Can you explain the architecture of a transformer model and its significance in language processing?