How to Improve Automatic Machine Translation Evaluation? Add Humans, Scientists Say

Scientists propose new system to standardize automatic evaluation of NLP systems by adding humans in the loop.