Wikidata:Project chat

Wikidata project chat

Place used to discuss any and all aspects of Wikidata: the project itself, policy and proposals, individual data items, technical issues, etc.
Please take a look at the frequently asked questions to see if your question has already been answered.
Requests for deletions and mergers can be made here.
IRC channel: #wikidata^connect

Start a new discussion

• Afrikaans
• العربية
• беларуская
• беларуская (тарашкевіца)
• български
• Banjar
• বাংলা
• brezhoneg
• bosanski
• català
• کوردی
• čeština
• словѣньскъ / ⰔⰎⰑⰂⰡⰐⰠⰔⰍⰟ
• dansk
• Deutsch
• Zazaki
• dolnoserbski
• Ελληνικά
• English
• Esperanto
• español
• eesti
• فارسی
• suomi
• føroyskt
• français
• Nordfriisk
• galego
• Alemannisch
• ગુજરાતી
• עברית
• हिन्दी
• hrvatski
• hornjoserbsce
• magyar
• հայերեն
• Bahasa Indonesia
• interlingua
• Ilokano
• íslenska
• italiano
• 日本語
• Jawa
• ქართული
• қазақша
• ಕನ್ನಡ
• 한국어
• kurdî
• Latina
• lietuvių
• latviešu
• Malagasy
• Minangkabau
• македонски
• മലയാളം
• मराठी
• Bahasa Melayu
• Mirandés
• مازِرونی
• Nedersaksies
• नेपाली
• Nederlands
• norsk bokmål
• norsk nynorsk
• occitan
• ଓଡ଼ିଆ
• ਪੰਜਾਬੀ
• polski
• پنجابی
• português
• Runa Simi
• română
• русский
• Scots
• davvisámegiella
• srpskohrvatski / српскохрватски
• සිංහල
• Simple English
• slovenčina
• slovenščina
• shqip
• српски / srpski
• svenska
• ꠍꠤꠟꠐꠤ
• ślůnski
• தமிழ்
• తెలుగు
• ไทย
• Tagalog
• Türkçe
• українська
• اردو
• oʻzbekcha / ўзбекча
• Tiếng Việt
• Yorùbá
• 中文

Edit

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 2026/10.

Notability and Commons categories

Wikidata:Notability says that an entity is eligible for inclusion in Wikidata if it "contains at least one valid sitelink to a Wikipedia, Wikivoyage, or Wikimedia Commons page". The latter is ambiguous; does "page" encompass categories? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 13:27, 16 November 2013 (UTC)

It should. Categories from WP and WV count. --Jakob (Scream about the things I've broken) 13:42, 16 November 2013 (UTC)

And from Commons? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 13:50, 16 November 2013 (UTC)

In principle, they should count as well. The problem is that due to ongoing discussion on whether Commons categories should be connected with Wikipedia pages or Wikipedia categories, very few of them have items.--Ymblanter (talk) 13:55, 16 November 2013 (UTC)

Side note: I have just closed that discussion. I will make a section here shortly. John F. Lewis (talk) 14:23, 16 November 2013 (UTC)

┌────────────────────────────────────────────────────────────────────────────────────────────────────┘

Thank you. In that case, please could an admin undelete Q15136093? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 15:23, 16 November 2013 (UTC)

 Done, but I removed the biographical information and links to userpages. --Jakob (Scream about the things I've broken) 15:47, 16 November 2013 (UTC)

Thank you; but why the removals? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 16:21, 16 November 2013 (UTC)

In fact, I now see you've deleted all the content, apart from the Commons link. Please restore it. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 17:47, 16 November 2013 (UTC)

Well, this far, content related to users do not meet notability. With the inclusion of Commons, because of pages like this, it's obvious that Wikidata:Notability is outdated. Andy Mabbet is an Author of content in Commons and we, sooner or later, have to accept that pages like this is a part of Wikidata. Though, that does not mean that I want to see pages like this about moi. -- Lavallen (talk) 18:12, 16 November 2013 (UTC)

I don't know who Andy Mabbet is; but Andy Mabbett (me), is the subject of a Commons content category, not just a creator of content. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 19:27, 16 November 2013 (UTC)

Pigsonthewing, the subject of the item is your images on commons, not you. --Jakob (Scream about the things I've broken) 18:18, 16 November 2013 (UTC)

Quite the contrary; the subject is (images - almost all created by other people - and an audio file, of) me. You will notice that, in the first part of this discussion, above, it is the unequivocal view of other editors that that warrants a Wikidata entry. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 19:27, 16 November 2013 (UTC)

Jakob, you might want to notice that Wikimedia Commons usually has multiple categories for content about a subject and content created by a subject. As Andy pointed out, Commons:Category:Andy Mabbett features media files depicting Andy Mabbett, while Commons:Category:Images by Andy Mabbett features images created by Andy Mabbett. This is commons practice on Commons (e.g. Commons:Category:Allan Warren and Commons:Category:Photographs by Allan Warren) and should be reflected on Wikidata accordingly. Commons:Commons:Categories#Categorization_tips uses the example of Rembrandt: Commons:Category:Rembrandt is categorised in other categories describing Rembrandt, his works are in sub-categories like Commons:Paintings by Rembrandt‎. Admittedly there are a lot of categories packed with content that should be moved into sub-categories, yet this doesn't affect these categories' logic. Regards, Christoph Braun (talk) 10:44, 17 November 2013 (UTC)

Policy Changes

Hello all,

Per my second addendum to the Commons links RfC, Notability was changed to specifically state a Commons category no longer meets notability. Commons site links are now exclusively (probably wrong word) for Commons pages or galleries. I don't have much time to specifically elaborate on this at the moment so if someone else wishes to, please feel free. John F. Lewis (talk) 21:25, 18 November 2013 (UTC)

Other than the single reference in your closing comment, I don't see any mention of notability in that discussion. Can you point it out, please? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 12:54, 20 November 2013 (UTC)

By my interpretation of WD:N, Commons category links are not allowed at all, which, from what I can tell, is not supported by any option on Wikidata:Requests for comment/Commons links. The Anonymouse (talk) 17:04, 20 November 2013 (UTC)

That is the recent change, to which I refer. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 09:20, 21 November 2013 (UTC)

OK, I think I understand now. Because option III will be the ultimate result, Commons category links should not be used until option III is available.

By the way, what is the development team's opinion on this matter (since it would require significant development to accomplish)? I though John F. Lewis said he talked to Lydia about this, but I'm not exactly sure. The Anonymouse (talk) 17:09, 21 November 2013 (UTC)

That doesn't match not my understanding of any of the options discussed in the RfC, and certainly not the option approved by the closing admin as an interim solution. However I can't claim to understand his changes to WD:N, including a change made since the above posts.

The development team's opinion on option III was explained at Wikidata:Contact the development team/Archive/2013/11#merging_items_for_Categories_with_items_for_the_Category_topic (unfortunately not until late during the RfC). Briefly, they think option III "does not seem impossible but would need a lot more investigation" and that given other priorities, this won't happen for the "forseeable future". --Avenue (talk) 20:16, 28 November 2013 (UTC)

Nudge. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 18:16, 26 November 2013 (UTC)

This has been discussed further elsewhere, most recently at Wikidata talk:Requests for comment/Commons links#Review of closure. --Avenue (talk) 11:36, 27 November 2013 (UTC)

Items without a claim per language

I was wondering how many items still lack claims and if I could see a pattern or difference per language. The totals are at http://tools.wmflabs.org/wikidata-todo/stats.php -> Statements per item (no anchor so I can't link directly). I couldn't find it linked from Special:Statistics so I did the query myself. I also included the items per language which link to Wikimedia category (Q4167836).

wiki	Total items	Items without claims	Percentage without claim	Category items	Percentage categories
enwiki	5345886	1437061	26,88%	1028734	19,24%
nlwiki	1787341	544923	30,49%	78669	4,40%
dewiki	1777860	373570	21,01%	181057	10,18%
svwiki	1688682	813612	48,18%	118712	7,03%
frwiki	1639157	456512	27,85%	190502	11,62%
itwiki	1295921	311384	24,03%	207374	16,00%
ruwiki	1272505	328650	25,83%	123963	9,74%
eswiki	1254818	329175	26,23%	206838	16,48%
plwiki	1108496	367814	33,18%	80162	7,23%
warwiki	1076622	608111	56,48%	29164	2,71%
cebwiki	996749	615066	61,71%	14019	1,41%
jawiki	989850	539723	54,53%	75283	7,61%
ptwiki	916022	257882	28,15%	112948	12,33%
zhwiki	778798	273453	35,11%	80724	10,37%
viwiki	696311	172985	24,84%	40395	5,80%
ukwiki	541649	165242	30,51%	47679	8,80%

Do you see anything interesting? Multichill (talk) 18:55, 20 November 2013 (UTC)

Is a dewiki-item an item that only has a dewiki site link? --Tobias1984 (talk) 19:45, 20 November 2013 (UTC)

Nope, any item that has a link to dewiki. An item which has multiple language links shows up multiple times here. Multichill (talk) 20:15, 20 November 2013 (UTC)

I see that sv/ceb/war-wiki who has high percentage of bot-created articles have a high percentage here. -- Lavallen (talk) 19:49, 20 November 2013 (UTC)

I'm thinking the same. Other option is that not all categories have been tagged yet. Ceb has 113K categories, but only 14K here, War has 128K categories, but only 29K here, SV has 255K, but 128K here. Multichill (talk) 20:15, 20 November 2013 (UTC)

My bot is working on Category, but I am started from biggest wiki. --ValterVB (talk) 20:31, 20 November 2013 (UTC)

I just compiled a list and my bot has been working on that for the last month. That covers most of the big wiki's so you wont get a lot of hits there. What's your approach? Multichill (talk) 20:46, 20 November 2013 (UTC)

I use Special All pages (Category namespace) if exist on Wikidata Bot Add instance of (P31)=Wikimedia category (Q4167836) and descriptions for it,fr,de,pt,pt-br,ru,sv. (example) and delete P107 (P107). --ValterVB (talk) 20:57, 20 November 2013 (UTC)

Ah, I see. Could you run on ruwiki, eswiki, viwiki and ukwiki? I haven't touched those yet so you should get a lot of hits. Multichill (talk) 21:17, 20 November 2013 (UTC)

eswiki is done, when I have finished svwiki I can do ruwiki, viwiki, ukwiki, cebwiki and warwiki. --ValterVB (talk) 21:29, 20 November 2013 (UTC)

How can I find items without a claim in my language please? — Ayack (talk) 20:05, 20 November 2013 (UTC)

en, fr, de nl, sv, ceb & war. Multichill (talk) 20:40, 20 November 2013 (UTC)

Thanks! — Ayack (talk) 21:07, 20 November 2013 (UTC)

Bot generated taxons

A quick look through the lists of items without claims reveals a lot of bot generated articles about taxons. Lot's and lot's of insects. I was wondering what to do with them. At first I thought about adding taxon rank (P105) and P74 (P74), so I looked at Wikidata:List of properties/Terms#Biology and ended up at Wikidata:Taxonomy task force. That task force page contains the statement that every item about a taxon should contain instance of (P31) taxon (Q16521) (or monotypic taxon (Q310890)). That's something a bot can do. That also makes it easier for that project to see what's in scope. We're talking about quite a few items here. Dutch Wikipedia will already give 830.000 items that are about taxons, Swedish Wikipedia even more than a million, War Wikipedia 840.000 (out of a total of 960.000 articles), Ceb Wikipedia 845.000 (out of total 892.000 articles). I fired up a bot for the Dutch Wikipedia (pwb.py claimit.py -family:wikipedia -lang:nl -transcludes:Taxobox -namespace:0 P31 Q16521). Would be nice to have three other operators for the other languages. Who wants to help? Multichill (talk) 17:44, 21 November 2013 (UTC)

Tagging items that way is not really helpful. --Succu (talk) 18:45, 21 November 2013 (UTC)

Why wouldn't that be helpful? With no claims you don't know what it is, now you know work on improving that subset, there is even a task force which has this in it's requirements. I leave adding other properties to the taxo experts, but now at least you have a good overview of what's in scope of the project. Multichill (talk) 18:54, 21 November 2013 (UTC)

I'm part of this taskforce. So I ask you again: stop this task and lets discuss first. --Succu (talk) 18:58, 21 November 2013 (UTC)

Since it's the same bot-operator who has created articles in sv, ceb and war, you will end op with the same set of items.-- Lavallen (talk) 19:02, 21 November 2013 (UTC)

Succu is not just part of the taskforce, he is the major part of the task force, putting in an incredible amount of work. Personally I don't like instance of (P31) taxon (Q16521); it just looks like clutter to me. What each item should really contain is taxon name (P225), taxon rank (P105) and parent taxon (P171) (the latter could be the genus for each species). In itself, I see no reason why these three could not be added by a bot, but as Succu is the one doing all the work, it is his opinion which weighs heaviest. - Brya (talk) 06:49, 22 November 2013 (UTC)

        P.S. The plural of "taxon" is "taxa" (only in French is it "taxons").

The problem with using simply claimit for P31 is that you do not distinguish between taxon and monotypic taxon. Also, the data quality of these bot-generated articles is quite low. Nonetheless, as we have the items, we need taxon name (P225), taxon rank (P105), and, if possible, parent taxon (P171) deployed to be able to identify the corresponding information in literature or taxonomic databases. Here, your help is very welcome! Once we have this done, we can choose to add instance of (P31) as monotypic or not based on sources and do not need to "repair" what is now naively deployed in masses.  — Felix Reimann (talk) 10:06, 22 November 2013 (UTC)

Now we have a lot of items with no claims at all. If these items get one or two of the right claims, it brings them in scope and reach of a task force. I don't see adding instance of (P31) taxon (Q16521) instead of monotypic taxon (Q310890) as an error. monotypic taxon (Q310890) is a subclass of taxon (Q16521) so it's semantically correct. I see it as something that can be improved. An error would be if I added railway station (Q55488). This is the way how it's handled in other domains too. It's just like adding railway station (Q55488) to Uhlandstraße (Q568224) and later updating it to metro station (Q928830)

I want to get this straight, because I do not like it to get attacked, accused of making errors and have legitimate edits reverted. Multichill (talk) 18:00, 22 November 2013 (UTC)

Multichill: If you want to get things straight and not have people telling you off then I suggest you listen to what people are telling you, even if it isn't what you want to hear. Go back and read all the response to your post above and then do what those responses are suggesting.

Succu: What is your idea for adding statements to these items? I have looked at a bunch of these and most of these are species with their latin binomial names as the wikipedia article title so in theory you can take the use the first half of the title for the 'parent taxon' (though many of these genus items are not created yet), the whole title for the 'taxon name' property and 'species' as the 'taxon rank'. Would that work? Filceolaire (talk) 22:02, 22 November 2013 (UTC)

Dear Multichill, at first, I want to thank you for dropping a note at the task force talk page. It would be great if all would inform the task forces before rearranging things where others already had lengthy discussions and perhaps even found working approaches. The active members of the taxonomy task force do a lot of manual work to fix all the errors which we have in the items. The bots which are already running in this field are specialized to create as less additional workload as possible by doing it the first time right. If you prefer to first import and then fix errors, this is also fine for me. But this means: You as the bot operator are also responsible for addressing the errors. And I guarantee that if using a plain claimit, you will get a lot of them. The problem in this field is manyfold: 1) In taxonomy you do not have a single reference source which defines what is correct. 2) Bigger Wikipedias have an extremely high data quality. If we ever (!) achieve to convince them to switch to use Wikidata, we need to prove that also our data has an acceptable level of quality. 3) Several other Wikipedias have bot created pages which, and this is the main problem, rely partially on databases which are of extremely bad quality or are hopelessly outdated. Thus, we definitely need your help. But importing junk data is not helpful.  — Felix Reimann (talk) 09:41, 25 November 2013 (UTC)

• As a member of taxonomy taskforce I would be glad if some bot add instance of (P31) taxon (Q16521) (or monotypic taxon (Q310890)) and taxon rank (P105). --Infovarius (talk) 06:44, 29 November 2013 (UTC)

Proposal to enable abuse filter blocking

If you've been browsing the block log, you'll notice a lot of blocks that came because spambots were triggering Special:AbuseFilter/5. Blocking the addresses has become a rather large chore (because we want to block them as soon as they are identified, since they might later try to evade the filter), and it would of great convenience if the filter could just do the blocking automatically. Technically, because these are all IPs, finite block durations would be required, but can be configured for IPs separately from accounts. The filter has been very accurate, but to further guard against false positives, it would warn before blocking the IP. The change requires community consensus in favor of it.

•  Support as proposer.--Jasper Deng (talk) 22:22, 21 November 2013 (UTC)

• I don't see how to do this just by modifying the filter. Are you proposing using an adminbot? --Jakob (Scream about the things I've broken) 22:31, 21 November 2013 (UTC)

• No, the extension itself would block it, but the technical change to enable that functionality needs consensus. See Meta's block log for an example of it in action ("Abuse filter (talk | contribs) blocked...").--Jasper Deng (talk) 22:37, 21 November 2013 (UTC)

  The blocking component has a built in account "AdminBot" that undertakes the blocking. The change that JD is proposing requires the change to the underlying config, so the consensus is required for the bugzilla request to be actioned.  — billinghurst sDrewth 00:53, 22 November 2013 (UTC)

•  Support In my opinion, blocking these accounts is a quite useless task to do but it is still better when they are blocked directly by the abuse filter rather than by bored admins whose energy could better be spent somewhere else (real life or maybe WD:RFD). Vogone talk 22:39, 21 November 2013 (UTC)
•  Support But this should only be used for very accurate filters. --Rschen7754 00:46, 22 November 2013 (UTC)

  My recommendation is that the application of this component requires rigorous in situ testing prior to throwing the switch, AND it requires permanent monitoring. No set and forget.  — billinghurst sDrewth 00:53, 22 November 2013 (UTC)

 Comment To note that if there is interest, this wiki can be part of the global abuse filters that are run by stewards. Basically there are a few filters that stewards set that identify spambots, and from meta we identify those and globally llock them straight away. Where there is a specific issue, it also allows stewards to see it, and if necessary, also undertake checkuser.  — billinghurst sDrewth 00:53, 22 November 2013 (UTC)

•  Support --DangSunM (talk) 02:14, 22 November 2013 (UTC)
•  Comment Can somebody explain what filter 5 is doing? I cannot read it! -- Lavallen (talk) 08:59, 22 November 2013 (UTC)
  • It checks for certain blacklisted links that are usually random gibberish, or a particular syntax often used by spambots.--Jasper Deng (talk) 09:17, 22 November 2013 (UTC)
•  Support Why not? --by Revi_TalkCMG at 10:30, 22 November 2013 (UTC)
•  Support original proposal with Rschen's caveat. TCN7JM 23:23, 22 November 2013 (UTC)
•  Support With Rschen's and Billinghurst's recommentations. — ΛΧΣ²¹ 06:13, 23 November 2013 (UTC)
•  Oppose (Edited later at 04:10, 25 November 2013 (UTC):  Neutral to weak support if the message clearly warns the user that (s)he will be blocked before it actually does block. This message must be translated on a wiki like Wikidata.) The potential benefits are not worth the risk of discouraging new users. Prefer soft security when feasible. Unless you are receiving an unmanageable amount of spam, I don't see the need for this. I know resistance is futile. This will surely pass. If it does, you should check each filter very carefully and use the following rules:
  1. Any blocking filter that does not match any spam in a week must be disabled, unless there is a very good reason for keeping it (read: LTA, not spambot pattern).
  2. The logs of every blocking filter should be checked regularly for false positives. If a false positive is found, the filter must be immediately disabled and reviewed until it is fixed. No exceptions, even for users who "look like spammers".
  3. The message for blocking filters should be very clear about what happened, should not be discouraging at all, and must provide clear instructions for how to appeal the block and report the filter as incorrect.

Just my 2 cents. Also, global filters should not block. πr² (t • c) 03:45, 25 November 2013 (UTC)

• • Did you even look at the relevant log? Please tell me that admin time is not better spent manually blocking all of that.--Jasper Deng (talk) 03:49, 25 November 2013 (UTC)

• I have opened bug 57681 amd submitted a patch for this. Consensus seems fine plus the adoption of such a feature would be extremely helpful. John F. Lewis (talk) 21:01, 27 November 2013 (UTC)

Block lengths for IPs

Here, we need to also decide on which block length to use. Account blocks will be indefinite (however, we do not plan to use a filter to block accounts anytime soon).--Jasper Deng (talk) 02:54, 28 November 2013 (UTC)

3 months

•  Support - most commonly used duration.--Jasper Deng (talk) 02:54, 28 November 2013 (UTC)
•  Support - Spam blocks last this long anyway. Ajraddatz (Talk) 02:54, 28 November 2013 (UTC)
•  Support --Rschen7754 07:25, 28 November 2013 (UTC)
•  Support --by Revi_TalkCMG at 08:49, 28 November 2013 (UTC)
•  Support - This makes the most sense. TCN7JM 19:41, 28 November 2013 (UTC)
•  Support--GZWDer (talk) 16:19, 30 November 2013 (UTC)
•  Support-- As User:TCN7JM said. --Clarkcj12 (talk) 01:07, 1 December 2013 (UTC)
•  Support-- Seems to strike a balance between stopping spambots and keeping IP addresses free. I'd also support a slightly longer block if these bots cause a lot of harm before being stopped. --Arctic.gnome (talk) 04:53, 1 December 2013 (UTC)
•  Weak support This would only help Wikidata. Matěj Suchánek (talk) 12:55, 1 December 2013 (UTC)

Default (indefinite)

•  Oppose - IPs should not be blocked indefinitely.--Jasper Deng (talk) 02:54, 28 November 2013 (UTC)
•  Oppose too long. --by Revi_TalkCMG at 03:31, 28 November 2013 (UTC)
•  Oppose - No indefblocks for IPs. TCN7JM 19:42, 28 November 2013 (UTC)
•  Oppose - as some can be shared IPs. Also I think it would be too long especially for like the first block. --Clarkcj12 (talk) 01:04, 1 December 2013 (UTC)

Quantities datatype available for testing

Hey everyone!

We have a first version of the quantities datatype available for testing now on test.wikidata.org. It still has a few rough edges that we need to fix before we can enable it on Wikidata but I wanted to give you a chance to test it now already and help find issues early.

Here's a test property using the quantities datatype: https://test.wikidata.org/wiki/Property:P63

Cheers --Lydia Pintscher (WMDE) (talk) 23:04, 21 November 2013 (UTC)

One bit of initial feedback - we should have the ability to specify a restricted domain for a particular property (like integers only (might be different data types), and more importantly positive/negative or a particular range).--Jasper Deng (talk) 23:10, 21 November 2013 (UTC)

Restricting the world again? ;) --Denny (talk) 23:36, 21 November 2013 (UTC)

It could be editable afterwards (although Lydia said on IRC that this won't happen) - it'll just enforce guidelines on the proper usage of the property.--Jasper Deng (talk) 23:39, 21 November 2013 (UTC)

Two questions: should the enter key save the value like it does in the other properties? Also, what are the rules on commas versus periods - is this determined by the language? --Rschen7754 00:45, 22 November 2013 (UTC)

Yes it should save on enter. Localisation isn't done yet. --Lydia Pintscher (WMDE) (talk) 08:14, 22 November 2013 (UTC)

We have bugzilla:56685 for localisation. --Lydia Pintscher (WMDE) (talk) 13:53, 26 November 2013 (UTC)

Exciting! Is the explicit signing of positive numbers intended to be consistently applied, or is that one of the "rough edges" you're referring to? For example, since a population can never be negative, to have it listed as "+5,000,000" just makes pages harder to read without any obvious benefit. I'm guessing it's a known issue but wanted to check. :-) --Eloquence (talk) 07:49, 22 November 2013 (UTC)

Jep. That's one of the rough edges ;-) The plus will go away. --Lydia Pintscher (WMDE) (talk) 08:14, 22 November 2013 (UTC)

Lydia: sometimes the plus sign is wanted, example: w:Oxidation state.--Micru (talk) 19:03, 22 November 2013 (UTC)

We have bugzilla:57589 for that now. --Lydia Pintscher (WMDE) (talk) 13:57, 26 November 2013 (UTC)

1. How will the unit look like when we deal with length/area etc?

2. How soon will we be able to modify the precision through the UI? I have a census-report with +/-50 in the numbers, ready to type in here. -- Lavallen (talk) 08:43, 22 November 2013 (UTC)

Two problems:

1. Scientific notation is not supported.
2. "+" should be omitted.

--GZWDer (talk) 10:16, 22 November 2013 (UTC)

The largest number one can enter seems to be 10^{126}-1, which seems to be reasonable large. I found only that the number of positions for V-Cube 7 (Q1788746) is larger (besides some mathematical constructions). --Zuphilip (talk) 12:25, 22 November 2013 (UTC)

• Scientific notation should be supported because we can hardly count the number of 0's in 100000000000000000000000000000000000000000000000000000000000000000000000000000000 (one googol).--GZWDer (talk) 15:05, 22 November 2013 (UTC)
  • We have bugzilla:57588 for that now. --Lydia Pintscher (WMDE) (talk) 14:59, 26 November 2013 (UTC)

Prefixes

Good to see the progress! It would be interesting if simple prefixes like ±, <, >, ≈, etc were also supported. Is this perhaps planned?--Micru (talk) 13:21, 22 November 2013 (UTC)

Not planned for the initial version. I'll have to see how to solve this later. No concrete plans yet. --Lydia Pintscher (WMDE) (talk) 14:56, 26 November 2013 (UTC)

P279 & P31 (& P361) for taxons (included species)

What should P279 & P31 be for taxons? Currently, we use P31=taxon (Q16521) and P279=organism (Q7239) in human (Q5). But I think we can probably copy P75 (P75) to P31 (or probably still use P31=taxon (Q16521)), and probably merge all of P273 (P273), P75 (P75), P76 (P76), P77 (P77), P70 (P70), P71 (P71), P74 (P74), P89 (P89) to P279 (or probably P361) with P75 as qualifier, and P273 (P273), P75 (P75), P76 (P76), P77 (P77), P70 (P70), P71 (P71), P74 (P74), P89 (P89) as well as parent taxon (P171) will be deprecated and can be deleted. Wikidata:Requests for deletions/Archive/2013/04/02#Property:P171 was a discussion before.--GZWDer (talk) 15:51, 22 November 2013 (UTC)

The recommendation in Wikidata:Taxonomy task force is P31: taxon (Q16521) or monotypic taxon (Q310890). I think it makes sense. It seems that parent taxon (P171) is not likely to be deleted, as it often makes life easier than subclass of (P279). I guess we can also add the above taxon in subclass of (P279) if that sounds useful. Hopefully, we will be able one day to document that a property is a subproperty of another so that we do not need to repeat things thigs way. --Zolo (talk) 16:05, 22 November 2013 (UTC)

I'm sure that I can speak for all active members of WD:Taxonomy task force (which you did not ask before starting this discussion) that we of course keep parent taxon (P171). Perhaps you should read the answers you got at some of the other threads you started for related topics.  — Felix Reimann (talk) 16:09, 22 November 2013 (UTC)

You can speak for me... :) @all owl-theoreticians: there are a lot of alternatives to model a domain. For a litte, but import part - ranks, taxonnames and how they are related - we have a good working solution. This solution is in agreement with real owl solutions out there. --Succu (talk) 16:40, 25 November 2013 (UTC)

@Zolo This is a case of class classification :) TomT0m (talk) 18:33, 22 November 2013 (UTC)

Don't use instance of (P31); use taxon rank (P105) instead - it's a specialised synonym for P31. Don't use subclass of (P279); use parent taxon (P171) instead. It's a synonym for P279 for use in taxonomy. Don't use part of (P361); it's for linking things to bigger things; not for linking classes to bigger classes. All taxonomy is about classes so there is not much scope for part of (P361) in taxonomy unless you are talking about anatomical features. Filceolaire (talk) 21:37, 22 November 2013 (UTC)

I don't like the idea we use sbecialised properties without proper support in the Wikibase software : this mean a sotware coded for any class won't work for taxons as is:This is why I advocate to class classes and not for upper taxon: TomT0m (talk) 00:43, 23 November 2013 (UTC)

So, remove P279 & P31 or set P31=taxon (Q16521) and P279=organism (Q7239)?--GZWDer (talk) 06:03, 23 November 2013 (UTC)

I'd say, remove p279==organism (Q7239) (but there might be other useful values for p279) and set P31=taxon (Q16521)

@tomT0m, Talk:Q16521 :-). Though the template is very much in alpha mode, the class documentation is segregated in Talk:Q16521/class so it can easily be transcluded in other places like task force pages. Hopefully, the can avoid some duplicates. --Zolo (talk) 08:45, 23 November 2013 (UTC)

Great work, congrats :) Just a thing that make me wonder : is really taxon a subclass of class ? TomT0m (talk) 15:26, 23 November 2013 (UTC)

Each Taxon is a collection of some organisms and excludes other organisms so (for me) it is a class.

If every item with taxon rank (P105) needs instance of (P31):taxon (Q16521) as well then I guess every items with P132 (P132) should also have instance of (P31):administrative territorial entity (Q56061)? Filceolaire (talk) 14:49, 25 November 2013 (UTC)

I actually asked Amir as he was running Dexobot over the items removing GND main type that those with P132 (P132) have added instance of (P31) as a duplicate of P132 (P132). This would help us delete P132. :^) --Izno (talk) 01:01, 26 November 2013 (UTC)

TomTom: As (I'm pretty sure) I'm the one who added that claim, I would argue that it is. A taxon is a grouping, a set, or a collection of things which all share certain properties, which by most definitions is a class. What is most telling of course is that species, genus, or any of the other taxonomic ranks are subclasses of taxon, which makes the P31 taxon a duplicate statement. The problem of course is that you silly taxonomy task group people have decided to duplicate P279 and P31 by using taxon parent and taxon rank respectively. :^) --Izno (talk) 01:01, 26 November 2013 (UTC)

You're right, I figured. If we take all the classes of objects, taxons are surely inside them. TomT0m (talk) 19:19, 26 November 2013 (UTC)

I've been following this discussion from the sidelines for a while, and just wanted to chime in and say that Izno's comment reflects my views precisely. As I argued in April, parent taxon (P171) is one of the clearest examples of redundancy with subclass of (P279) on Wikidata, and should be deleted. Having domain-specific 'subclass of' and 'instance of' properties is a bad idea, as has been agreed upon in many property deletion discussions; having a domain-specific 'subclass of' property for biological taxonomy is no exception.

The idea that parent taxon (P171) is "in agreement with real owl solutions out there" is, to my knowledge, incorrect. Show me an example of a biological taxonomy encoded in OWL that uses 'parent taxon' instead of 'subclass of'. I haven't seen one. Instead, what I've seen is that very large, established biological ontologies use the standard subclass of (P279) property and not a custom 'parent taxon' property to define parent taxon. UniProt, a consortium of some of the biggest national bioinformatics organizations in the world, uses 'rdfs:subClassOf' (on which 'subclass of' P279 is based) to specify parent taxon. See for example UniProt's structured data on 'dog', defining the parent taxon Canis lupus (gray wolf) with 'subclass of', not 'parent taxon'. This makes the data standards-compliant and easily queryable, as explained here. Emw (talk) 17:05, 30 November 2013 (UTC)

Property idea: "same as the compound property"

This is a slightly long question and I've mentioned the problem before, but bare with me because I think we still need to sort this out and I'm wondering if a new qualifier is the answer.

I keep wanting to give something a statement with the qualifier P642 (P642) alongside a compound statement that says the same thing. For example, under Barack Obama (Q76)=>position held (P39), we need to have both President of the United States (Q11696) and president (Q30461)=>P642 (P642)=>United States (Q30) so that he will be included when people gather data about presidents and when people gather data about presidents of the United States. But it seems inefficient to repeat all of the same qualifier information for both (like start date, end date, preceded by, succeeded by, and subject of). This is already being done on a number of pages, such as Rob Ford (Q169303). Under this system, the Obama entry would look like this:

• Barack Obama (Q76)=>position held (P39)
  • President of the United States (Q11696)
    • follows (P155)=>George W. Bush (Q207)
    • start time (P580)=>20 January 2009
    • statement is subject of (P805)=>presidency of Barack Obama (Q1379733)
  • president (Q30461)
    • P642 (P642)=>United States (Q30)
    • follows (P155)=>George W. Bush (Q207)
    • start time (P580)=>20 January 2009
    • statement is subject of (P805)=>presidency of Barack Obama (Q1379733)

Could there be a way to condense the general title and the compound title into one statement? One option would be to use "president=>of=>US" with statement is subject of (P805)=>President of the United States, but there are two problems with that. First, "subject of" is already being used for more specific entries, like "Presidency of Barack Obama". Second, if we removed "position held=>President of the United States" from Obama in favour of "position held=>president=>of=>United States", would that make it more difficult to sort the data?

I'm wondering if we should have a new parameter that links to compound titles, so in my Obama example we would have

• Barack Obama (Q76)=>position held (P39)
  • president (Q30461)
    • P642 (P642)=>United States (Q30)
    • follows (P155)=>George W. Bush (Q207)
    • start time (P580)=>20 January 2009
    • statement is subject of (P805)=>presidency of Barack Obama (Q1379733)
    • same as the compound property=>President of the United States (Q11696)

Would that make sense? --Arctic.gnome (talk) 01:12, 23 November 2013 (UTC)

Another approach : take the president of the united states item, and express the nature of this item of . Then if someone searches the president of something he as to make is query in to steps (or a little more complex query) : first find the item for the united states, then the president associated. We then can use both of your approaches without to duplicate anything, just to create an item for each kind of presidents. TomT0m (talk) 15:21, 23 November 2013 (UTC)

That is in my opinion the only reasonable solution. A query for all presidents can then be performed by something like http://208.80.153.172/wdq/?q=claim[39:%28CLAIM[31:248577]%29]. But there seems to be some inconsistency with the statements:

• instance of (P31) President of the Republic (Q248577): 5 results
• subclass of (P279) President of the Republic (Q248577): 5 different results
• instance of (P31) president (Q30461): 11 results
• subclass of (P279) president (Q30461): 6 results

Moreover, there is Q7241207. --Zuphilip (talk) 17:54, 23 November 2013 (UTC)

OK. Now they are all subclass of (P279) president (Q30461) except for one where the title was explicitly "President of the Republic" rather than "President". Filceolaire (talk) 21:58, 23 November 2013 (UTC)

• I want to make sure I'm understanding User:TomT0m and User:Zuphilip's suggestion. On the person's page we would have position held => President of Foobar (or Mayor of Foobar, Emperor of Foobar) and on the President of Foobar page we would have subclass of => president => of => Foobar. Is that right? When I brought up a related issue, a couple editors strongly wanted us to use "president" with the "of" qualifier on the person's page, but Zuphilip's example shows that searches can be done either way. The one this that still confuses me is that earlier in the month we had a discussion where I thought you said "Prime Minister of Foobar" was an instance of "prime minister" rather than a subclass. --Arctic.gnome (talk) 17:48, 25 November 2013 (UTC)

There is a misunderstanding, I see nothing like "Prime Minister of Foobar" was an instance of "prime minister" anywhere in this conversation. How do you deduce that ? TomT0m (talk) 18:59, 26 November 2013 (UTC)

On the page President of the United States (Q11696) there is now the information that President of the United States (Q11696) subclass of (P279) president (Q30461) . --Zuphilip (talk) 11:42, 29 November 2013 (UTC)

Work on disambiguations

I have discovered a way to find some items to merge – many disambiguations. Some items between Q9,000,000 and Q15,000,000 are disambiguations having only one link. Many of these can be merged with another lower-Qid-item.

The way to find duplicates:

1. Open Special:Search and have only ticked main namespace.
2. Enter an affix which usually identifies a disambiguation page.
   • I don't recommend entering "disambiguation", "Begriffsklärung" etc., because they are used very often. Use some less used like "olika betydelser" or "discretiva".
   • If there is a punctuation within the affix, the part starting with the letter must be replaced by an asterisk (due to issues with the search engine) – "rozcestník" must be replaced by "rozcestn*".
   • You should see something like this. If not, add -disamb* to the searched string.
3. If you are fortunate, you will see result containing items having only one disambiguation link. Now you can find pages to merge using search (while searching, add dhomon* to the key word because thanks to a bot all disambiguations have "d'homonymie" within French label).

I hope it is interesting and someone joins in.

Maybe a bot can be made for this because this is much work for few humans. Matěj Suchánek (talk) 19:55, 24 November 2013 (UTC)

• Please see Wikidata:Project_chat/Archive/2013/11#Disambig_page_link: There're no consensus about which disambiguations can be linked. Maybe we should only merge items with links with SAME name in different languages first.--GZWDer (talk) 10:32, 26 November 2013 (UTC)

"Instance of" as qualifier?

Whenever we needed a general property that can be customized, we used the pattern "general property=>item" with qualifier "type of general property=>item that describes the type". An example of this can be seen in relative (P1038) // kinship to subject (P1039). Recently I'm seeing instance of (P31) used as a qualifier of the general property instead of using a specific qualifier. I'm not sure that "instance of" should be used as qualifier because to me it only makes sense as a membership property. However, this kind of use of p31 reflects the need of a property that can be used that way. Are you fine using p31 as qualifier or should we have a dedicated property for such cases? --Micru (talk) 23:27, 24 November 2013 (UTC)

I don't appreciate that use myself, but I have seen it. Can you provide an example that you've seen recently of it, so that we might describe the use case more exactly? --Izno (talk) 00:53, 25 November 2013 (UTC)

Don't remember which item, but on the property proposal for "upper stage" there was the suggestion to use "instance of" as qualifier. Maybe Tobias1984 can present his thoughts on the subject?--Micru (talk) 10:16, 25 November 2013 (UTC)

At this point in time I really don't know if using instance of (P31) in a qualifier is a good idea or not. We just know too little about how the query system will work. The most important thing is probably that we stay consistent in as many items as possible. Maybe a developer, or semantic-web expert can weigh in, or even start an RfC about the correct usage of qualifiers. --Tobias1984 (talk) 11:00, 25 November 2013 (UTC)

I have started the qualifier proposal "specifically" to cover these kind of cases.--Micru (talk) 09:44, 26 November 2013 (UTC)

Deleting P107 (P107) and Constraint violations

Please, you, who are massively removing P107 (P107) from items, chhange it in property talk pages, because this property is widely used as Constraint violations, see e. g. [1]. JAn Dudík (talk) 07:36, 26 November 2013 (UTC)

people who are writing systems of constraint, should rewrite their systems not people who are removing P107. Amir (talk) 07:53, 26 November 2013 (UTC)

In other words: We destroy, the other should do the clean up ;( Imho it's no improvement to remove P107 and leave thousands of items without any info. --88.130.163.18 08:14, 26 November 2013 (UTC)

People who are removing make usually some instance of (P31) = Q1234 instead. So is no problem to replace it or to delete it. JAn Dudík (talk) 08:31, 26 November 2013 (UTC)

A GND main type usually can't be replaced by exactly one "instance of" in the constraints lists (and of course, if I replace one property in one item, I'm not going to check a thousand properties if I can update them accordingly). Is there a list compiled somewhere what possible replacements there are for the GND main types? Like:

• person (Q215627) -> human, fictional character, deity, ...
• organization (Q43229) -> company, band, ...
• event (Q1656682) -> ...
• creative work (Q386724) -> book, film, painting, ...
• term (Q1969448) -> ...
• place, geographical feature (Q618123) -> ...
• disambiguation page (Q11651459) -> disambiguation page

Is this even possible in a meaningful way without the constraint system knowing about subclasses and hierarchies? --YMS (talk) 08:46, 26 November 2013 (UTC)

The constraint system knows about subclasses and hierarchies, so you may as well use it and replace {{Constraint:Item|property=P107|item=QXX}} with {{Constraint:Type|class=QXX|relation=instance}}. Of course, more specific constraints are often better. --Zolo (talk) 08:57, 26 November 2013 (UTC)

Attributing museum directors II

In Wikidata:Project_chat/Archive/2013/11#Attributing_museum_directors I asked for a solution on attributing museum directors. Thanks to the effort of The Anonymouse we now have director / manager (P1037). After having a closer look at w:Corporate title I think that manager/director is too ambiguous. A manager is a middle management title while a director is a senior management title, therefore it wouldn't be possible to run meaningful queries, once people start populating items with this property. I'd propose to separate these properties into two distinct ones. Some people thought about creating a more general property such as "corporate title". Note that having atomic and unique properties allow us to add meaningful qualifiers. property:manager -> item:john doe -> qualifier:start date -> item:2013 would be a good approach my opinion. I'd love to hear your opinions, thanks. Regards, Christoph Braun (talk) 12:53, 26 November 2013 (UTC)

This kind of property seems too specific, but you can use a qualifier for the same effect, either P794 (P794) or the recently suggested "specifically". Then you would have property:manager/director -> item:john doe // qualifier1:specifically/as -> item:curator // qualifier2:start date -> item:2013.--Micru (talk) 17:49, 26 November 2013 (UTC)

In that case all position titles should be obsolete as properties. An inverted "position held" property might be helpful (currently we can state: John Doe -> position held -> director. But we can't state: museum -> available position -> director -> position held by -> john doe). I'm not sure about the proper wording ('available position' doesn't sound right). Regards, Christoph Braun (talk) 00:29, 27 November 2013 (UTC)

Structure type conundrum

I was unable to find a way how instance of (P31), P168 (P168) and has use (P366) interrelate with each other, so decided to post a question here. Currently, description for P31 includes: "Use more specific properties when applicable, e.g. occupation (P106) instead of "is a " "; P168 is described as "structure type or intended primary function according to structural characteristics, e.g. bridge, tower"; and P366 is "main use of the subject". Further, P168 has following restrains: the object using it should be an instance of built structure (Q811979) or its subclass, and its value likewise. What do we have in reality:

• Eiffel Tower (Q243) has instance of (P31) = built structure (Q811979), P168 (P168) = lattice tower (Q1440476);
• Forth Bridge (Q275) has instance of (P31) = cantilever bridge (Q1429218), P168 (P168) = both cantilever bridge (Q1429218) and bridge (Q12280);
• Burj Khalifa (Q12495) has instance of (P31) = both skyscraper (Q11303) and tower block (Q18142), has use (P366) = both office building (Q1021645) and hotel (Q27686);
• The Pentagon (Q11208) has instance of (P31) = built structure (Q811979), P168 (P168) = office building (Q1021645);
• Gare de Lyon-Vaise (Q2011) has instance of (P31) = railway station (Q55488), P168 (P168) = railway station (Q55488);
• Père Lachaise (Q176582) has instance of (P31) = metro station (Q928830), P168 (P168) = railway station (Q55488).

So the questions are:

1. Should various kinds of buildings be used via instance of (P31), or should it always equal built structure (Q811979)?
2. Can or should the same value railway station (Q55488) be used in both instance of (P31) and in P168 (P168) at the same time?
3. What traits of the building should be taken into account for P168 (P168) and has use (P366)? If some structural details are necessary for the building to be of some particular use, like office rooms, should it have P168 (P168) = office building (Q1021645) or has use (P366) = office building (Q1021645)? If there are rail tracks that can be used for either railway or metro (but are used only for metro), what property should it have and with what value?
4. I have proposed to introduce a new property with a provisional name "Structural type" (suggestions on a better name are welcome), that would be used to define a particular kind of a building within a larger class basing on is structure, engineering details etc., like lattice tower (Q1440476) and cantilever bridge (Q1429218). The idea is that like "writer" a particular kind of "human" pertaining to his occupation, so a "lattice tower" is a kind of "tower" but pertaining to different aspects than its purpose, use, location etc. In this case, "lattice tower" is a subclass of "tower"; but for example, deep single-vault station (Q2630536) should not to be a subclass of metro station (Q928830), as underground railway stations can also have this structure. So, should P168 (P168) have a narrow value like "lattice tower" or a simple value like "tower"? Should metro stations have both "metro station" and "shallow column station" as values of P168 (P168), or should they be in different properties?

-- YLSS (talk) 13:09, 26 November 2013 (UTC)

IMHO, we should delete P168 (P168) and use only instance of (P31). — Ayack (talk) 14:20, 26 November 2013 (UTC)

(edit conflict)In my opinion:

• What is now P168 (P168) should become instance of (P31). See also this property deletion request.
• The properties that now are marked as instance of (P31), actually belong to the item linked as a subclass of (P279). It would be something like:
  • Eiffel Tower (Q243) is a instance of (P31) = lattice tower (Q1440476), and lattice tower (Q1440476) is a subclass of (P279) = built structure (Q811979);

And answering to your questions:

1. use p31, delete p168
2. delete p168, use p31 with "railway station". An "railway station" is a "sublclass of" "station", and a "station" is a "subclass of" "building" which in turn is a subclass of "architectural structure".
3. property "use" is handy when the use is different than the original one (a station reconverted into a night club)
4. you can do the same with "instance of", like "instance of": "metro station" and "vault station"

I also would recommend to document all this creating a Task force (see example: Wikidata:Physics task force)--Micru (talk) 14:30, 26 November 2013 (UTC)

Well, I don't mind such an option: everything would be simpler. However, how in such a case would a Wikipedia infobox choose one of the values of P31? E.g. ru:Шаблон:Станция метро (Metro station) has a field for "type", i.e. a narrow subclass of underground station, like deep single-vault station (Q2630536). What algorithm to use? YLSS (talk) 16:02, 26 November 2013 (UTC)

Actually you just need "instance of:vault station" since vault station is already a "subclass of:metro station".--Micru (talk) 17:40, 26 November 2013 (UTC)

As I said above, it's not correct to treat e.g. "single-vault station" as a subclass of "metro station": any underground railway station can be classified as such, not only metro. w:Gare de Châtelet – Les Halles, for example, seems to have a shallow column station. So both values should be present simultaneously. YLSS (talk) 19:43, 26 November 2013 (UTC)

Alternatively, you could simply make it subclass of rail station. :) --Izno (talk) 22:27, 27 November 2013 (UTC)

In all probability it should be, and so a station of Moscow Metro could be simultaneously an instance of: metro station (Q928830), deep single-vault station (Q2630536), and possibly also cultural heritage site in Russia (Q8346700). How would an infobox choose only the second one? YLSS (talk) 08:24, 28 November 2013 (UTC)

 Comment The argument above does not seem to look at similar situations across the board and how factors are tied in with specific instances. I believe that the "architectural structure" allows for the simple violation check of "architect", whereas if you are going to promote every structural type to an "instance" status then you start to build the same check across many properties. The reason that it makes sense is that there are numbers of other types of structures that all have elements in common and allow for a reasonable presentation of the components and the violation checks that can be run.

Can I also reflect that the commentary around railway stations has not been made with any view to what is happening in that space where there are at least five different terms used to identify a station where a train stops to collect passengers.  — billinghurst sDrewth 12:42, 28 November 2013 (UTC)

Other-language descriptions

My preferred (and only) language is English, Occasionally, though I have sufficient knowledge to apply a non-English description to an object. How may I do so? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 18:14, 26 November 2013 (UTC)

You can add descriptions and labels in other languages by adding babel to your user page. Other way is to use labellister gadget. --Stryn (talk) 18:24, 26 November 2013 (UTC)

E.g. I have {{#babel:fi|en-2|sv-1|et-0|nb-0}}, so I can add labels and descriptions on those languages without labellister gadget. --Stryn (talk) 18:26, 26 November 2013 (UTC)

Thank you. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 18:37, 26 November 2013 (UTC)

For people who don't want to mess with templates or gadgets or are not logged in - On the top of the page you will see your current set language. Near that are a couple of mysterious symbols which bring you to a menu which lets you change your set language and see the page in a different language. Filceolaire (talk) 12:21, 27 November 2013 (UTC)

Coordinate data and Nearby Pages

Hi all! Is there any plan to put all the geographic coordinate data right here, into the Wikidata, and then make the new "Nearby Pages" just access these information here and bring it to the articles on the many Wikipedia or, even better, into OSM maps also! Once in the Wikidata, projects like OSM could easily access them also! (migrated from here)Sturm (talk) 23:42, 26 November 2013 (UTC)

1. Yes. The plan is to put all the geographic data in wikidata and then make it available to all the WMF projects.
2. No one can really make anyone else do anything round here. By having the data in one place, where it will (we hope ) be checked and maintained and machine readable, we hope that the other WMF projects will choose to use the data from wikidata. This of course includes the 'Nearby' function. Using data from Wikidata 'Nearby' should get extra cool features like being able to filter which articles get shown - just certain languages or just train stations etc.
3. As well as making the geodata available to WMF projects we also want it to be easily usable by external projects such as OSM and Google maps and everybody else. Wikidata is certainly (we hope) designed to make this possible.

Hope this helps. Filceolaire (talk) 12:06, 27 November 2013 (UTC)

Shall We Use Wikidata to Handle Disambiguation

Currently the disambiguation information is only stored by the Wikipedia pages. But this structure will prohibit any automation on disambiguation processing. If Wikidata can store disambiguation information, we can:

1. save all manual work to write something like {{otheruses|...}}
2. reduce duplicated work to update disambiguation pages on each language with the same title (especially for alphabet abbreviations)
3. generate automatic links from an ambiguous named page to the disambiguation page

I propose a structure like this:

• Labels in any language of the same disambiguation page should all use the same/equivalent title
• Original language as a property
• May mean as a property, values are the relevant choices, quantifiers may include:
  • in language (without this quantifier means all linked language has this disambiguation choice, this help to handle special choices not shared while most choices are shared)
  • commonness (a mean to decide when and in what style to generate automatic links)
  • field/subject (many disambiguation group choices according to the related subjects)
• All choices pages may include an automatic link to the disambiguation page, or an "otheruses" template when not more than 2 other choices
• All Wikipedia with the language links will include a generated part, with:
  • links to the choices that the current language has a page for, each followed by the description of the target page, grouped by field/subject and "instance of" values
  • show the original language labels of the choices in brackets as translations when Original language is not the current language

金亦天 (talk) 03:48, 27 November 2013 (UTC)

Interesting proposal for Phase 3 and deep idea! Can you provide some example of future item structure for that case? --Infovarius (talk) 06:50, 29 November 2013 (UTC)

There are 2 kinds of examples that support the automation of disambiguation pages:

1. Widely shared disambiguation pages: an example is abbreviation like ABC(Q286874), it is linked to the disambiguation pages in different languages, but most of them are incomplete, users want to know more still need to refer to the English version. To make the update easy and instant for every language, a shared knowledge source like Wikidata that helps generating the language specific disambiguation pages is necessary.
2. Disambiguation choice page: like w:en:Rapunzel, a choice page usually carry an "otheruses" template, they include some redundant description words when using the template. Automating this kind of disambiguation reminder can save effort and prevents user from forgetting to add "otheruses" in the default named page.

金亦天 (talk) 08:54, 29 November 2013 (UTC)

Seems I was not answering your question, but I am not clear what you are asking for. You want examples of items/pages being applied the new structure or you want more detail about the new structure?金亦天 (talk) 09:32, 29 November 2013 (UTC)

Some problems with numbers

I was looking into some old census-reports of Digerberget (Q15240713), when I discovered something I do not know how to handle correctly here.

The 1960-report tells me that it had 1130 inhabitants. The following report from 1965, tells me that the numbers from 1960 has been revised to 883. The problem is that I do not know if this revision comes from that the method of counting has been revised or if there were something wrong in the counting 1960.

I have seen this several times, more often in more modern reports, but then it is always because the method has been revised. I know I can add both numbers here, and I will, but I do not know which one to prefer when that becomes possible. -- Lavallen (talk) 10:47, 27 November 2013 (UTC)

Usually statistical revisions are due to better data becoming available, better methods being used, or errors being corrected. I'd generally presume the most recently published figure should be preferred, unless there are specific grounds to think otherwise. --Avenue (talk) 11:41, 27 November 2013 (UTC)

The change in method from 2005 to 2010, was how many holiday homes were allowed in a locality. So the 2010 revision of of 2005-census is not necessarly better numbers, it's a new definition. But I guess I have to make the asumption that the 1960-numbers where wrong in this case. In the 1960-census the people who made the counting, could modify their definitions depending on circumstances. That is never done today. -- Lavallen (talk) 16:04, 27 November 2013 (UTC)

Yes, statistical organisations have tightened up on such things over time. The UK now has an 11-page official protocol on revisions, for instance.[2] Ideally any revision of the size you saw in the 1960s figures would be well documented, but I imagine this would be less likely the further back in time one goes. There may also be greater problems locating any such documentation. --Avenue (talk) 21:07, 27 November 2013 (UTC)

Watchlists

Is it possible to make an item on wikidata about user watchlists so all the watchlists for all languages are linked. Since the links of the watchlists are not user dependend it would be possible to link the watchlist pages. So if you are on your English Wathclist page (en:Special:Watchlist) that you can click to your Dutch watchlist (nl:Speciaal:Volglijst) for instance . The thing is that the item Special:Watchlist now links straight to the watchlist on wikidata, so it's not possible to create that item. Sander.v.Ginkel (talk) 15:09, 27 November 2013 (UTC)

The current policy on Wikidata is to not make items for Wikimedians (though I personnaly see only good reasons to do so.) But maybe if we are engouh to think differently we can contest this :) Amongst the reasons : link the user pages, having item to site on commons for artwork authorship when a Wikimedian did it, ... Drawbacks : None. TomT0m (talk) 15:22, 27 November 2013 (UTC)

I can image that the userpage is a bit different from your contribution page and the wachtlist page. Besides that the link is username depended, you can create links on your pages yourself. This is not possible for for instance your watchlist page. I'm actually struggeling all the time finding the foreign watchlists. Sander.v.Ginkel (talk) 15:49, 27 November 2013 (UTC)

Yes and no. It is possible to create an item for Special:Watchlist, however notability 'forbids' the creation of items for Special: pages. John F. Lewis (talk) 17:38, 27 November 2013 (UTC)

Creating a dozen item is not really a big deal. TomT0m (talk) 18:25, 27 November 2013 (UTC)

I am really not sure how a list in the sidebar with hundreds of interwiki links for watchlists (as we do have hundreds of Wikipedias) would help us here … Vogone talk 18:59, 27 November 2013 (UTC)

It does not help Wikidata, but it helps in the crosswiki-work. -- Lavallen (talk) 19:40, 27 November 2013 (UTC)

I doubt this. It is far easier to write the iw prefix + "Special:Watchlist" in the search field than scrolling through a list with over 800 items, in my opinion. Vogone talk 19:51, 27 November 2013 (UTC)

Pretty much what Vogone said. John F. Lewis (talk) 20:01, 27 November 2013 (UTC)

There is already one for Special:RecentChanges, but that is being used to replace iw links that already existed. --Rschen7754 02:52, 28 November 2013 (UTC)

Thanks for that argument. -- Lavallen (talk) 07:43, 28 November 2013 (UTC)

That item is Q6293548. There is one big difference to the watchlist: Any Wikimedia project actually has recent changes, while the vast majority of Wikimedia users will not have a meaningful watchlist in the vast majority of Wikimedia projects. --YMS (talk) 08:20, 28 November 2013 (UTC)

I am concerned though about if we did this. We might wind up with a list of links longer than the length of most peoples' watchlists; it would take a while to load, and be ugly. --Rschen7754 08:57, 29 November 2013 (UTC)

What about this gadget? JAn Dudík (talk) 21:53, 27 November 2013 (UTC)

Well, it works in one project. -- Lavallen (talk) 07:06, 29 November 2013 (UTC)

See also bugzilla:28604. Helder 15:47, 29 November 2013 (UTC)

RfD 2013 archives

The page Wikidata:Requests for deletions/Archive/2013 only lists archives up to the middle of August. Could someone please have a look? It is done by some templates which I did not manage to fix. Thanks in advance.--Ymblanter (talk) 17:16, 27 November 2013 (UTC)

 Done Though it really should be done automatically. Redirects such as this need to be created or it doesn't work properly. Del_♉sion23 (talk) 21:01, 27 November 2013 (UTC)

Thank you. I thought there was a bot doing it, but apparently smth did not work out.--Ymblanter (talk) 21:29, 27 November 2013 (UTC)

Esperanto logo on site

Hi, as I have just mentioned in Esperant Project chat it would by nice to have Esperanto logo on site for Esperanto as user interface language, similar like is for Slovak. Thanks in advance! --KuboF (talk) 21:26, 27 November 2013 (UTC)

Problems with Open Library ID

An identifier should identify a person. Property:P648 (Open Library ID) does not fulfill this demand:

1. Abba Eban (Q305776): OL2686699A
2. Dean Koontz (Q272076): OL7119293A

No. 1 (Abba Eban) is just one of six author pages with that name to be found in OL. No. 2 (Dean Koontz, b. 1945) is a person but the OL page only provides the reader with a copy of the Wikipedia article and "0 works". -- Tom, 28 November 2013

Regarding (1): you can either correct the id or correct the info on OL. Regarding (2): OL is like Wikipedia, that it lists 0 works doesn't mean that an author has 0 works, it means that there are 0 works recorded and you can add the missing ones.--Micru (talk) 12:10, 28 November 2013 (UTC)

I added this information. Search powered by Openlibrary. Site gave one identifier. In Freebase has links to several (up to 6) identifiers. It makes sense to add it?--Пробегающий (talk) 13:15, 28 November 2013 (UTC)

The main Open Library author page for Abba Eban (Q305776) is:

OL399071A: Abba Solomon Eban (1915-2002)

Other pages like OL2686699A, OL3417217A, or OL6185545A are what VIAF calls "undifferentiated" (see also: Property talk:P227#Usage note). About Freebase: I would only add well selected links. A page that is 100 % about the item (person) and provides valuable information. Freebase is an automated collection of data (harvesting) and does not provide authority files like GND or LCCN. --Kolja21 (talk) 16:57, 28 November 2013 (UTC)

For Dean Koontz (Q272076) in Freebase 2 ID from Open Library, in OpenLibrary 9 ID. All redirect to OL7119293A. Note: errors may be on any sources.--Пробегающий (talk) 19:14, 28 November 2013 (UTC)

Verified statements

Following up a discussion on the books task force, I would like to ask if there would be interest in discussing a system for verifying statements similar to the one used in some en-wp infoboxes (example: w:MPP+). It could be thought as a property automatically added by a gadget (verified by:username as string), or as part of the system (it could take longer). The meaning of such verification would be that a user has checked that statement and source match (not the case of many VIAF statements), and it would help focus efforts. Now there is no way of conveying that a statement has been verified, so we are checking once, twice, or more times the same statement without conveying that we have quality data.--Micru (talk) 12:52, 28 November 2013 (UTC)

Yes, absolutely.--Ymblanter (talk) 13:03, 28 November 2013 (UTC)

A 'checked by' property with a user name as the object would be ideal except that isn't one of the datatypes. String datatype instead? Filceolaire (talk) 14:59, 28 November 2013 (UTC)

Wikidata-link to the user name and the second line the date of checks.--Пробегающий (talk) 15:10, 28 November 2013 (UTC)

Yes, it can be done with a string that links to the userpage as we do with other properties. If someone makes a gadget, it can add "point in time" too. When or if the datatype "mediawiki user" is available, it can be moved easily. I have started the property proposal "checked by".--Micru (talk) 15:29, 28 November 2013 (UTC)

I like the idea, but I would prefer a software solution like in ns:Page in Wikisource. -- Lavallen (talk) 15:37, 28 November 2013 (UTC)

This is a cool idea. We will need some way to identify people anyway for Commons for example. However I am currently not sure if this should be a separate datatype (person) or just use URL. (URL because we will need to link to people on-wiki and off-wiki. --Lydia Pintscher (WMDE) (talk) 16:18, 28 November 2013 (UTC)

Urls are probably OK, but this does not allow to show information from Wikidata about the person from it, as we need internaly an item and we cannot make queries for datas outside Wikidata. If we want to store a name or an occupation, we will need an item, especially if we want to show these informations. Is it really different to create an item an associate the URL to this item ? TomT0m (talk) 17:06, 28 November 2013 (UTC)

Both would work. However keep in mind that for the same thing you'll probably at some point (latest when commons gets real) use this for people here on-wiki as well as outside. Imagine for example a creator property on Commons. Now you could create items for all of them but this seems a bit cumbersome to me. --Lydia Pintscher (WMDE) (talk) 22:42, 29 November 2013 (UTC)

@Lydia Pintscher (WMDE): To me, it's like saying a drop in the ocean is cumbersome :) If we can't manage that without overhead we have big problems imho. Just by seeing the over side, I see plenty of applications, a Wikimedian would not have to put its Babel boxes on any of its personal page, Interwiki in its accounts managed, and some users likes to have a clean personal page to demonstrate their techical skills, it can be a playground to learn Wikidata. We just should have a policy exactly as the user pages : the user is responsible for its Wikidata datas as he is for its personal page. TomT0m (talk) 11:29, 30 November 2013 (UTC)

Hi folks, I'm a strong supporter of this kind of initiatives (@Snipre: can confirm :)) What about storing the datas as statements or qualifier such as Verified by or source annotations ? TomT0m (talk) 17:06, 28 November 2013 (UTC)

I will play the role of the bad guy: not really a good thing because the problem exposed by @Micru: is not a problem if the source is correctly stated. This is in my opinion just a way to avoid sourcing correctly. The only situation where this can be useful is when entering statement about identifiers: but for the cases I know like the case of Chemspider, it is possible to source with stated in (P248) and retrieved (P813). For a example look at ethanol (Q153) for the statement about PubChem CID (P662). Snipre (talk) 18:24, 28 November 2013 (UTC)

Not only identifiers, also imported data, and imported sources. In the ideal world we would have all the sources at hand and we could check that the information corresponds with what the source says, but in reality we have many sources imported by users who didn't have access to the source themselves, so this can add extra confidence. "Date retrieved" doesn't add that extra quality seal because on the first sight on PubChem CID (P662) it cannot be seen who retrieved it, a bot parsing or cross-mapping millions of registers and maybe prone to errors or the careful @Snipre:? Additionally, if we had a visual indication that some check has been done (which I doubt it could be done relying only on p813), the effort could be focused on the ones that are still pending.--Micru (talk) 23:56, 28 November 2013 (UTC)

@@Micru: What you are proposing is a new concept: entering data with source without checking them. Until now the principle was that the contributor who enters data is "responsible" of the data he enters and checks them. With your new property you define a new concept: a data is true not when entered with its source but when confirmed by a second user. This the concept of the german wikipedia where each modification has to checked by a confirmed user to be displayed by the system (flagged revision).

I am not oppose to that new principle but this can't be done just with a new property: this is a radical change of the management of the contributions. To be efficient we need an interface which can do the job and a decision of the community. This adds a new level of administration work. Snipre (talk) 11:29, 29 November 2013 (UTC)

TBH, what worries me the most at the moment are bot-imported identifiers. We have plenty of them already and we should take steps to increase the trust on the data so that at least those identifiers can be used confidently *now* by Wikipedias. Why should they switch to Wikidata if the incentive is not high enough or even against it in some cases like the Chembox? The extra effort to validate that data doesn't worry much since we are already doing it, but without keeping track of those efforts. Think too that this effort can be outsourced to wikipedias, where users could have the option to click to add a "checked by" automatically here in Wikidata, thus spreading this info to other Wikipedias.

Even being hard to assess with no statistics (clicks per identifier, corrections, etc), I presume that the sum of the time spent by each individual user checking independently the pages linked is much greater than the time it would take one single user to check the statement and mark it with a gadget as "checked". An integrated flagging system would be ideal, however since that would take development resources and it is not a priority at the moment, I am not sure if there is much use in discussing it now... discussions too waste resources if from the beginning it is clear that they cannot lead to any useful outcome.--Micru (talk) 12:27, 29 November 2013 (UTC)

Just to recall Wikidata is not about Truth ... The only thing in Wikidata itself is the difference between a claim, a statement, and ranks. TomT0m (talk) 12:29, 29 November 2013 (UTC)

@Micru:. Instead of creating a new system why not use the existing system: if you check the identifiers this means you have a source so why not add the source used to check the identifier ?

If you have a CAS number or a PubChem ID why is it not possible to add the source used by you, me or someone else to check that the imported value is correct ? If I use a book, I can put the reference of the book, if I use a website, I can use the URL, if I use an online database, I can use an URL or the item of the database to specify the source use to check: no need of a new property "check by". And finally to measure the progress of the check we can juts delete the "imported from" property each time we check the value. For me the property "imported from" should disappear with the time and the use of this property should be restricted and banned in a near future. I really don't understand the need of a new tool as we have already everything. And in the case of the en:WP with their infobox and their system of "checked" this is just a bad way to avoid to source correctly their data.

And as I worked a little on identifiers, most of them can be extracted automatically from the database website because there are free and open data. We just have to organize the extraction. Snipre (talk) 00:29, 30 November 2013 (UTC)

@Snipre:: I agree that "imported from" should stop being used as a source property, but many people disagree, so most likely it will continue existing unless a distinction is made in Wikidata repo between sources and provenance. If at least the statements that only have "imported from" as a source were displayed as "0 sources" that would be IMHO an advance, since I wouldn't have to click on each source to make sure that "imported from", also known as the "incarnation of sources distrust", is not lurking there.

As for organizing data extraction I think at some point we'll have to document the mappings to external data as it is done now in dbpedia, but for that we would need a similar extraction framework to the one they have... I'm sure @Lydia Pintscher (WMDE): would be glad to hear more requests :P--Micru (talk) 18:39, 30 November 2013 (UTC)

It is quite justified to worry about bot runs. I have just been looking at a few entries which were created twice at the same Wikipedia by different bots. Elsewhere, there were entries which were correctly copied from a database twice (once in the wrong spelling, once in the right spelling). I agree that "imported from" should be deprecated. Personally, if I add a source, I almost invariably will have seen the source but in a Wikipedia anything is possible ... - Brya (talk) 18:01, 30 November 2013 (UTC)

Sports results

The inclusion of sports results has been raised a few times, such as Wikidata:Project_chat/Archive/2013/04#Election, Sport and other game results in Wikidata => Timetable? with Kahlores and Kompakt. I have started Wikidata:Sports results task_force to bring these discussions together. Ideally we would be using authoritative sources. I have access to Olympic Data Feed, which might be a legal mess, but I am sure we obtain well described metadata from other sports federations that have a more open data licensing mentality. I expect an RFC will be needed before any large scale implementation occurs, and we may need to wait for technology advances. Anyway, I have had a play with this area and noted my results at Wikidata_talk:Sport_results_task_force#Medals. I would appreciate some feedback on this general direction. John Vandenberg (talk) 14:25, 28 November 2013 (UTC)

Citations

Hi, just discovered this place, I like how it sparkles. ;) Can somebody point me towards information on what has been done so far on integrating bibliographical items, citations and quotations, and whatever is required to make it available to client wikis? Also, is there a way to query Wikidata from my local wiki? Paradoctor (talk) 16:27, 28 November 2013 (UTC)

Hi Paradoctor, there is a Wikidata:Books task force. More about citations you will find at Help:Sources. --Kolja21 (talk) 17:27, 28 November 2013 (UTC)

Thanks. Paradoctor (talk) 17:47, 28 November 2013 (UTC)

To your second question, mw:Extension:Wikibase_Client might be a good answer. πr² (t • c) 06:36, 29 November 2013 (UTC)

Sadly not. I wanted to access Wikidata from my personal wiki, just like you can do with Commons files, but that's not yet available. If you want it done right... ;) Thanks for the info, though. Paradoctor (talk) 00:56, 30 November 2013 (UTC)

P131 and sources

We are supposed to use the smallest administrative unit when we add located in the administrative territorial entity (P131), but that is not always easy to add a source for.

The smallest administrative units in Sweden are the parishes. But they will be deprecated any minute, therefor have I used municipalities instead here. Soon will a new division, made up by districts come instead. We do not know exactly how it will look like until 2016.

Until now, have the reports from Statistics Sweden I have been using, not told me anything about which parish the items are located inside. They have only told me about which municipality. Not a big problem, since parishes are deprecated today, as I told above. But how are we supposed to do when a source tells me something is located in a municipality, but the Property asks me to look for an even smaller administrative unit? P131 is often considered as a trivial piece of information that you do not need to add sources for, but in reality, I would disagree in many cases. To know which parish I live in, I have to look into the title (Q13629195), since the border between the parishes here looks like the borders in the area around Baarle-Hertog (Q244959). Maps are good to use, but they are not always reliable. -- Lavallen (talk) 09:10, 29 November 2013 (UTC)

If I got it right, we are waiting for the new administrative division. In this case, I would not any info which is imminent to be changed, and just wait until we can reliable source it.--Ymblanter (talk) 21:05, 29 November 2013 (UTC)

User:Lavallen: I think that would be a good case in which time qualifiers would be helpful. I wouldn't delete the statements working with parishes, but simply give them and end date. --Tobias1984 (talk) 17:18, 30 November 2013 (UTC)

• I agree with Tobias; give them all end dates, or if you don't know exact end dates, put "point in time => 2013" so someone will look into it in the future. --Arctic.gnome (talk) 04:48, 1 December 2013 (UTC)

As I said above, I do not have much sources for which place is located in which parish. I only find them in OSM, but only in areas where there are devoted OSM-users.

One problem is also that I do not know which enddate is valid. The parishes became defunct as a govermental geographic division as of 1999, without being replaced by districts until 2016. But they still exists as the smallest organisational division of the Lutheran church of Sweden. From 2000 the Lutheran church have been dramaticly reorganised, and maybe only a third of those parishes still exists. -- Lavallen (talk) 09:45, 1 December 2013 (UTC)

Dates in Julian / Gregorian calendar

I want to refresh and extend Denny's address at User talk:Kizar#Dates in Julian / Gregorian calendar. Here are some more bots that have erroneously inserted dates in Julian calendar to the (proleptic Gregorian) data type properties:

• Ladsgroup's Dexbot: e.g. [3], [4],
• Ivan A. Krestinin's KrBot: e.g. [5],
• Viscontino's ViscoBot: e.g. [6], [7],
• Kizar's KLBot2: e.g. [8], [9].

Since I have found them by unsystematic experimenting there are probably more false dates by other bots or human users. How can we detect and correct them? I am quite sure that there are too much of them for a quick manual correction. Is there an other possibility than rivising all dates before 1923 when Greece adopted the Gregorian calendar? And have all the bots stopped these edits yet? (Dexbot at least has.) Best, --Marsupium (talk) 17:56, 29 November 2013 (UTC)

calendarmodel display

BTW: I would appreciate if the bots did not set a value for the calendarmodel displayed other than "auto" or even removed those yet setted. I think the indication of the display is perhaps not useful at all. The way some bots use it is for sure meaningless, for example in the statement <Leo I (Q183776)> date of death (P570) <+00000000474-01-18T00:00:00ZGregorian> added by User:KLBot2. The indication of the Gregorian display for a date nobody has given in Gregorian calendar, neither when Leo I (Q183776) lived, nor today, seems completely useless to me. --Marsupium (talk) 17:56, 29 November 2013 (UTC)

Interwiki section links

Hi. I ran into the problem of Interwiki #section links not working, and eventually found this discussion: Wikidata:Interwiki conflicts#Interwikis where there is no one-to-one match.

One option mentioned there (by User:Littledogboy) seems interesting, and I would like to discuss it further: Creating a redirect at the foreign language that only has a subsection mention.

E.g. The English article en:List of classical music composers by era (a graphical timeline) interwiki-links to the German subsection de:Komponist#Komponisten (Klassik) - Grafische Übersicht (a graphical timeline). Could/should I create a redirect at de:Komponisten (Klassik) - Grafische Übersicht (or something), so that the German content can be indexed in Wikidata?

What use-cases, or policies/guidelines, would conflict with this solution? (It's new territory for me, so I might be overlooking something obvious).

FAQ: Please, whatever the solution we come out with, add something to Help:FAQ (and/or the top of Wikidata:Interwiki conflicts), because it took a while for me to uncover, and is probably a common problem. Thanks! Quiddity (talk) 18:20, 29 November 2013 (UTC)

1. The software do not fully support sitelinks to redirects yet. (You have to temporary make it a "real" article first to be able to create the sitelink.)

2. The information in the Wikidata-item will not be linked to the de-article "Komponist", only to the redirect, where templates and modules cannot be added.

3. This solution is maybe not backward-compatible with the future solution the developers will create.

It the purpose only is to create interwiki, I would recomend you to wait. -- Lavallen (talk) 19:01, 29 November 2013 (UTC)

I agree with Lavallen, too. It's one of the biggest technical issues we have currently. --Ricordisamoa 19:34, 29 November 2013 (UTC)

Thanks for the insights. Could someone add 1 sentence explaining that, at the top of Wikidata:Interwiki conflicts and/or the bottom of Help:FAQ? Please and thank you! Quiddity (talk) 18:42, 30 November 2013 (UTC)

wikidata coordinates database?

Is there regularly updated Wikidata coordinates database on toollabs? Something like ghel coordinate databases with wikipedia coordinates on toolserver. --WikedKentaur (talk) 07:38, 30 November 2013 (UTC)

Maybe this and this will help. --LBE (talk) 08:34, 30 November 2013 (UTC)

Handicap physique (fr) /Körperbehinderung (de)

Hi, I need some help : I try to link the french article Handicap physique to the german one Körperbehinderung, the subject is the same, but there is seemingly a problem because each article is linked to others (same subject) with different item numbers. Thanks for your advice and help. --Franz53sda (talk) 19:28, 30 November 2013 (UTC)

I agree and have  Merged items Q3126720 and Q1179623. Littledogboy (talk) 23:09, 30 November 2013 (UTC)

File replacements and global usage

I struck an issue that there was a file in use at roWP that was being replaced from gif to png at Commons. After large amounts of wading through foreign language templates (always a pain), I found that the image was called at roWP through a property call to WD. So, okay, I have updated WD to the change the image, however this sort of issue is going to become worse and worse. The standard file replacement processes at Commons are not going to work, and there is no ready identifier at Commons to direct users that they need to look at WD. We clearly need to have a means that a file reference at WD shows up somehow, I am just thinking that surely that it could be referenced in Global Usage, eg. Special:GlobalUsage/Slavutich_prapor.png has a WD linkage somehow.  — billinghurst sDrewth 01:45, 1 December 2013 (UTC)

@Lydia Pintscher (WMDE): This sounds like something that needs to be handled from a technical side. --Izno (talk) 02:49, 1 December 2013 (UTC)

Yes indeed. We have bugzilla:46358 for that which could use some input from you. --Lydia Pintscher (WMDE) (talk) 08:31, 1 December 2013 (UTC)

@billinghurst: For the future, for techy sounding problems, you should probably leave a comment to WD:Contact the development team. --Izno (talk) 02:49, 1 December 2013 (UTC)

A soultion on the Commons end I can imagine is to update CommonsDelinker and other file replacing tools like Fastily's tool to also replace files in WD items. I have no idea how easy this is.--Ymblanter (talk) 09:18, 1 December 2013 (UTC)

From the Commons side there it knows nothing about WD, except for the rare times we physically link an image. So until the linkage issue is resolved, we cannot even talk to Commons about how we handle replacements. Even then there will be issues as it will be specific cases where it occurs and where it needs fixing. There will be a significant discussion at that time as there is complexity and xwiki progress in play here.  — billinghurst sDrewth 10:22, 1 December 2013 (UTC)

Top of the subclass tree

What should go at the top of the subclass tree? I've been using Q15244161, which in retrospect might not be the best option. User:Emw replaced it with no value on one page, but that seems to break Template:tree, so that option is no good unless that bug is fixed. Should we just not have anything at the top? Follow-up question: If we don't have any item at the top, what items should we allow as property constraints of subclass of (P279)? As of now, the items at the top of the subclass trees appear to be entity (Q35120), space (Q107), mind (Q450), principle (Q211364), and physical phenomenon (Q15257329). That list covers everything I can think of. --Arctic.gnome (talk) 04:43, 1 December 2013 (UTC)

• I don't think it is for us to decide how the upper parts of the subclass of (P279)-tree looks like. We just have to add the values that different sources give. Probably different sources will generate different trees that might even contradict each other. Another problem is that we might end up having items like "Universe", where even philosophers can't agree on the exact meaning. This will still take a lot of work and thought on our part, on how to model that. --Tobias1984 (talk) 12:42, 1 December 2013 (UTC)

I don't think we should punt on this. Without a sound and stable upper ontology, subclass of (P279) would be drastically less useful than it should be. As I explain below, entity (Q35120) is the standard root node for major, broad ontologies. Wikidata is in the same boat: many contributors are interested in their specific domains, but at the end of the day these domains must be coherent to each other. Having a consensus root node for subclass of (P279) is a basic requirement for that. Emw (talk) 13:46, 1 December 2013 (UTC)

• I've been cultivating the top of the subclass tree since subclass of (P279) was created in March, but I've never given a decent explanation of why I've been using entity (Q35120) (alias: thing) as the root node of Wikidata's type hierarchy. The reason is convention. General-purpose ontologies typically root their taxonomic hierarchy at an item labeled "Entity" or "Thing". (The Thing (Q11039766) is a Wikipedia disambiguation page and the corresponding Wikipedia page for entity (Q35120) describes the concept in rough ontological terms, so I went with the latter and made "thing" an alias of it.) Examples of ontologies rooted at such an item are any ontologies built with OWL, SUMO, Cyc, and upper ontologies presented in widely-used textbooks like Artificial Intelligence: A Modern Approach, 3rd edition (see figure 12.1 on page 438).

Shared vocabulary and interoperability are important, so since other major ontologies use it there, I think keeping entity (Q35120) as the root node of our class hierarchy makes sense. The basic idea of a root class is that it is something that everything else can be said to be a more specific type of. This criterion easily disqualifies space (Q107), mind (Q450), principle (Q211364) and physical phenomenon (Q15257329) as root nodes: one can imagine an item that is not a subclass of each. On the other hand, it is clear that 'space', 'mind', 'principle' and 'physical phenomenon' are all entities, i.e. things.

(To answer Arctic's operational questions: "Should we just not have anything at the top?" With the possible exception discussed below, all items should be linked to entity (Q35120) by either P279 or P31. entity (Q35120) should have no P279 claims, nor any P31 claims. "What items should we allow as property constraints of subclass of (P279)?" Only entity (Q35120), with the possible exception discussed below, per the rationale in this post.)

While I think having "entity" and not something like "mind" as a general-purpose root node is uncontroversial, there are important open questions about related aspects of the top of our so-called subclass tree. I think the most pressing question is how we should root fictional entities. One of the soft conclusions of that linked discussion is that we would use fictional entity (Q14897293) as the root for things in fictional universes (e.g. the Harry Potter universe, the Star Wars universe). In this case, Wikidata would have two hierarchies: one for things that exist in the real universe, and one for things that exist in any non-real universes. I encourage anyone interested in this to read through the previous discussion on fictional entities. Emw (talk) 13:25, 1 December 2013 (UTC)

Items of "project pages" and "project pages" of Wikidata

Wikidata items collect also "project pages" as templates and categories of Wikipedia, Wikivoyage and Commons (at now available). What about Wikidata templates and categories? For example Template:Archive box (Q5828032) collects the interlinks of archive box templates of 71 Wikipedia and 2 Wikivoyage projects, but not our own template. In these cases could be useful add also Wikidata project pages? --Paperoastro (talk) 12:02, 1 December 2013 (UTC)

bugzilla:55570 and Wikidata:Contact_the_development_team/Archive/2013/10#Links to Wikidata pages in items. Matěj Suchánek (talk) 12:22, 1 December 2013 (UTC)