That's certainly the approach our translation team prefers! Assembling a grammatically-correct statement from multiple discrete tokens is hard in languages with grammatical gender, complex pluralization rules, word ordering, etc. And there's not much benefit of doing it halfway right.