Fetching the paper…

Minimum Word Error Rate Training for Attention-based Sequence-to-Sequence Models · Around