Fetching the paper…

Efficient minimum word error rate training of RNN-Transducer for end-to-end speech recognition · Around