Using Google as a super corpus to drive written language learning: a comparison with the British National Corpus