Hacker News Reader

Comment by xupybd | Hacker News Reader

This is a hard problem to solve. The amount of work required is huge. I understand the problem. I work on Italian CNC machines. Half my time debugging is in Google translate trying to figure out comments, variable names and error messages. I don't think it's impossible to translate all of the material to English but it would cost hundreds of thousands. I Know FOSS doesn't have the same economic model but it does have opportunity costs. If someone wants to advocate for this great, I'd be cautious to ever disparage a project for not doing this. Simply because it is a tremendous amount of work.

View on HN

Is translation really that expensive? Once they're in a spreadsheet there are plenty of services which will translate them.

Unfortunately it is not like that. We experienced various difficulties translating only few sections in the Django documentation because, also for technical text, mechanical translation is not effective, but you have to adapt the text.

For good translation you need some one who is fluent in both languages and fluent in the technical domain you want translated and is skilled at translation. You're asking for skilled labor at the intersection of 2-4 different skills (depending on how you count it) and depending on the domain one of those skills might already be pretty rare.

Bunch of short UI strings in a spreadsheet don't work well. You need screenshots or English builds if possible, and the shorter sentences become, more context explanation become necessary, unless you actually want:

- "Kill them, they're Russians"

in place of "Remember - [say]No Russian[words]"[1]

- "Province"

in place of "State[as in status]"[2]

- "Scheduled installation time:"

in place of "Installation start in[at following directory]:"[3]

- "Preparing Phacochoerus africanus"

instead of ... "Loading Warthog"?[4]

- "An error report has been created. Would you like to send it?"

when it should say "[x] Send error reports automatically"(IIRC this one was in DisplayLink Windows driver)

Human language relies a lot on contexts, and without it there could be multiple technically valid candidates of interpretation that make no sense. Translators need the picture of the situation the text is used to narrow down the valid ones, and the easiest way is convey the situation is to just give them bunch of pictures.

Even something simple as "Back" could mean:

- "[show] Back [side]"

- "[go] Back [one step]"

- "[on] Back [of body]"

and there's no way for translators to know from "English: "Back", Translated: _____" form. And only one of them is correct.

Machine translations are hilariously worse by the way[5][6]:

  English:     "DO give methods names that are verbs or verb phrases."
  MT Japanese: "Do-not-verb or verb phrases specify name of a Method."

Ref., (most is in Japanese, sorry):

1: https://news.denfaminicogamer.jp/gamenewsplus/200402i

2: https://blog.netandfield.com/shar/2020/07/sql-server-3.html

3: https://sccm.jp/2020/12/06/post-2871/

4: https://twitter.com/skb_apos/status/1338001444138467329/phot...

5: https://github.com/dotnet/docs.ja-jp/issues/118

6: https://cpplover.blogspot.com/2018/07/blog-post.html

Hi, thanks for your feedback. My suggestion in the article was more to take care of all people in the community than to translate everything. Starting to use a better/simpler English would be a great starting point.

Reply on news.ycombinator.com