Managing translations

We currently have a large number of skilled translators, a small number of dedicated, regular contributors, and a small collection of regularly changing content (articles and messages) to be translated into many languages. What follows is a discussion of how to manage fast, reliable translation of this content, assuming

  • ~100 articles and a few urgent messages to manage
  • 5-15 supported languages, 40-100 other languages.
  • 1-4 source languages for original content

At the end are some generalizations to encyclopedia-sized collections of content.

I. Central Collection and Prioritization edit

  1. Maintain a single list of key content in one language, prioritized by importance.
    • If content exists, but has yet to be published, create a link to the future page that will hold the published content.
    • Link to published|draft versions of that item in each supported language, two links per language; mouseover text should indicate when each version was last updated.
    • The original/source language for each piece of content should have bolded links
    • Example:   Multilingualism: de, en, fr, ja
    • Each language may also note the % completion of its translation next to its link.
    • A parallel page will list existing links to translations in undersupported languages. (Some of these may be published in a different way, to make allowances for the sparsity of pages in those languages)
  2. Maintain a separate list of secondary content
    • Format as above
  3. Maintain a template with urgent/100-lang content above both lists.
    • Each link in this template will lead to a page devoted to that piece of content; depending on the length and complexity of the content, its page format will vary (e.g., single sentences may all go on one page).

Future improvements edit

Eventually: The choice of the language in which the article names is presented will be an option on the page; the list of links, with information-enhanced mouseover text, will be database-generated. A self-reported level of completeness of the translation will be stored with each edit. Each published article will have an explicit version number, which will increment with each major change. Translators will be asked to note which article version was being used as a source; edit summaries (and the db) will note which source language, with which version-number, was being translated at that time. This completeness %, source-language, and source-version information will also be included in the [mouseover text].

II. Translators and Translation edit

Translators edit

Anyone can help translate and copyedit draft translations. Non-fluent people should be encouraged to help in this way.

There should be a designation for trusted (perhaps low-availability) translators and proofreaders, who will be able to modify final content in their fluent languages, and within that group a core of high-availability translators, for coordinating important messages and requests and for overseeing the translation process.

Let us call trusted translators-and-proofreaders TPs, and high-availability translators HATs. These should be noted as contributors trusted to produce final copies of content in their language. Only they will be able to modify final translated content.

Trusted translators and proofreaders:
Any editor can ask to be a TP, with a simple vote of approval by their language community and the other translators. There should never be only one TP in a given language, to ensure competent peer-review of any 'final' translation; either a translator in another language should be able to double as a TP in the new language, or the first two+ TPs in a language should be created at the same time. TPs will be able to edit the final article pages; under the current system, this might mean involve protecting those pages and granting TPs meta-adminship.

High-availability translators:
HATs should be active contributors who watch relevant channels of communication -- meta: translation pages, the mailing list, their local project lists & VPs, &c. They can be appointed informally by community support and at some point blessed by the board, to the extent of getting editing rights on the WMF site. They will help coordinate translators and ensure key articles and site-sections are kept up-to-date.

There needs to be a way to contact a large list of occasional translators for urgent messages that need to be widely distributed (Ex: server issues; translation of important requests; coordinated press releases). This can be handled separately by the HATs in each language; later a single mechanism for doing this (via a global "New message" banner alert?) will be handy.

Supported languages The WMF site will have a public set of supported languages -- ones in which the core of the site will be kept up-to-date, for instance. Each of these languages should have at least 5 TPs and 2 HATs.


Translation edit

Translators can visit the list of items needing translation, and set to work. Urgent messages come first. Other content should generally be translated in order of priority. Anyone can update draft translations; only TPs can update final translations.

Copyedits by TPs to final translations should be marked "minor"; major edits should refer only to updates in translated content. These major edits should be made rarely, after much draft revision, to simplify the process of determining whether other languages are in synch with the latest major revision.

Short messages that need quick translation into many languages will generally be more open, and have no barrier to translation.

III. Overview and Review of Translations edit

TPs and other active translators should regularly update summaries of their efforts -- directly on the list of articles (the % completion of each translation) and on an overview page for that lang (which articles of particular importance to that lang community need work, who is actively translating / available on which days of the week, where to look for lang-specific resources, which non-translators on the local project are available for help with proofreading, &c).

TPs and other fluent proofreaders should regularly check the latest draft, announce any mischievous translators who are altering meaning, and (when appropriate) update the final version of that article. Each final version should be checked by a fluent proofreader/copyeditor before publication (one might require that the last person to proofread/edit the article should be different from the HAT updating the live site with it, who should also review it first).


IV. Available resources edit

As of July, 2004:

There are ~100 translators, and ~200 {translator, skilled-from-language, fluent-to-language} triples, on the List of translators on meta. There are perhaps twice as many people across the projects who have expressed an active interest in multilingualism, translation, or embassy-work at some point.
There are 2-5 key Wikimedia Foundation pages -- | main page | donations | projects | about WMF | contact us |
There are 12 Wikipedia languages with over 10,000 articles -- da|de|en|eo|es|fr|it|ja|nl|pl|sv|zh
There are at least 3 native speakers of each of these languages, active on their projects and occasionally on meta:, who have offered to help translate content.

Current Translation structure:

Requests for translation into all languages; Translation-into-local-lang projects on local Wikipedias.
broadcasting announcements (via mailing list, IRC, & Goings-on) that there is something new to translate; broadcasting an editable English message and hoping local projects translate it appropriately.
creating pages on meta: which say in English that they are intended to contain translations of a given message into as many languages as possible.