Page MenuHomePhabricator

Make an integrated tool for convenient mass translation of Wikidata labels
Closed, DuplicatePublic


This is a big feature request :)

The Wikimedia world has a convenient tool for translating user interface strings of MediaWiki core and extension, and also wiki pages: the Translate extension. It's used by thousands of Wikipedians effectively every day to translate new features and community pages in Meta,, etc.

Unfortunately, there is no such robust tool for translating labels and descriptions of Wikidata items and properties, even though it would be very useful. The current ways to translate them are:

  • To open each item or property page in a separate browser tab. This is very different from how the Translate extension works: it shows all the translatable strings grouped in a list on one web page, automatically progresses to the next translation after a translation is published, and shows translation tools for each translatable item (machine translation, translation memory, documentation, etc.)
  • To use tools such as Terminator or Tabernacle. They are supposed to provide similar functionality. Unfortunately, they have several disadvantages: They are not localized the way MediaWiki extensions are; they are often unstable; and they require a lot of specialized knowledge in queries, instead of organizing the knowledge in easily selectable groups, as the Translate extension does.

Having such a tool, which would be robust and properly integrated with Wikibase will go a long way towards the following goals:

  • Attracting new multilingual contributors to Wikidata and other Wikibase sites.
  • Making it possible to show more diverse information in tools such as Article Placeholder and in Wikidata-connected templates.

Another thing to note is that labels and descriptions are usually short and could be translated on mobile devices quite easily. (Unfortunately, the Translate extension doesn't support mobile devices well itself, and of course it should.)

Of course, the nature of Wikidata labels is significantly different from the nature of user interface strings in MediaWiki extensions: Wikidata labels are much more numerous, new labels are created much more quickly than UI strings, and their topics is much more varied. Nonetheless, it would be useful to reuse the experience that Wikimedians have from translating UI strings at least partly, and perhaps to reuse at least some parts of the Translate extension itself.

Event Timeline

Amire80 created this task.Dec 19 2018, 6:00 PM
Restricted Application added a subscriber: Aklapper. · View Herald TranscriptDec 19 2018, 6:00 PM
Amire80 updated the task description. (Show Details)Dec 19 2018, 6:03 PM
Ijon added a subscriber: Ijon.EditedDec 23 2018, 5:03 PM

Experience suggests that stabilizing the @Magnus tools would be a likelier path than getting WMF to allocate feature development resources.

AFAIK Ops has been working with Magnus to stabilize the VM for PetScan, which has become a core, day-to-day working tool for many volunteers across many communities. Similar investment could be made in Tabernacle, which is excellent.

Your point about wanting to not depend on SPARQL-queries can be addressed fairly easily by adding a drop-down with useful pre-set queries for novices to use when wanting to translate useful labels in Tabernacle.

Ijon added a subscriber: Magnus.Dec 23 2018, 5:03 PM