Re: PO-based Documentation Translation

[Keld Jørn Simonsen <keld dkuug dk>]

Some ideas for .po based translation tools:

In OpenOffice.org translations we used a tool to generate
some raw translations. This was based on simple string replacement with
longest string matching first, so it should be simple to program.
The tool we are using in OOo is not open source and is oriented towards
the OOo document format, so it is not useful for GTP. 

Some consideration needs to be done towards efficiency, as the number of
strings may become quite large, like over 100000. But the technique is
simple and gives quite good results, at least for my language Danish,
translated from English. 

My other idea was then to use this technique on the database or corpora
that we have in GTP (or even kde and TP and other Open Source
collections of translations). In this way we can feed the translation
system with a huge database from the start of relations between a
source and a target translation.

The database could be built from the longest strings consisting of whole 
sentences or whole paragraphs or messages, down to just two or one word.
A process after creating the translation database would be to check 
if the automatic generated list is correct. The list itself could be
formatted in .po - to help administering changes, additions etc.

I note that it in some cases may be productive to also use other source 
translations than English. I have myself had good succes with
translation from Norwegian and Swedish, as the grammar is closer to my
destination language Danish, and especially the Swedish are so damn good
in translating everything:-)

I give you these ideas as I realize that my plate is too full to do
something about it in the near future, and I hope I can inspire somebody
to do something about it.

best regards
Keld