TL;DR: Mozilla is killing localization on Support Mozilla, overwriting articles written by humans with machine generated translations. Although Mozilla knows that their AI doesn’t localize or adhere to style guides, Mozilla is going live with it anyway. I thank locale leaders and localizers for their tireless efforts. Locale leaders seem to be obviated by AI, and Mozilla has nothing to say about it.

you are viewing a single comment's thread
view the rest of the comments
[+] -40 points 11 months ago (8 children)

In all objectivity, this seems like the most rational thing to do.

Now that you can have good translation by AI, why would anyone want to waste their time on localization?

  • source
  • hideshow 8 child comments
  • [–] 18 points 11 months ago

    The main issue, and this is also mentioned in the blog post, is that the bot only does translation and not localisation.

    The first is just taking the words from one language and changing to another.

    The second is to actually make sure the text in the new language makes proper sense. Maybe the English article uses some analogy that does realy make sense in the new language. Localisation is to find some other suitable analogy to use instead, so that the point from the main article is kept, but it still makes sense.

  • source
  • parent
  • [–] 16 points 11 months ago*

    I do some volunteer work on localizing for Welsh and some Spanish. AI translations are hot garbage still even for the larger languages like Spanish. You can get a general idea of the intent most of the time, but that's about it. It's nowhere near good enough to replace humans with yet, definitely not good enough to overwrite what they've already translated.

  • source
  • parent
  • [–] 7 points 11 months ago

    As a translator, I have to say machine translations are really good if you just want to get the gist of things. If all you want is to understand 90% of what's written and you can live with the margin of error, then I'm not gonna try to convince you to hire me.

    But if you need to understand 99.9% then a human is required because a machine just will not understand the nuance or have a larger context unless instructed. Localizing is another issue too, you'll have to take into consideration the cultural nuance of the target language. Localizing software is even more niche because you're often limited in character count for the source language which machines often misunderstand, and the same limitation for the target language that a machine may not account for.

    tl;dr - hire me if you want good localization

  • source
  • parent
  • [–] 11 points 11 months ago (3 children)

    I've never seen any machine translation that I would consider anywhere near good.

  • source
  • parent
  • hideshow 3 child comments
  • [–] -1 points 11 months ago (2 children)

    deepl.com (text translation) has been useful to me

  • source
  • parent
  • hideshow 2 child comments
  • [–] 3 points 11 months ago (1 child)

    It is good, specially on mid size text. But it is not good enough. When text is long or too short, it gets lost and makes tons of context mistakes. It also tends to be unnatural for the target language preferring original language phrasing.

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 11 months ago* (last edited 11 months ago)

    I have to admit I've only ever used it to translate a paragraph or two at a time.... where I was just looking for the gist of a text.

    Not too surprising, considering that for centuries many people well-versed in two languages have made a very good living as translators ... and often having to get delicate nuances across (for poets as well as statesmen). It's as much art as science.

    overwriting articles written by humans with machine generated translations. I really don't understand that! But then, there are truckloads of worthwhile texts from throughout history that will never see translations otherwise ... so that's a worthy cause. Over time it may be improved, IF the algos are given feedback that allows them to learn from mistakes.

  • source
  • parent