Community Wishlist Survey 2022/Translation/Add DeepL as a machine translation option in ContentTranslation

Add DeepL as a machine translation option in ContentTranslation

  • Problem: Content translation allows pre-filling the text using machine translation provided by either Google Translate or Yandex Translate. This is better than nothing, but the quality of their translations is poor.
  • Proposed solution: Add DeepL (https://www.deepl.com/) as a machine translation option. In my experience their output is much better than the other two, although they only support 24 languages. Probably make it the default for the supported languages, too. Human review is still required, but I think this will reduce the effort spent on fixing boring things, e.g. correcting the pronouns in languages with gendered nouns.
  • Who would benefit: Translators to/from 24 languages
  • More comments: (I've mostly used it when translating English to Polish, also occasionally for various languages to English, and their result was the best almost every time I tried comparing with the other websites.)
  • Phabricator tickets: phab:T301561
  • Proposer: Matma Rex (talk) 03:39, 22 January 2022 (UTC)

Discussion

  • 100% agree! I came here to propose the same. In my experience DeepL is really better than Google T. Javiermes (talk) 11:58, 22 January 2022 (UTC)
  • How does licensing work for this? I know there was a special arrangement worked out for Google Translate. czar 23:59, 29 January 2022 (UTC)
    @Czar I researched it out of curiosity. The terms of use for DeepL's paid services [1] state that "DeepL does not assume any copyrights to the translations made by Customer using the Products", so there is probably no licensing problem. (Also, I found this page about our Google Translate arrangement with a bit more detail than the page you linked.) Matma Rex (talk) 00:33, 30 January 2022 (UTC)
  • Would have supported if I had known voting ended at 18 UTC. ~~~~
    User:1234qwer1234qwer4 (talk)
    22:32, 11 February 2022 (UTC)

Voting

  •   Support Javiermes (talk) 19:23, 28 January 2022 (UTC)
  •   Support It's the highest-quality machine translation out there. Would be great. Femke (talk) 19:37, 28 January 2022 (UTC)
  •   Support Lahmenfurst (talk) 21:01, 28 January 2022 (UTC)
  •   Support Nad.roz (talk) 22:43, 28 January 2022 (UTC)
  •   Support Izno (talk) 00:02, 29 January 2022 (UTC)
  •   Support Edigodiuss (talk) 02:26, 29 January 2022 (UTC)
  •   Support Shizhao (talk) 04:02, 29 January 2022 (UTC)
  •   Support Ottawajin (talk) 06:08, 29 January 2022 (UTC)
  •   Support 𝑇𝑚𝑣 (𝑡𝑎𝑙𝑘) 07:19, 29 January 2022 (UTC)
  •   Support Curios7ty (talk) 08:55, 29 January 2022 (UTC)
  •   Support +1 Afernand74 (talk) 08:59, 29 January 2022 (UTC)
  •   Support Sajbadina (talk) 10:20, 29 January 2022 (UTC)
  •   Oppose risk of multiplication of automatic translation (even good) without human review too hight Triton (talk) 11:23, 29 January 2022 (UTC)
  •   Support Terber (talk) 11:45, 29 January 2022 (UTC)
  •   Support Nardog (talk) 14:32, 29 January 2022 (UTC)
  •   Support Aca (talk) 15:09, 29 January 2022 (UTC)
  •   Support Likibp (talk) 15:37, 29 January 2022 (UTC)
  •   Support — Omnilaika02 (talk) 19:44, 29 January 2022 (UTC)
  •   Support Vincent Simar (talk) 21:24, 29 January 2022 (UTC)
  •   Support It's my preferred translator for bulk text, requiring minimal corrections during review. DeepL's quality is noticeable. I'm very pleased using it to translate text from English to Lithuanian (my native language) compared to Google Translator, which often comes short. Kunigas Michailas (talk) 22:54, 29 January 2022 (UTC)
  •   Support Added a question above czar 23:59, 29 January 2022 (UTC)
  •   Support Tgr (talk) 00:27, 30 January 2022 (UTC)
  •   Support Ali Imran Awan (talk) 07:18, 30 January 2022 (UTC)
  •   Support Lectrician1 (talk) 07:33, 30 January 2022 (UTC)
  •   Support TheInternetGnome (talk) 08:28, 30 January 2022 (UTC)
  •   Support Ameisenigel (talk) 09:20, 30 January 2022 (UTC)
  •   SupportBilorv (talk) 20:45, 30 January 2022 (UTC)
  •   Support Hkoala (talk) 06:37, 31 January 2022 (UTC)
  •   Support Yes it seems quite good. Gam3 (talk) 07:50, 31 January 2022 (UTC)
  •   Support Elenaib (talk) 12:34, 31 January 2022 (UTC)
  •   Support -- big, big +1 Trizek from FR 13:47, 31 January 2022 (UTC)
  •   Support the wub "?!" 15:02, 31 January 2022 (UTC)
  •   Support Djiboun (talk) 15:05, 31 January 2022 (UTC)
  •   Support Matma Rex (talk) 16:32, 31 January 2022 (UTC)
  •   Support Bencemac (talk) 18:12, 31 January 2022 (UTC)
  •   Support Zapipedia (talk) 19:19, 31 January 2022 (UTC)
  •   Support PatriHorrillo (talk) 19:53, 31 January 2022 (UTC)
  •   Support Lalviarez (talk) 20:54, 31 January 2022 (UTC)
  •   Support MONUMENTA (talk) 23:59, 31 January 2022 (UTC)
  •   Support Kazkaskazkasako (talk) 09:08, 1 February 2022 (UTC)
  •   Support Havang(nl) (talk) 09:49, 1 February 2022 (UTC)
  •   Support Eduardo.noeda (talk) 12:00, 1 February 2022 (UTC)
  •   Support Szymonel (talk) 13:30, 1 February 2022 (UTC)
  •   Support Absolutely agree! RaquelFlorez (talk) 14:00, 1 February 2022 (UTC)
  •   Support Worrida (talk) 14:41, 1 February 2022 (UTC)
  •   Support Thibaut (talk) 16:15, 1 February 2022 (UTC)
  •   Support XavierItzm (talk) 20:14, 1 February 2022 (UTC)
  •   Support Thingofme (talk) 02:39, 2 February 2022 (UTC)
  •   Support Geert Van Pamel (WMBE) (talk) 16:38, 2 February 2022 (UTC)
  •   Support Sargento - A sus órdenes 17:41, 2 February 2022 (UTC)
  •   Support Aimwin66166 (talk) 06:17, 3 February 2022 (UTC)
  •   Support I came to say propose the same thing. Rotavdrag (talk) 11:30, 3 February 2022 (UTC)
  •   Support In my opinion DeepL is the best CAT in the web. Giacomo Alessandroni let's talk!   14:32, 3 February 2022 (UTC)
  •   Support Paucabot (talk) 15:51, 3 February 2022 (UTC)
  •   Support Pintakuda (talk) 09:46, 4 February 2022 (UTC)
  •   Support Dsalerno (talk) 00:58, 5 February 2022 (UTC)
  •   Support - Darwin Ahoy! 02:10, 5 February 2022 (UTC)
  •   Support Rdr3141 (talk) 09:03, 5 February 2022 (UTC)
  •   Support Kpjas (talk) 12:52, 5 February 2022 (UTC)
  •   Support Exilexi (talk) 17:59, 5 February 2022 (UTC)
  •   Oppose --Ciao • Bestoernesto 03:46, 6 February 2022 (UTC)
  •   Support Fehufanga (talk) 07:50, 6 February 2022 (UTC)
  •   Support One of the best machine translation service on the net--Vulp❯❯❯here! 09:01, 6 February 2022 (UTC)
  •   Support —— Eric LiuTalk 10:00, 6 February 2022 (UTC)
  •   SupportThanks for the fish! talkcontrib (he/him) 17:12, 6 February 2022 (UTC)
  •   Support Ayumu Ozaki (talk) 02:21, 7 February 2022 (UTC)
  •   Support DesPsyCHo (talk) 09:35, 7 February 2022 (UTC)
  •   Support Ryse93 (talk) 12:46, 7 February 2022 (UTC)
  •   Support Serg!o (talk) 13:00, 7 February 2022 (UTC)
  •   Support Tom Ja (talk) 17:48, 7 February 2022 (UTC)
  •   Support Readers need a warning, "this page is machine translated from [source page]", sort of like webpage links framed as "ad-supported article." Otherwise readers may be confused by bits of garbage in the machine result. Gpwitteveen (talk) 21:38, 7 February 2022 (UTC)
    Publishing purely machine-translated text is not the point of mw:Content translation (and would be not quite in line with the standards on the various wikis); this proposal only aims to extend the range of available machine translation tools used to aid manual translation. ~~~~
    User:1234qwer1234qwer4 (talk)
    22:30, 11 February 2022 (UTC)
  •   Support ~Cybularny Speak? 23:39, 9 February 2022 (UTC)
  •   Support Quiddity (talk) 08:48, 10 February 2022 (UTC)
  •   Support DeepL should be prefered for the languages it supports, and then google, and in the last Yandex, and in the absolute worst case scenario opt for Bing. Pauloroboto (talk) 10:13, 10 February 2022 (UTC)
  •   Support Wikiusuarios (talk) 20:23, 10 February 2022 (UTC)
  •   Support Marcok (talk) 12:52, 11 February 2022 (UTC)
  •   Support AngryBiceps (talk) 16:07, 11 February 2022 (UTC)