Long time since I have logged in here, so hi again PS
As I was reading this it occurred to me that the approach you are taking is possibly a lot slower than could be taken ...
A very simple PHP program could have a database of all the strings used as they are in English and present any user with the option to add a translation andor verify an existing translation.
By keeping track of how many people verify a particular translation, 'super' users in each language could be presented with a frequency sorted list of translations suggested by the population at large.
The 'super' user could select the translation that they felt was most appropriate and that person could be saved a lot of time typing.
Additionally, when new words or phrases are added, they andor currently untranslated phrases could be presented first to people willing to help, so that the database becomes rapidly populated.
But .. the real advantage would be that language specific XML files could then be auto-generated from the database tables, eliminating the high likelihood of typos.