TR2019-158

A Comparative Study on Transformer Vs RNN in Speech Applications


    •  Karita, S., Chen, N., Hayashi, T., Hori, T., Inaguma, H., Jiang, Z., Someki, M., Enrique Yalta Soplin, N., Yamamoto, R., Wang, X., Watanabe, S., Yoshimura, T., Zhang, W., "A Comparative Study on Transformer Vs RNN in Speech Applications", IEEE Workshop on Automatic Speech Recognition and Understanding (ASRU), DOI: 10.1109/​ASRU46091.2019.9003750, December 2019, pp. 449-456.
      BibTeX TR2019-158 PDF
      • @inproceedings{Karita2019dec,
      • author = {Karita, Shigeki and Chen, Nanxin and Hayashi, Tomoki and Hori, Takaaki and Inaguma, Hirofumi and Jiang, Ziyan and Someki, Masao and Enrique Yalta Soplin, Nelson and Yamamoto, Ryuichi and Wang, Xiaofei and Watanabe, Shinji and Yoshimura, Takenori and Zhang, Wangyou},
      • title = {A Comparative Study on Transformer Vs RNN in Speech Applications},
      • booktitle = {IEEE Workshop on Automatic Speech Recognition and Understanding (ASRU)},
      • year = 2019,
      • pages = {449--456},
      • month = dec,
      • doi = {10.1109/ASRU46091.2019.9003750},
      • url = {https://www.merl.com/publications/TR2019-158}
      • }
  • Research Areas:

    Artificial Intelligence, Machine Learning, Speech & Audio

Abstract:

Sequence-to-sequence models have been widely used in end-toend speech processing, for example, automatic speech recognition (ASR), speech translation (ST), and text-to-speech (TTS). This paper focuses on an emergent sequence-to-sequence model called Transformer, which achieves state-of-the-art performance in neural machine translation and other natural language processing applications. We undertook intensive studies in which we experimentally compared and analyzed Transformer and conventional recurrent neural networks (RNN) in a total of 15 ASR, one multilingual ASR, one ST, and two TTS benchmarks. Our experiments revealed various training tips and significant performance benefits obtained with Transformer for each task including the surprising superiority of Transformer in 13/15 ASR benchmarks in comparison with RNN. We are preparing to release Kaldi-style reproducible recipes using open source and publicly available datasets for all the ASR, ST, and TTS tasks for the community to succeed our exciting outcomes.