Large-scale language models like BERT achieve near-human WSD performance on benchmark tasks by lever...