4 ms·

Some grievances were vague. It doesn't follow our translation guidelines. What specific guidelines did it not follow? It doesn't respect current localization fo

Some grievances were vague. It doesn't follow our translation guidelines. What specific guidelines did it not follow? It doesn't respect current localization for Japanese users, so they were lost. What was not localized? Mozilla's response should not be limited to clarifying these grievances. But it could have been all the staff member who responded could do.

totetsu 1y ago
These ones? https://github.com/mozilla-japan/translation/wiki/L10N-Guideline https://github.com/mozilla-japan/translation/wiki/L10N-Guide... Looking through that wiki there seems to be a lot of things that ML would get wrong.
- pseudalopex 1y ago
  I edited my comment to clarify I hope. Imagining what it could have done wrong and knowing what it did wrong are different.
  - tsimionescu 1y ago
    Do you expect someone who has just watched a bot replace 20 years of their work, with no prior consultation or review, to now write a detailed post about how translations by the bot are not specifically wrong? The core issue here is the way the bot was deployed. The fact that they had the poor taste to make it auto-replace articles written by their own volunteers is idiotic and disrespectful in the extreme. A new bot should work entirely in the back end, sending proposals for translations to the volunteers, who can choose to accept them or ignore them. Once the rate of acceptance is very high, for a specific individual language, then you might consider automating further. And yes, this effort needs to be done for each language separately. Just because the bot works well in Italian doesn't in any way guarantee that it will work well in Japanese. Machine translation quality varies wildly by language, this is a well known and obvious fact.
    - shaky-carrousel 1y ago
      This is Mozilla as usual, arrogant and tone deaf.
- numpad0 1y ago
  lmao. This is a "research team and five years" task with current state of LLM.
  - pmontra 1y ago
    Yes, but now that's the answer to questions like "how do we deploy AI without pissing off our communities?"
    - eloisant 1y ago
      I have your answer: you provide a tool for your translators, you don't unleash a bot that makes changes left and right and creates new pages. Typically, when a new page is written in English, don't automatically generate a version in all languages. When a translator starts creating the page in their language, provide a button to pre-fill with ML translation if they want to. And for users, you can display the English version with a message, "this is not translated in your language yet but you can read an ML version if you want".
      - taikahessu 1y ago
        This sounds about right, and should be incorporated everywhere.
  - bw86 1y ago
    For reference: https://xkcd.com/1425/ https://xkcd.com/1425/
- eloisant 1y ago
  Interestingly, it looks like some of the "bad examples" are precisely the kind of things ML would produce. Others are what non-native speakers would produce (starting with "あなたは" to translate a sentence starting by "you"). I'm sure some translators were using ML before it was integrated, and those guidelines are here in particular to tell them about those problems. Also, ML is now really good to translate between European languages, but Japanese is very different in its structure so ML from English to Japanese is not as good. I'm sure some people who only know English/French/Spanish/German saw that ML is pretty good, and don't realize that for some other language it just doesn't work.
  - v1ne 1y ago
    > Also, ML is now really good to translate between European languages As somebody who has to regularly bear "German" machine-translated UIs and manuals that originate in English, I can only say: No, it's not. It's atrocious.
    - clausecker 1y ago
      Best one was when gedit had the option to syntax highlight for a language named “Los.”
      - flobosg 1y ago
        Not a bad name, to be honest!
sitharus 1y ago
Those would be the guidelines that all translation contributors are expected to follow, which are given to all prospective translators. It sounds like Mozilla just turned on the machine without consulting the human translators to see if the machine actually worked in a useful manner.
- op00to 1y ago
  Specifically which guidelines? Not a URL. Not hand wavey “oh you know the guidelines”. A text list of the guidelines that are not followed.
  - exe34 1y ago
    It might be more helpful to point out which guidelines it did follow. Humans are expected to read and obey these things - so presumably whoever deployed them will be aware and can demonstrate that they were followed.
- pseudalopex 1y ago
  > It sounds like Mozilla just turned on the machine without consulting the human translators to see if the machine actually worked in a useful manner. Yes. And someone should make a real apology. But learning what the machine did wrong is part of fixing a machine.
  - sitharus 1y ago
    Yes, that's why you engage with the people doing the work first and run it on a staging environment to see what would be overwritten. You test until it's working well enough to enhance the effort done by the translators.
    - pseudalopex 1y ago
      And someone should make a real apology. Which I said.
    - danielscrubs 1y ago
crazygringo 1y ago
I don't know why you're being downvoted, you're exactly right. The person replying is probably not an expert in this. But they want to get more details so they can figure out how to get it to the right people with more information. This is how it's supposed to work.
- ApolloFortyNine 1y ago
  * * *
  - BoredPositron 1y ago
    The market fit of your words matter and the non apology "sorry you feel that way" doesn't help. It's one of those gnarly phrases that only get used if you feel the need to act but in reality don't want to. Maybe the person is sincere his words are tainted by the overuse in shady corp responses. If you are engaging with your community you need to know this otherwise you'll always look like a clown.
  - exe34 1y ago
    > Same reason people here are taking "let's jump on a call" as some personal attack. Some people just like drama. No, some of us can see into the future, because it flows from the past. When management shits on 20 years of work and breaks everything after not listening to your warnings, they don't suddenly start listening and understanding out of nowhere.
  - kuschku 1y ago
    Tools should work for their users. In this case, the bot should work (and be controlled by) the localization teams for each language. If the bot has the power to overrule the volunteer translation teams, the entire power structure is wrong from the get go.
    - crazygringo 1y ago
      Or maybe it was just a poorly thought out piece of software that will get rolled back? Saying "the entire power structure is wrong from the get go" is a huge conclusion to draw here. People write internal tools and make mistakes. Mistakes can be fixed.

totetsu 1y ago

These ones? https://github.com/mozilla-japan/translation/wiki/L10N-Guideline https://github.com/mozilla-japan/translation/wiki/L10N-Guide... Looking through that wiki there seems to be a lot of things that ML would get wrong.

pseudalopex 1y ago
I edited my comment to clarify I hope. Imagining what it could have done wrong and knowing what it did wrong are different.
- tsimionescu 1y ago
  Do you expect someone who has just watched a bot replace 20 years of their work, with no prior consultation or review, to now write a detailed post about how translations by the bot are not specifically wrong? The core issue here is the way the bot was deployed. The fact that they had the poor taste to make it auto-replace articles written by their own volunteers is idiotic and disrespectful in the extreme. A new bot should work entirely in the back end, sending proposals for translations to the volunteers, who can choose to accept them or ignore them. Once the rate of acceptance is very high, for a specific individual language, then you might consider automating further. And yes, this effort needs to be done for each language separately. Just because the bot works well in Italian doesn't in any way guarantee that it will work well in Japanese. Machine translation quality varies wildly by language, this is a well known and obvious fact.
  - shaky-carrousel 1y ago
    This is Mozilla as usual, arrogant and tone deaf.
numpad0 1y ago
lmao. This is a "research team and five years" task with current state of LLM.
- pmontra 1y ago
  Yes, but now that's the answer to questions like "how do we deploy AI without pissing off our communities?"
  - eloisant 1y ago
    I have your answer: you provide a tool for your translators, you don't unleash a bot that makes changes left and right and creates new pages. Typically, when a new page is written in English, don't automatically generate a version in all languages. When a translator starts creating the page in their language, provide a button to pre-fill with ML translation if they want to. And for users, you can display the English version with a message, "this is not translated in your language yet but you can read an ML version if you want".
    - taikahessu 1y ago
      This sounds about right, and should be incorporated everywhere.
- bw86 1y ago
  For reference: https://xkcd.com/1425/ https://xkcd.com/1425/
eloisant 1y ago
Interestingly, it looks like some of the "bad examples" are precisely the kind of things ML would produce. Others are what non-native speakers would produce (starting with "あなたは" to translate a sentence starting by "you"). I'm sure some translators were using ML before it was integrated, and those guidelines are here in particular to tell them about those problems. Also, ML is now really good to translate between European languages, but Japanese is very different in its structure so ML from English to Japanese is not as good. I'm sure some people who only know English/French/Spanish/German saw that ML is pretty good, and don't realize that for some other language it just doesn't work.
- v1ne 1y ago
  > Also, ML is now really good to translate between European languages As somebody who has to regularly bear "German" machine-translated UIs and manuals that originate in English, I can only say: No, it's not. It's atrocious.
  - clausecker 1y ago
    Best one was when gedit had the option to syntax highlight for a language named “Los.”
    - flobosg 1y ago
      Not a bad name, to be honest!

sitharus 1y ago

Those would be the guidelines that all translation contributors are expected to follow, which are given to all prospective translators. It sounds like Mozilla just turned on the machine without consulting the human translators to see if the machine actually worked in a useful manner.

op00to 1y ago
Specifically which guidelines? Not a URL. Not hand wavey “oh you know the guidelines”. A text list of the guidelines that are not followed.
- exe34 1y ago
  It might be more helpful to point out which guidelines it did follow. Humans are expected to read and obey these things - so presumably whoever deployed them will be aware and can demonstrate that they were followed.
pseudalopex 1y ago
> It sounds like Mozilla just turned on the machine without consulting the human translators to see if the machine actually worked in a useful manner. Yes. And someone should make a real apology. But learning what the machine did wrong is part of fixing a machine.
- sitharus 1y ago
  Yes, that's why you engage with the people doing the work first and run it on a staging environment to see what would be overwritten. You test until it's working well enough to enhance the effort done by the translators.
  - pseudalopex 1y ago
    And someone should make a real apology. Which I said.
  - danielscrubs 1y ago

crazygringo 1y ago

I don't know why you're being downvoted, you're exactly right. The person replying is probably not an expert in this. But they want to get more details so they can figure out how to get it to the right people with more information. This is how it's supposed to work.

ApolloFortyNine 1y ago
* * *
- BoredPositron 1y ago
  The market fit of your words matter and the non apology "sorry you feel that way" doesn't help. It's one of those gnarly phrases that only get used if you feel the need to act but in reality don't want to. Maybe the person is sincere his words are tainted by the overuse in shady corp responses. If you are engaging with your community you need to know this otherwise you'll always look like a clown.
- exe34 1y ago
  > Same reason people here are taking "let's jump on a call" as some personal attack. Some people just like drama. No, some of us can see into the future, because it flows from the past. When management shits on 20 years of work and breaks everything after not listening to your warnings, they don't suddenly start listening and understanding out of nowhere.
- kuschku 1y ago
  Tools should work for their users. In this case, the bot should work (and be controlled by) the localization teams for each language. If the bot has the power to overrule the volunteer translation teams, the entire power structure is wrong from the get go.
  - crazygringo 1y ago
    Or maybe it was just a poorly thought out piece of software that will get rolled back? Saying "the entire power structure is wrong from the get go" is a huge conclusion to draw here. People write internal tools and make mistakes. Mistakes can be fixed.