A New Training Pipeline for an Improved Neural Transducer

Albert Zeyer, André Merboldt, Ralf Schlüter, Hermann Ney

The RNN transducer is a promising end-to-end model candidate. We compare the original training criterion with the full marginalization over all alignments, to the commonly used maximum approximation, which simplifies, improves and speeds up our training. We also generalize from the original neural network model and study more powerful models, made possible due to the maximum approximation. We further generalize the output label topology to cover RNN-T, RNA and CTC. We perform several studies among all these aspects, including a study on the effect of external alignments. We find that the transducer model generalizes much better on longer sequences than the attention model. Our final transducer model outperforms our attention model on Switchboard 300h by over 6% relative WER.

 DOI: 10.21437/Interspeech.2020-1855

Cite as: Zeyer, A., Merboldt, A., Schlüter, R., Ney, H. (2020) A New Training Pipeline for an Improved Neural Transducer. Proc. Interspeech 2020, 2812-2816, DOI: 10.21437/Interspeech.2020-1855.

  author={Albert Zeyer and André Merboldt and Ralf Schlüter and Hermann Ney},
  title={{A New Training Pipeline for an Improved Neural Transducer}},
  booktitle={Proc. Interspeech 2020},