It does make sense Filipe! It almost looks like you are studying computational linguistics...:=) I do, by the way. That is true as well. So, the more data (parallela corpora), the better the translation. This also means that in 10 years time, the translations will be much better, as google will have much more data than now. The lack of language data is indeed a big problem, especially for exotic, low-resourced-languages.