Hacker News Reader

Comment by mikestew | Hacker News Reader

When I worked at Microsoft as a localization lead, we had actual human beings that would look over this stuff before calling machine translation good (of course, we had actual human translators back then, too). But that was a long time ago, back when Microsoft had human software testers, too.

And what options are available when I right click on a .zip/.postcode file? Say, perhaps I wished to uncompress it?

View on HN

I think the worst example of this overzealous localization was when some person translated the VB(A) reserved keywords to the appropriate language (I want to say it was german or swedish). Happened in the 1990s.

Your programming language reserved keywords changing depending on your selected language was a big facepalm.

Searching for this problem on the web doesn't produce any valid relevant results, so the web has forgotten or maybe the horror has been plastered over.

Maybe you can safely believe it was all just a nightmare... certainly sounds like one!

E.g. Excel does that still.

Yes. The best is when you get an Excel file from Japan and the formulas are in Japanese. Who needs encryption or obfuscation.

No? Formulas are stored in a language-independent way, you should see English formulas in that file.

The missing part of that story is probably the file is a .xlsm (or even an older .xls with macros) and therefore has VBA scripts, which is not saved in a language-independent way.

Probably someone manipulating with formulas with VBA and improperly using FormulaLocal property. Formulas are stored in language neutral way and it shouldn't be a problem under normal circumstances.

But is it Shit-JIS, or are you lucky enough to get UTF-8?

So obnoxious, although apparently in newer versions you can modify what language you want.

I remember seeing the dictionary to translate between German and English function names[1], one time I googled for "Excel German translation" and Google proudly showed the Translate interace pre-filled with "English: excel / German: übertreffen"

[1] https://www.perfectxl.com/excel-glossary/what-is-excel-funct...

The feature itself isn't that crazy – it's even a good thing, especially considering that Excel (and to lesser degree, VBA) were designed for ordinary people and that in the 80s/90s English was less common than it is today.

Where it went wrong is that as soon as you choose a language you were stuck with that. It should have been a display option.

https://news.ycombinator.com/item?id=21352796

VBA is still localized in many languages, I saw macros in italian. It's not just VBA keywords either, even excel functions are localized. e.g. this is from microsoft's docs[0]

    =CERCA.VERT(B2;C2:E7;3;VERO)

[0] https://support.microsoft.com/it-it/office/cercare-valori-co...

There was a famous translation mistake back in the days of IE 5 for Mac or so where in one of the settings screens they translated the country code for Norway "no" as "nej" (literally No) in Swedish

This is still a problem today with YAML. Say you have something like a list of locales, the word 'no' get's interpreted as a boolean.

I get that in the latest version of the Panasonic aircon apps. They have the term "No." (with ha dot, as the column header for the column "number" in a table). So in Swedish the app says

    Nej.  
    1
    2

i18n is HARD. But it's not THIS step that is hard. This is day one of i18n school. You can't translate from english terms. Just stop translating things from english, translate from a neutral key!

    key         en-US     sv-SE
    num_short   "No."     "No."
    yesno_no    "No"      "Nej"
    yesno_yes   "Yes"     "Ja"

This seems to me like a case of a dev that used the zipPostalCode instead of the zipFilename key from the string table on accident or out of laziness.

Still, any machine translator worth your time would know what the phrase zip file is, unless it's not a machine but someone manually changing it.

Since translation is given less and less thought, the underlying system is also becoming more basic and less context-aware.

I don't remember exact occurrences but I've noticed quite a few similar problems on macOS that are just the result of words being translated out of context.

On a popular forum software I used to do support for, "flag" (for moderation) was translated in French as "drapeau" for example. Which is technically correct I suppose, drapeau does mean flag, but the wavy kind of flag that countries like to use as symbols. Nothing else. And it's not a verb.

In this case, there is simply the string "zip" somewhere that is used both in some address forms and for the file type, and the translator just happened to translate the first occurrence probably without even knowing that is was also used in another context as well. Or without any means to ask the dev team to separate them into two terms. Which might not even be possible if they used too simplistic a translation system.

The programmers and the translators are different people who have probably never met eachother.

Translators usually just get to see the string they are translatinbg, and if they're lucky a 1 sentence explanation written by a programmer.

With no context, a "Zip file" might indeed mean a "postcode file" - perhaps it is a file for storing all the zip codes, and now it will be used for storing all the postcodes?

I’ve drapeaued this comment to favorites :)

J’ai drapeaué

To be fair, translation systems genuinely are a pain, and it's easy to make a mistake either as a developer or as a translator who has thousands of strings to translate.

It doesn't help that a lot of developers probably never translated an application; when I added translations the first thing I did was translate it myself, but that's not really feasible for larger systems and/or with people who don't speak a second language well enough to actually do translations.

I want to share this gem of a screenshot I took on the Belgian Amazon (Amazon.com.be) settings page a few weeks back:

https://media.discordapp.net/attachments/323831494718128130/...

This is French, English, German and JavaScript all in one sentence.

Two trillion dollar company.

Other localizations of Amazon are just as hilarious. And to understand the product listings, you have to know (pretty complete) English, because the translations only make any sense if you know the English description and can reverse the translation in your head back to English.

So, you have to understand 2 languages to understand these listings.

> that is used both in some address forms and for the file type

Oh yes, the classic "I just picked the string from my IDE's dropdown list, it was there so it's fine right?"

They can't work on the per-word basis, translations only works for long enough/large enough texts and images. Words("zip") and partial sentences("expand zip") are garbage in. Out comes "postcodes" and "elaborate addresses" or whatever the PRNG chooses.

I’d assume they have to use Bing Translate instead of something decent like DeepL.

Reply on news.ycombinator.com