Effective Training of RNN Transducer Models on Diverse Sources of Speech and Text Data.

Takashi Fukuda, Samuel Thomas

Browse the full ICASSP paper archive.