Re: translation automation?

[chaddadun at bluewin dot ch]

Salam,

Experience proved that "machine translation" is still far from being a perfect
solution, as texts translated this way require heavy editing and usually
more time. While we might gain 10%, editing the remaining 90% will take
more time and effort.

Other translation tools exist, however. Translation Memory software, such
as Trados and Transit (and OmegaT under all platforms including Linux) is
only the most famous. It is only suitable for highly repetitive text. It
works more or less like Kbabel.

Another more recent alternative is Corpus-based translation tools, such
as MultiTrans of MultiCorpora. It is, in my opinion, the best solution available
for translating documents with relatively less repetitive occurences at
the sentence level, but more repetitive occurences of more than 2 words,
such as computer books. A comparision between both systems is available
here:

http://www.multicorpora.com/papers/new/White%20Paper%20Multilingual%20Corpus_en.pdf

Another effort is underway at the United Nations University: the Universal
Networking Language (UNL). I do not know much about it. Here is the link:

http://www.undl.org/unlsys/

While I cannot still give a hand in any effort in this regard because of
my injury, I would like to say that such software under Linux is becoming
a real need in many international organizations sufferring from abusive
practices of some proprietary software owners and wishing to move towards
open source.

Salam,

Yasser

>-- Message original --
>Date: Tue, 12 Oct 2004 23:55:44 -0700 (PDT)
>From: Nadim Shaikli <shaikli at yahoo dot com>
>To: Documentation and Translation <doc at arabeyes dot org>
>Subject: Re: translation automation?
>Reply-To: Documentation and Translation <doc at arabeyes dot org>
>
>
>--- Abdulaziz Al-Arfaj <alarfaj0 at yahoo dot com> wrote:
>> You mean like this?
>>   http://babelfish.altavista.com/
>>
>> I would very much like to see an English->Arabic and then Arabic->English
>> automated translation, even if it's wrong 90% of the time. If its wrong
>90%
>> of the time then its right 10% of the time, so thats 10% less work for
>us.
>>
>> This however is a _huge_ and complex project, and for someone to write
>> it he needs a very firm grasp of the syntax, semantics and morphology
of
>> Arabic, plus programming prowess. It doesnt look like we have someone
>> like that here... :-(
>
>I think we have some of the ingredients to actually give it a go, what
it
>does require is someone with the enthusiasm and curiosity to stick with
it
>and develop it as much as possible.
>
>The following projects could come in handy,
> + Duali     (http://www.arabeyes.org/project.php?proj=Duali)
> + Farahidi  (http://www.arabeyes.org/project.php?proj=Adawat)
> + Wordlist  (http://www.arabeyes.org/project.php?proj=Wordlist)
>
>I'm by no means saying we can actually do it and get it right, etc.  But
>I am inclined to think that we have some tools in-place to start on such
>an adventure given there is dedication and interest in a long-term project.
>Universities should also be consulted as this is something that would very
>much salivate over (or they should).
>
>Sounds like something to add to Arabeyes' Global TODO list.
>
>Salam.
>
> - Nadim
>
>
>
>
>_______________________________
>Do you Yahoo!?
>Declare Yourself - Register online to vote today!
>http://vote.yahoo.com
>_______________________________________________
>Doc mailing list
>Doc at arabeyes dot org
>http://lists.arabeyes.org/mailman/listinfo/doc