[haiku-development] Re: [haiku-doc] Re: Machine translation (was: Re: Wiki for translation/localization teams)

  • From: Sean Healy <jalopeura@xxxxxxxxxxx>
  • To: haiku-development@xxxxxxxxxxxxx
  • Date: Sun, 01 Nov 2009 19:46:09 +0100

Jonas Sundström wrote:

I personally wouldn't want to be restricted to some simplified
language, and I resent the idea of having a single language - likely English - as a master language which one has to use for
content creation, and from which all translations are generated.
English often has that role, e.g. in programming, but formalizing
this practice also for natural languages doesn't sit well with me.

As I mentioned, the lack of desire to obey such restrictions is one reason a lot of machine translation efforts fail.

And as for your resentment of English as a "master language", that really is not the fault of the field of machine translation, which is just following a general worldwide trend. English is widely enough spoken that most content providers are going to start with English, because they can reach more people that way. Those English documents become the de facto normative bases for translation. Not for everyone, of course, but for a very very large majority.

And even if the language is not English, there is usually a "master language" that is the normative version. The only exception I can think of is the European Union, and it's not working out very well for them. They're running into problems because every language is normative. This leads to conflicting interpretations of laws, and a judge eventually has to come down in favor of one interpretation. As it is, French, English, and German have become "more equal" than the other languages in practice.

And to apply this to Haiku, whether or not you think it's "formalized", the fact is that English is the normative language for the Haiku documentation. The documentation is written in English, the translators translate from English to the target language. So machine translation has nothing to do with setting English or any other language up as a "master language"; it's simply reflecting the world around it.

If you think machine translation fits the needs of Haiku, you will
likely have to put in much of the work yourself. We have people
wanting to translate, given tools, and we have developers able to
write web services, etc, but we have few actual linguists to maintain the rule set(?) of a machine translation setup and define/refine a master language.

Actually, I don't think it fits Haiku's needs. Someone else suggested or it. (Or rather, someone else suggested suggested something that turned out to be machine-assisted translation, and I misinterpreted.) I just wanted to point out the problems inherent in the approach, and offer my help if the team decided to go that way.

I think manual translation allows for more freedom of linguistic expression in the docs, and provides a more natural sounding translation.

So please, everyone, understand that I am not trying to push a pet project here. I am trying to make sure everyone concerned understands what's involved.

Other related posts: