Endangered Languages Project

May 14, 2015 admin 0 Comments

Team

Though this project was mostly translation, the main thing was that I learned how to use Transifex. I could have also mentioned Gettext, but in fact .po files we translated did not have any difference from any other files having any other extension – it was the task of localization engineers to externalize strings and use .pot files to embed translations, so I was not really exposed to Gettext.

However, it was the first time I used a TMS for a big and complex project. And I got several surprises.

First, the progress in percentage is displayed based on the segment count, not the character count (the system we use in Russia to evaluate the size of the project: 1800 characters with spaces equal 1 page, and we are paid per page). Based on this character count, I had 25 pages to translate.

However, when I started translating, Transifex showed I progressed by 25% within an hour! This is simply impossible, but I was not able to see how many pages I actually translated (definitely less than 6). Within the next 3 hours I progressed by another 50%. As for the remaining 25%, it took me several days, so I consider this imbalance the main drawback of Transifex. The thing is that I always keep track of how many pages I translate every hour as it must not be less than 2, and I check every half-hour to make sure I translated 1 page for the last 30 min – if I have less, I speed up. This is, again, because of payment considerations, and I have had this system for years, but Transifex only counts segments. The result is: if I do not keep track of how many characters I translate, I always slow down, so I could have probably completed this project faster. The solution is to sign off and sign on again – a pop-up window shows how many characters remain untranslated, but signing-off-on every 30 minutes is not a very good idea.

Another drawback, due to the nature of this project (website translation), was that we did not have the context. The majority of segments were just random strings, and sometimes it was simply impossible to understand what a segment is about – some segments were just random words. It was a very funny feeling as I am not used to this lack of context. I had to go to the actual website all the time, but was still unable to find most strings. Because of that, quality could have suffered considerably, but I found a trade-off solution (though far from a perfect one): I was the last translator, all the other languages had already been delivered – so I simply referred to Spanish and Portuguese translations of the same stings and gave Russian translations accordingly. I do not speak Spanish or Portuguese, but I can understand Romance languages. The main problem was that Spanish or Portuguese translators had been in the same position as myself, i.e. they could have misunderstood the meaning of these strings, but at least we had consistency between the 3 translations.

Finally, another interesting finding is that Transifex offers several tabs for Russian if we have to translate variables – I had to populate each tab with different plural forms.