 (UTC)

Convert Q to P[edit]

Hi. Capacity factor (Q1755148) should be a property. Could someone help convert this, or guide me on how to go about with it please? The term is user frequently in the energy industry. Thanks! Rehman 11:16, 26 June 2017 (UTC)

It's not a matter of converting. The existing item is fine. There's the option for creating a corresponding property but that wouldn't mean we would get rid of the existing item that's linked to interwiki links. If you want a new property, the process is to write a property proposal. ChristianKl (talk) 11:56, 26 June 2017 (UTC)

Languages on Wikidata, Glottolog catalogue matching[edit]

Hi all

Currently there are 15,000 or so unmatched items which are being done manually. Because Glottolog already has links to ISO ISO 639 codes I think it may be possible to a very large amount of automajic matching to existing items which will both speed up matching and reduce errors in matching.

The only issue I can see with match items is with dialects which may shame ISO numbers with the 'main language'. I think this could be fixed with running two queries afterwards for items that share the same ISO code and for items with multiple ISO codes.

I'm talking about the UNESCO Atlas of World Languages in Danger on Wikidata at the Wikipedia Languages Conference in early July and would be really amazing to show the Glottolog catalogue as well.

Is there any way to know who added the catalogue for matching? It would be very very good to have both data about the languages on Wikidata but also include the relationships between them e.g their language families.

Also is there any way to know if all the ISO 639 (parts 1, 2 and 3) have been added to Wikidata?


--John Cummings (talk) 13:44, 26 June 2017 (UTC)

A language code may comprise multiple parts. The first is the language code (preferably ISO-639-3). The easiest thing to do is make a list of all the different codes when the list is not too long I am happy to have a look and come up with some recommendations. Thanks, GerardM (talk) 15:21, 26 June 2017 (UTC)

Wikidata weekly summary #266[edit]

Dispute about webapp-related "said to be the same as" claims[edit]

I don't know if I'm in the right place, but I wanted to get a third opinion on a dispute involving three statements added this morning by User:D1gggg:

The first two showed up on my watchlist; they sounded wrong to me so I reverted. But the user re-reverted again without an explanation. So I started a discussion on their talk page (Topic:Tt6ir4v8yjzxodgf), but made no progress, it seems we're talking past each other. Can anyone chime in with a third opinion please? Intgr (talk) 15:23, 26 June 2017 (UTC)

Pictogram voting comment.svg Comment be aware that my questions were ignored 123 times... d1g (talk) 15:43, 26 June 2017 (UTC)

I don't see a reason for why they should be tagged with said to be the same as (P460). Not all software that runs on a server is a web framework or a web server. The German Wikipedia lists 17 different types of servers and web servers are only one of those 17. The server software that manages access to printers on a network is neither a web framework nor web server software. print server (Q1162303) is a subclass of server software (Q1371279) and it would also make sense to classify web application (Q189210) as subclass. ChristianKl (talk) 15:56, 26 June 2017 (UTC)
we can change these 3 claims or every claim, I don't see a reason to use these items.
"web framework" or "web server" add no value over server software (Q1371279)/client (Q528166).
I don't see classification for Rich Site Summary (Q45432) in German Wikipedia.
Print server are using print protocols Internet Printing Protocol (Q1667982).
depends on software (P1547) should be used for "frameworks". d1g (talk) 16:10, 26 June 2017 (UTC)
I agree these things are not the same, and here are a few examples for User:D1gggg:
So these concepts are related, but too loosely to be identified with this property. − Pintoch (talk) 16:30, 26 June 2017 (UTC)
I don't think web framework (Q1330336) is in meaning equivalent to web server (Q11288) and Django does qualify as web framework (Q1330336) but otherwise I agree with Pintoch. ChristianKl (talk) 16:47, 26 June 2017 (UTC)
@Pintoch: HTTP server, so definitely server software (Q1371279) - which I'm pointing
"web framework" - is a subjective game around server (Q44127) and server software (Q1371279) in non-{{Q|Q251}] world.
Servlets are the Java platform technology of choice for extending and enhancing Web servers. - then what it is if not a server software (Q1371279)? subclass of "web framework"? Synonym? d1g (talk) 17:19, 26 June 2017 (UTC)
I do not understand your sentence "HTTP server, so definitely server software (Q1371279) - which I'm pointing" (and the following ones are also unclear). Can you please write full sentences? − Pintoch (talk) 17:42, 26 June 2017 (UTC)
When somebody says "HTTP" it can only mean two things: HTTP client (Q528166) or HTTP server software (Q1371279)
  • client A program that establishes connections for the purpose of sending requests.
  • server An application program that accepts connections in order to service requests by sending back responses. Any given program may be capable of being both a client and a server; our use of these refers only to the role being performed by the program for a particular connection, rather than to the program's capabilities in general. Likewise, any server may act as an origin server, proxy, gateway, or tunnel, switching behavior based on the nature of each request.
When one says "web framework" it can mean anything depending on the perspective of the speaker and his grip in technology. There is no standard for "web framework". 17:54, 26 June 2017 (UTC)
Thanks a lot, this is much clearer! But I do not really understand why you keep bringing up the HTTP protocol. Not every web-related concept is described by the standard for the HTTP protocol. (As a random pick, I would take Gangnam Style (Q890): not sure whether it is more of a server or a client. But it is definitely a very important component of the web!) − Pintoch (talk) 18:32, 26 June 2017 (UTC)

Pictogram voting comment.svg Comment I'm not change my opinion that "web ..." aren't useful because how complicated "web standards" are now.

We can use spend time on references and who says what, but results wouldn't be useful as specifying each protocol and their client/server implementation. Java Servlet (Q375673), Internet Protocol version 6 (Q2551624), HTTP/2 (Q739120), Hypertext Transfer Protocol over TLS/SSL (Q44484). It was HTTP 1.1 long, long time ago to call everything "web". d1g (talk) 16:56, 26 June 2017 (UTC)

d1g the world wide web was invented by Tim Berners-Lee in 1989 or thereabouts (en:World Wide Web). Servers have been around since the internet began, at least 20 years earlier - see en:Server (computing) ("In computing, "server" dates at least to RFC 5 (1969)"). An email server, for example, responds to SMTP requests, not to HTTP, so it is not a web server, and certainly not a web application, but it is server software. Your use of "said to be the same as" here is clearly wrong, and your arguments are not making any sense at all. ArthurPSmith (talk) 18:18, 26 June 2017 (UTC)
That's my point?.. "server"/"client" was there since day 1 or so; you can't define and source "web framework" consistently.
"web framework" is meaningless: commonly it means "server", but sometimes it is used for "client" too. d1g (talk) 18:28, 26 June 2017 (UTC)
so why are you trying to claim something that you assert is "meaningless" is "the same as" something that is clearly well defined and "was there since day 1"? In fact, while "web framework" may be somewhat buzzwordish, in my experience it does have a tangible meaning distinguishing that kind of software environment from the other components of an http(s)-based internet service. Which is in fact why we have a whole enwiki page on it: en:Web framework. enwiki doesn't usually have entire long pages about things that are "meaningless". ArthurPSmith (talk) 19:12, 26 June 2017 (UTC)
I used "roughly same as", not exactly as (exact match (P2888)) and I removed these claims and with replaced them with facet of (P1269).
http/https changed a lot. and ""web framework" gained popularity only after 1992" I also wonder why there is a peak in 60s...
So, "web framework" usage is very limited (1992-) as opposed to "servers" and "clients".
Furthermore, number of application programming interface (Q165194) is grown so that "client" in HTTP sense is more relevant than before.
Basically, internet of things (Q251212) is a call to API-ize every hardware, even more programs would be "clients" of everything else. d1g (talk) 20:19, 26 June 2017 (UTC)

Qualifiers in QuickStatements 2?[edit]

QuickStatements is a great tool and I use it extensively however it is lacking documentation (as we already discussed at Project chat) so the main way of figuring things out is try and error. Reading User:Magnus Manske/quick statements2 page suggests that the tool might be supporting Qualifiers. Did anybody figured out how to do it? I would like to be adding statements like date of birth is "circa 500 BC" (as in Pythagoras (Q10261)). --Jarekt (talk) 18:46, 26 June 2017 (UTC)

  • Add TAB P1480 TAB Q5727902 behind the statements that would normally add 500 BC and before the sources. ChristianKl (talk) 18:59, 26 June 2017 (UTC)
  • You will find the documentation in the old version. Matěj Suchánek (talk) 19:07, 26 June 2017 (UTC)
Matěj Suchánek & ChristianKl, thank you. I misunderstood the documentation in the old version. I thought one can add multiple properties with a single statement, but it is one property and multiple qualitiers. That makes much more sense. Thanks again. --Jarekt (talk) 19:37, 26 June 2017 (UTC)

Ugh category duplication! Surely we can do better[edit]

It has been previously requested of Emaus and GZWDer to watch their category creations as they have had a tendency to create duplicates. In the past few days EmausBot (talkcontribslogs) and GZWDer (flood) (talkcontribslogs) have created 840 duplicate categories for Commons <-> EnWiki alone [2], and at a guess many others. This is unsustainable and an abuse of volunteer time and patience to have to resolve that sort of rubbish. There needs to be a more mindful approach to the creation of categories than the current apparent "wham bam thank you maam" approach that is taking place.  — billinghurst sDrewth 04:18, 27 June 2017 (UTC)

Given the amount of created category-items that seems to be an error rate that's less than 1%. It would be good if it would be lower, but it's not that bad.
If I look at the created items I don't understand why no label (Q30652379) got created by EmausBot (talkcontribslogs) without any labels. ChristianKl (talk) 08:41, 27 June 2017 (UTC)

Maybe @Magnus Manske:'s merge game can be changed in a way that lists of merge candidates like the one that Pasleim created go to the top? Given that currently, the game has a high false positive rate, merging might get more productive with presorted lists. ChristianKl (talk) 08:59, 27 June 2017 (UTC)


Does anyone know why the item guitar (Q6607) that describes the guitar is called bass? Mycomp (talk) 06:44, 27 June 2017 (UTC)

It does no longer. It was vandalised yesterday. Matěj Suchánek (talk) 07:05, 27 June 2017 (UTC)
I see. Thank you. Mycomp (talk) 07:48, 27 June 2017 (UTC)

help with a instance of (P31) in Jerusalem in religion items[edit]

What instance of (P31) and/or subclass of (P279) do I add to Jerusalem in Islam (Q6185220), Jerusalem in Judaism (Q2920442) and Jerusalem in Christianity (Q6185216)? DGtal (talk) 08:47, 27 June 2017 (UTC)

Use facet of (P1269). Sjoerd de Bruin (talk) 12:19, 27 June 2017 (UTC)

Documenting old names of TV stations[edit]

The name of Las Estrellas (Q80478) was Canal de las Estrellas until the 22 august 2017 when they changed their name to Las Estrellas. In the infobox of TV shows that started before that date the name of the station should be displayed as Canal de las Estrellas and for the ones after that date it should be displayed as Las Estrellas. When loading the data from Wikidata we use the label, but that way it's impossible knowing when each of the names should be displayed. How should the name change be added as Wikidata properties to allow us automatically taking the correct one? I've been thinking on using official name (P1448), but I don't know if there is any property that would be more appropriate for the current use case. Thanks. -- Agabi10 (talk) 11:39, 27 June 2017 (UTC)

You have the correct property, be sure to use date qualifiers and you are all set! :D Sjoerd de Bruin (talk) 12:19, 27 June 2017 (UTC)
I think that this discussion is resolved and can be archived. If you disagree, don't hesitate to replace this template with your comment. Agabi10 (talk) 14:42, 27 June 2017 (UTC)

Interacting with the Data Import Hub[edit]

Under there are presently a few projects that are about important data into Wikidata but most of the projects don't get any comments. It seems to me like it would be useful if we would provide more guidance at this page. ChristianKl (talk) 15:49, 27 June 2017 (UTC)

Thanks for pointing that out! I really need to write a tutorial to explain how to use OpenRefine for Wikidata imports. This is exactly what this tool is good for, but people are just not aware it exists. − Pintoch (talk) 18:10, 27 June 2017 (UTC)
It's not entirely clear whether the hub is for people trying to figure out how to do it (without asking for help on specific points) or people trying to find someone else to do it for them. Supposedly, the second group would better ask at Wikidata:Bot requests.
--- Jura 18:19, 27 June 2017 (UTC)

Changing constraints in a property[edit]

I'd like to add either clothing (Q11460) (my preference) or folk costume (Q3172759) as acceptable types for indigenous to (P2341). Do I need to formally propose this somewhere? - PKM (talk) 18:47, 27 June 2017 (UTC)

@PKM: You can propose it on its talk page, pinging those involved in its proposal. Mahir256 (talk) 19:54, 27 June 2017 (UTC)

Titles of films in other regions[edit]

When adding has quality (P1552) with title for Spain (Q27847754) and title for Hispanic America (Q27834579) for the film Saving Mr. Banks (Q3474574) as Jura1 said at the Infobox film of eswiki I realized that the addition triggers a constraint report. Is this way of defining the titles in other regions incorrect? And if it's correct where can the constraint be fixed? I can't find it neither in the talk page nor in the property page. -- Agabi10 (talk) 19:47, 27 June 2017 (UTC)

Once in a while, constraint definitions get re-loaded from property talk pages. Should be ok next month ..
--- Jura 19:52, 27 June 2017 (UTC)

Named time and style periods[edit]

There is a tiny problem with named time periods (read HUGE problems). Such periods not only have different names in different countries, but they also have different start and end times. They can even appear in different order in different countries. Very confusing. Very very confusing. In Norway we had the Merovinger time between younger and older iron age. In Sweden they had the Vendel time. In Denmark they had the Younger Germanic iron age. Iron age in Norway goes from 500 BC to approx 1000 AD, unless we are talking about Sami people which had an iron age from 0 AD to 1500 AD. It is a complete mess.

So what to do? It seems like Iron Age (Q11764) have a sort of solution, but incoming references might be wrong if they don't use additional specifier to identify which period and where it applies. And to whom, like for the Sami people. Use applies to part (P518) with an ethnic group?

An other strange example. Assume a building described as belonging to a specific architectural period like Swiss chalet style (Q2256729). This architectural period starts at another year in a neighboring country. Use Norway and Sweden as an example. In Norway the period started about 1840 and in Sweden about 1850. In 1855 it had already started to fall out of popularity in Norway, being replaced by Dragestil (Q4562834). In Norway there were factories mass producing such houses, exporting them to Sweden, and there being sold and built as Swiss chalet style. So, is a building built in Norway in late 1880-ish belonging to the Swiss chalet style if it is described on the Swedish Wikipedia? What style period does no label (Q14942980) belong to on Swedish vs Norwegian Wikipedia? Can we say that it is the same style period anyhow?

So, the description for a style period should probably both have a reference to where the object is, and where the object originated. There are probably some problems associated with the observer too, but this is more than complex enough for now.

This starts to get really messy. Anyone with an idea how this could be simplified? Jeblad (talk) 20:16, 27 June 2017 (UTC)