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

Open in new window