Re: Using Windows Language Portal for IDE string translation

2018-04-04 11:15 GMT-03:00 <[email protected]>:

>
>
> On Tue, Apr 3, 2018, 11:49 PM <[email protected]> wrote:
>
>> Hi Brian
>>
>> I just realised that to automatically translate and manually check the
>> IDE could be a good solution. For the documentation, I think most used
>> translators are offered as web service for free with some limits of
>> quantity, but there is the Apertium platform (GNU):
>>
>> http://wiki.apertium.org/wiki/Main_Page
>> https://github.com/apertium
>>
>> The wiki page is really not newbie friendly but I have used it for an
>> installation art, it runs as a server in the local machine and works pretty
>> well (there is even a ppa for Ubuntu). Maybe it could be used as a quark
>> to translate the documentation then and there, it has language packages for
>> many language pairs. I could try that and tell.
>>
>
> Sorry, I genuinely have no idea what I'm looking at here; browsing the
> wiki and GitHub only left me more confused. I found a "language pair list"
> page which makes it sound like only a few languages are well-supported -
> Spanishis that correct? If you think it could be useful maybe you should
> try it out and explain what the workflow would look like.
>

Yep, that wiki is terrible, I know. I think it is in heavy development
right now.

The enter point is to install nightly builds repo, apertium, apertium-apy
and languages data (see below).

http://wiki.apertium.org/wiki/Apertium-apy/Debian
http://wiki.apertium.org/wiki/Install_Apertium_core_using_packaging
http://wiki.apertium.org/wiki/Apertium-apy

I can't start the server as a service but the command:

$ apertium-apy /usr/share/apertium

works (that's hot new)

There are many packages for languages pairs, in the repo are for example:

apertium-en-es # i.e. lang data pair English-Spanish
apertium-en-fr # i.e. lang pair English-French
apertium-eng # IIRC with 3 letters are analyzers, generators, morphemes and
taggers packages for each language or pair (don't ask me more)
apertium-spa
apertium-fra

Languages progress have different states (incubator, nursery, staging) like
Debian is not just Spanish.

Then programs can make request and process the answer as usual:

curl http://localhost:2737/listPairs
{"responseStatus": 200, "responseData": [{"targetLanguage": "spa",
"sourceLanguage": "eng"}, {"targetLanguage": "eng", "sourceLanguage":
"spa"}, {"targetLanguage": "eng_US", "sourceLanguage": "spa"}],
"responseDetails": null}

curl 'http://localhost:2737/translate?langpair=es|en&q=está'
{"responseStatus": 200, "responseData": {"translatedText": "It is"},
"responseDetails": null}

curl 'http://localhost:2737/translate?langpair=en|es&q=there+is'
{"responseStatus": 200, "responseData": {"translatedText": " hay"},
"responseDetails": null}

I don't know how well it will perform for complex sentences, Google's
engine is surely stronger but it can't be used unlimited and clients need
to be registered. These things are kind of green yet but an approach like
translating while parsing the docs for rendering (preferable with a free
engine, or a least free for personal use) solves synchronisation and effort
problems and would be easy than creating a platform for translations. Or a
step towards this last, I don't know, it's green, 'super green'.


> Also, I feel like you have missed the point with the MS translations -
> they are professionally produced, accurate UI translations that MS uses for
> their own products. This is the best kind of database for automated
> translation.
>

Personally I'm perfectly fine with this if doesn't generate dependency to
MS in the future.

Best
Lucas