Fast-MD: Fast Multi-Decoder End-to-End Speech Translation with Non-Autoregressive Hidden Intermediates

Hirofumi Inaguma, Siddharth Dalmia, Brian Yan, Shinji Watanabe

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

The multi-decoder (MD) end-to-end speech translation model has demonstrated high translation quality by searching for better intermediate automatic speech recognition (ASR) decoder states as hidden intermediates (HI). It is a two-pass decoding model decomposing the overall task into ASR and machine translation sub-tasks. However, the decoding speed is not fast enough for real-world applications because it conducts beam search for both sub-tasks during inference. We propose Fast-MD, a fast MD model that generates HI by non-autoregressive (NAR) decoding based on connectionist temporal classification (CTC) outputs followed by an ASR decoder. We investigated two types of NAR HI: (1) parallel HI by using an autoregressive Transformer ASR decoder and (2) masked HI by using Mask-CTC, which combines CTC and the conditional masked language model. To reduce a mismatch in the ASR decoder between teacher-forcing during training and conditioning on CTC outputs during testing, we also propose sampling CTC outputs during training. Experimental evaluations on three corpora show that Fast-MD achieved about 2× and 4× faster decoding speed than that of the naïve MD model on GPU and CPU with comparable translation quality. Adopting the Conformer encoder and intermediate CTC loss further boosts its quality without sacrificing decoding speed.

Original languageEnglish
Title of host publication2021 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2021 - Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages922-929
Number of pages8
ISBN (Electronic)9781665437394
DOIs
Publication statusPublished - 2021
Externally publishedYes
Event2021 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2021 - Cartagena, Colombia
Duration: 2021 Dec 132021 Dec 17

Publication series

Name2021 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2021 - Proceedings

Conference

Conference2021 IEEE Automatic Speech Recognition and Understanding Workshop, ASRU 2021
Country/TerritoryColombia
CityCartagena
Period21/12/1321/12/17

Keywords

  • CTC
  • End-to-end speech translation
  • Mask-CTC
  • multi-decoder
  • non-autoregressive decoding

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Signal Processing
  • Linguistics and Language

Fingerprint

Dive into the research topics of 'Fast-MD: Fast Multi-Decoder End-to-End Speech Translation with Non-Autoregressive Hidden Intermediates'. Together they form a unique fingerprint.

Cite this