Shortcut: WD:XLINK

Wikidata:WikiProject Cross Items Interwikis

From Wikidata
Jump to navigation Jump to search

Other languages:
العربية • ‎Deutsch • ‎English • ‎français • ‎русский • ‎svenska

The goal of this project is to coordinate the efforts, code, and ideas about interwikis when the corresponding article in a former Wikipedia is not linked to the same item on the local wiki.

This is necessary because Wikidata has a different interwiki granularity than pre-wikidata language links : on Wikidata, one concept = one item, whereas articles could be linked to articles with a close but not identical subject

For example, there is an item for Bonnie and Clyde (Q219937), and two for Bonnie Parker (Q2319886) View with Reasonator View with SQID and Clyde Barrow (Q3320282) View with Reasonator View with SQID. Some Wikipedias have an article about the duo, others have articles for the two members of the duo.

This means that using the sitelinks on the wikidata item associated with a Wiki article will not find all the relevant sitelinks - Wikipedias with articles on Bonny Parker won't have links to wikipedias with articles on Bonny and Clyde and vice versa.

Ideas [edit]

The current state of the global reflexion is that interwikis will be old style interwikis generated automatically with a template, which will inspect the content of items to find relevant articles in the other languages.

experiments
here (Raven (Poe)) by Innocent bystander, with the module : [1]

How to find relevant interwikis[edit]

Collection of problems[edit]

  1. Bonnie and Clyde : articles about groups, that also discuss the various parts of the group. Wikidata needs separate items for each part as well as an item for the whole. Wikipedias either have an article about the duo or separate articles about each of the two. We can use redirects to add links to the 'Bonnie and Clyde' article from the 'Bonny Parker' article but how do we add links to the 'Bonnie Parker' and the 'Clyde Barrow' articles from the 'Bonnie and Clyde' articles?
  2. Other whole/part splits.. Where wikipedias have an article about a town that also covers the same name district surrounding the town. An article about a genus and the species that make up that species. An article about a band and the band members. An article about a book and about the characters in the book.
  3. Fruits / taxons
  4. Conflicting definitions : North America
  5. FRBR Editions (at Wikisource) with Works (at Wikisource or Wikipedia)
  6. Encyclopedic articles (at Wikisource) with main topics (at Wikipedias)
  7. Wikipedias with two articles in two different languages. Example: Q12715471 (nynorsk) and Q25164 (högnorsk) on nnwiki.
    ... or even two articles in different scripts of the same language. Example: Q333 (lad:Astronomiya in Ladino, or Judeo-Spanish, Latin alphabet) and lad:אסטרונומייה (in Ladino in Hebrew alphabet, now at Q20819872).
  8. Wikisource-pages embedded in Wikipedias, like frr:Text:Hoodsid and als:Text:Houptsyte
  9. Forenames with multiple slightly different spellings
  10. Chemical elements : definition as pure substance chemical substance (Q79529) View with Reasonator View with SQID versus definition as types of atoms (fr, de, ...). linked by :
    < pure hydrogen substance > has part (P527) View with SQID < hydrogen atoms >
    .
    also stereoisomeres, racemats, salts, and tautomeric mixtures
  11. Company item mixed with the sofware they edited item, example : WOT Services (Q1391304) and WOT (Q21126538)
  12. WikiProject Taxonomy : monotypic taxon (Q310890) View with Reasonator View with SQID instances : taxons that have unique (immediate) lower rank taxon son. They have all the same instances, so they could be in another classification system considered as a unique class. Wikipedias can have articles about an arbitrary rank taxon. A testcase :
    A good test case is the taxonomic hierarchy: order Amborellales – family Amborellaceae – genus Amborella – species Amborella trichopoda. The order, family and genus are monotypic. The English Wikipedia has one article at the genus; the German Wikipedia has one article at the species; the French Wikipedia has articles at the order, family and species; the Spanish Wikipedia has one article at the family; and so on. At present these aren't properly connected. There needs to be a proper solution. Over to you guys, as I don't work here. 86.113.153.215 13:24, 4 May 2016 (UTC)
  13. Regular transport route, one specific instance of it (e.g. when there was a crash), the vehicle operating the specific instance. en.wp plane crash articles are commonly about all three, but wikidata can have three items - which gets the interwiki links?
  14. Production process / product. Example : aromatised wine (Q19903355) View with Reasonator View with SQID and flavored wine cooking (Q926736) View with Reasonator View with SQID.
  15. Item_A/list about item_A; item_B/list about Q5 + qualifier P39=item_B


Collection of ideas on how to solve[edit]

(if solution specific to a problem in the previous section, please precise the number)

  1. Well, we have some properties that can be considered as child-to-parent ones. They obviously need to be directional, acyclic, and preferably single-value:
    One can use those properties to obtain "parent" item and insert it's interwiki into current page (via standard interwiki syntax). -- Vlsergey (talk) 17:54, 16 June 2015 (UTC)
    This can help with the whole/part splits - to add links back to the articles on the whole from the articles on each part (though this can also be done using wikipedia redirects). Filceolaire (talk) 12:53, 18 June 2015 (UTC)
  2. To links from the articles on the whole (book/band/genus) to articles in other languages on each of the parts (character/band member/species) I think the model should be to
    • add an info box for each part to the article for the whole - to be located in the section for that part
    • This infobox should include a drop down list for other languages, revealing links to articles in other languages related to that part. Filceolaire (talk) 12:53, 18 June 2015 (UTC)
  3. An alternative way to link from the whole to the parts is for the article about the whole to have a table with a line for each part (rather than a text article with a section for each part). Wikidata info is used to populate all the table cells/fields and the drop down 'languages' links then go in the last field of the table. This is more suitable for things like squad lists for sports teams. Filceolaire (talk) 12:53, 18 June 2015 (UTC)
  4. Domback (Q1879056) is an example of a multi issue item. See sv:Dombäck how that is solved in the infobox today. Since (at least) I prefer to have one statistical entity in each item, Talk:Q14629121 now shows how the statitics from "Dombäck north" and "Dombäck south" could be added to one single article. -- Innocent bystander (talk) 13:44, 18 June 2015 (UTC)
  5. Innocent bystander developed a great solution for item #7 above, and it turns out to have some real utility for part of the Bonnie and Clyde problem, too.
    1. The key to solving the problem is in a Lua module he developed, which is installed in two places so far (Module:Interwiki (Q20819069)). Innocent bystander developed this in part for Wikisource, and I can't fully speak to how it works there. With respect to Wikipedia, once that module is deployed in a given Wikipedia, though, here's what you can do:
      • The answer to what the module do on Wikisource is "Nothing". Wikisource got Arbitrary access before Wikipedia and was therefor a good place to test some features. The sv.Wikisource-module is in fact located in a Sandbox. -- User:Innocent bystander
    2. To address item #7 above, let's say that I have to link two articles through Wikidata, lad:Astronomiya and lad:אסטרונומיה. These are essentially the same article, one written in Latin-script Ladino, and one written in Hebrew-script Ladino. Only one can be directly linked through Wikidata, and by convention we're linking the Latin-script ones. Hence only the Latin-script one is included at astronomy (Q333), while we had to create an alternate Wikidata entry for the Hebrew script one at no label (Q20819872). In the Wikidata item for the Hebrew script one, I include [[P:460]] permanent duplicated item (P2959) with the value of Q333. Then, if I go on the Hebrew-script page and add {{#invoke:Interwiki|interwiki}}, all the Wikidata links from Q333 appear on the Hebrew-script page, too. Obviously, it's only a one way solution. If you click on any interwiki link there, then clicking back to Ladino takes you back to the Latin-script page, not the Hebrew-script page. But since all pages on ladwiki with dual versions have tabbed links to each other, it only takes a second click to complete the round-trip.
    3. It turns out, though, that there is more to it than that. There's a pretty common template on Wikipedias, Template:InterWiki (Q6275256). For some reason the ladwiki version was only connected manually, so I connected it through Wikidata instead. But there's a similar template (Template:InterwikiCountry (Q20819350)), intended to go on country articles instead of language articles, that only appears on ladwiki and eswiki. (Since Ladino, or Judeo-Spanish, is largely built from a medieval version of Spanish, eswiki often turns out to the be the source of much of the infrastructure of ladwiki.) In both wikis, there were a lot of interwiki links—and in both cases, the links were manual, and to items in Q6275256. So the first thing I did was link them to each other, hence Q20819350. After that, though, I tried adding back the rest of the Interwiki links to 6275256 by using the module. Here, I used a syntax not depending on P460; instead, I used {{#invoke:Interwiki|interwiki|qid=Qnnnnn}}, where here "Qnnnnn" was Q6275256. And it worked: On the ladino version of this template, all the links go to templates on Q6275256 except eswiki, which links normally to the counterpart at Q20819350.
      So what does this mean? If a wiki has a Bonnie Parker article and a Clyde Barrow article, and you have the module installed, then if you put {{#template:Interwiki|interwiki|Q219937}} on each of those two pages, here's what will happen: the iwlink to any Wikipedia with a separate article will be the link to the separate article. But the iw link to any Wikipedia without separate articles, but with the combined article, will be the link to the combined article.
      In principle you could use it the other way, too, but then you'd have to pick whether to point to Bonnie or Clyde, which isn't really something that can be solved by software. But the beauty of this solution is that it doesn't even depend on assigning any particular properties or hierarchy to Bonnie, Clyde or Bonnie-and-Clyde. It only requires you to decide that the "fallback" iwlink source will be Bonnie-and-Clyde, and it works. StevenJ81 (talk) 20:49, 16 August 2015 (UTC)
  6. Problem 12 (monotypic taxons) : To link with a parent taxon article that could exist on another language wp : follow Parent taxon search and load the item. To go from a parent taxon to his son : maybe union of (P2737) View with SQID and disjoint union of (P2738) View with SQID could be serve as inverse properties. TBD with WikiProject Taxonomy. author  TomT0m / talk page 10:46, 5 May 2016 (UTC)
  7. Problem 14 (production / product) they are linked throug produce search. Maybe there is a need for an inverse property. author  TomT0m / talk page 17:46, 11 September 2016 (UTC)
  8. ... add yours here

Possible blockers[edit]

We can go from the the subject to the object of a property, it's impossible to do the inverse. It could be a problem for example in the monotypic genuses (genus with only one species) with current model : If we have

< species > parent taxon (P171) View with SQID < genus >

and wiki A has an article about the species and wiki B has an article about the genus, then we will be able to generate a link via lua from wiki A to wiki B, but not from wiki A to wiki B.

Workflow[edit]

It could be cool if we made all experiments here, then find a way to propagate the changes to the collaborating Wikipedia to centralize the efforts.

modules[edit]

Code: Module:Interwiki (Q15727702) View with Reasonator View with SQID (get interwikis with said to be the same as (P460) View with SQID)

Templates :

Groups[edit]

Hello. 1986–87 Cypriot Fourth Division (Q17436662) was split into three groups. We have three winners winner (P1346), one for each group. And we have teams that were promoted (P2881) from each group and teams that were relegated (P2882) from each group. Which qualifier can I use to show for example that Libanos Kormakiti (Q18228530) was the winner of Nicosia-Keryneia Group? (I don't want to have 3 different Wikidata pages since in Wikipedia we used to have only one article with all the groups and not to separate articles, one for each group). Even the Cyprus Football Association referred to 1986–87 Cypriot Fourth Division (Q17436662) as a single championship, not 3 separates. I don't think there will be a Wikipedia that will create 3 different articles, one per group. Xaris333 (talk) 17:37, 7 June 2016 (UTC)

Participants[edit]

[+] Add yourself to the list

The participants listed below can be notified using the following template in discussions:

{{Ping project|Cross Items Interwikis}}