It’s no tower of Babel, it’s South Africa

Originally in AsiaSource blog reposted in FNsRamblings

By Frederick Noronha

Countries from the South, emerging from a number of problems, can seem like a 
Tower of Babel. But when it comes to coping with diversity, Dwayne Bailey's 
story from South Africa makes the point nicely. Free/Libre and Open Source 
Software can indeed offer interesting solutions even to smaller languages.

From 2001, this slim South African started the Translate.org.za project.

WIKIPEDIA notes: South Africa has eleven official languages, which is second 
only to India. As a result, there are many official names for the country. It 
also recognises eight non-official languages (Fanagalo, Lobedu, Northern 
Ndebele, Phuthi, Sign Language, Khoe, Nama, and San).?

Dwayne explains what this is all about: "At Translate.org.za we're localising 
Free and Open Source Software into 11 South African languages. One is English, 
it's quite an easy one (smiles). The others are Afrikaans, Zulu, Khosa, Venda, 
Tsonga, Tswana, Siswati, Northern Sotho, Southern Sotho, Ndebele."

All these languages use the Roman script. Bailey (33) points out that this 
makes the task easier. But it is a heavy task nonetheless. Some South African 
languages have special characters, but these are Romanised. "We have no 
problems with fonts or keyboards. With the Asian languages, it's the other way 
round, and there have been big problems."

Bailey sees the project as doing well. Open Office, the free alternative word 
processor, has been done in four languages. Bailey expects six more to be 
completed by June 2005. Mozilla, the attention-grabbing useful web-browser, has 
also been done in seven languages. KDE, the desktop environment, has been done 
in four languages but fell behind because of a "changed idea of what should be 
done". Gnome has been completed in three languages, and the team is busy doing 
Fedora in three more.

What's the secret for the apparent success of such an ambitious and, in many 
ways, daunting project? "Invent what you're doing. Nobody else is doing that. 
We have social objectives, that helps us define what we do. Otherwise you get 
pulled and pushed by everyone."

Our golden rules, says Bailey, is that applications chosen for translation 
should be focussed at the end-user. "Our logic that (it should benefit) the 
people whom language would most affect. Someone who could program could have 
probably mastered English already. Localisation must be aimed at the end-user."

Besides, Bailey clearly favours Free Software. "We're using donor's and state 
money. So, the work should be available to the people and Free Software allows 
us to do that. If we were using proprietary software, we would probably still 
be negotiating for the right to do the software.

Lastly, he prefers cross-platform tools. Ones that are able to run on both 
Windows and GNU/Linux. "Clearly, the majority of people still use Windows. It's 
not our job to change that, though my personal views might be different," says 
he.

Translate.co.za has been helping other countries that would like to learn from 
their experiences. And mistakes.

How easy, or difficult, is it to actually localise a Roman script language? 
Says Bailey: "If you can avoid keyboard stuff, then you just look at 
translating. It's very easy to translate, but difficult to translate well. 
That's where we start looking at things in building glossaries."

"We want people to translate, but not everyone should be doing it. Like the 
Linux kernel -- everyone wants to add code, but not all are allowed to do so 
(unless of proven competence). There are barriers to entry," says he.

Bailey's approach is to usually reject an English word as an easy way out while 
translating. "If it's already in the culture, either in English or corrupted 
English, then it's fine," says he.

To give an example: the team insisted that there was no word for 'password' in 
Zulu. They insisted there was no equivalent for 'proxy-server'. "Proxy is 
someone acting on your behalf. The Zulu have a word for guards that stand at 
the gate of the king's court. People often don't think long enough and hard 
enough about concepts." Some languages don't have a positive connotations for 
witches and wizards. It really means helper.

"In fact, the best people to translate are not computer people but language 
people," he says. Finding people who are excited about their language, or care 
passionately for them, make for the best chances of success.

How long does it take? A good translator can manage approx a thousand words a 
day. So, Open Office, with its 75,000 words would need some 75 working days. 
Then, the documentation is an additional 500,000 words. Quality assurance would 
take some more time.

What's the kind of feedback he has received? Most has come from the language of 
Afrikaans. "Most just accept. But sometimes out of the 800,000 words you 
translated, they'll pick one and haggle. You could have 200 emails about 
opinions and yet no valid feedback," says he, undaunted by what can be a 
thankless job at times.

Earlier, this project was founded by the millionaire-founded Shuttleworth 
Foundation. Now too, it gets some funds from there.

Says Dwayne: "When we started we made lots of mistakes. Nobody else who 
interacts with us makes the same mistakes. Since we are dealing with 11 
languages, the situation in India (with its multiplicity of languages) is the 
closest thing to the work we have to do." But one shortcoming he sees with his 
own project is getting out the results adequately to potential users.

Tags: , ,