Quantifying Exposure Bias for Neural Language Generation

Open in new window