I would definitely vote for the extra wrapper tag. Otherwise your extracting code has no way of telling which tag group it should use to fall back on in case of a missing language.

Having multiple tags with the same ID seems counter-intuitive to me, and probably would confuse parsers as well. ID's are supposed to be unique, I think.