1 |
Impact of Encoding and Segmentation Strategies on End-to-End Simultaneous Speech Translation
|
|
|
|
In: INTERSPEECH 2021 ; https://hal.archives-ouvertes.fr/hal-03372487 ; INTERSPEECH 2021, Aug 2021, Brno, Czech Republic (2021)
|
|
Abstract:
International audience ; Boosted by the simultaneous translation shared task at IWSLT 2020, promising end-to-end online speech translation approaches were recently proposed. They consist in incrementally encoding a speech input (in a source language) and decoding the corresponding text (in a target language) with the best possible trade-off between latency and translation quality. This paper investigates two key aspects of end-to-end simultaneous speech translation: (a) how to encode efficiently the continuous speech flow, and (b) how to segment the speech flow in order to alternate optimally between reading (R: encoding input) and writing (W: decoding output) operations. We extend our previously proposed end-to-end online decoding strategy and show that while replacing BLSTM by ULSTM encoding degrades performance in offline mode, it actually improves both efficiency and performance in online mode. We also measure the impact of different methods to segment the speech signal (using fixed interval boundaries, oracle word boundaries or randomly set boundaries) and show that our best end-to-end online decoding strategy is surprisingly the one that alternates R/W operations on fixed size blocks on our English-German speech translation setup.
|
|
Keyword:
[INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]; efficient speech technologies; online sequenceto-sequence models; simultaneous speech translation; speech segmentation
|
|
URL: https://hal.archives-ouvertes.fr/hal-03372487/file/2104.14470.pdf https://hal.archives-ouvertes.fr/hal-03372487/document https://hal.archives-ouvertes.fr/hal-03372487
|
|
BASE
|
|
Hide details
|
|
2 |
LeBenchmark: A Reproducible Framework for Assessing Self-Supervised Representation Learning from Speech
|
|
|
|
In: INTERSPEECH 2021: Conference of the International Speech Communication Association ; https://hal.archives-ouvertes.fr/hal-03317730 ; INTERSPEECH 2021: Conference of the International Speech Communication Association, Aug 2021, Brno, Czech Republic (2021)
|
|
BASE
|
|
Show details
|
|
3 |
LeBenchmark: A Reproducible Framework for Assessing Self-Supervised Representation Learning from Speech
|
|
|
|
In: INTERSPEECH 2021: ; INTERSPEECH 2021: Conference of the International Speech Communication Association ; https://hal.archives-ouvertes.fr/hal-03317730 ; INTERSPEECH 2021: Conference of the International Speech Communication Association, Aug 2021, Brno, Czech Republic (2021)
|
|
BASE
|
|
Show details
|
|
4 |
LeBenchmark: A Reproducible Framework for Assessing Self-Supervised Representation Learning from Speech
|
|
|
|
In: INTERSPEECH 2021: ; INTERSPEECH 2021: Conference of the International Speech Communication Association ; https://hal.archives-ouvertes.fr/hal-03317730 ; INTERSPEECH 2021: Conference of the International Speech Communication Association, Aug 2021, Brno, Czech Republic (2021)
|
|
BASE
|
|
Show details
|
|
5 |
ON-TRAC' systems for the IWSLT 2021 low-resource speech translation and multilingual speech translation shared tasks
|
|
|
|
In: Proceedings of the 18th International Conference on Spoken Language Translation, ; International Conference on Spoken Language Translation (IWSLT) ; https://hal.archives-ouvertes.fr/hal-03298854 ; International Conference on Spoken Language Translation (IWSLT), Aug 2021, Bangkok (virtual), Thailand. ⟨10.18653/v1/2021.iwslt-1.20⟩ (2021)
|
|
BASE
|
|
Show details
|
|
6 |
ON-TRAC Consortium for End-to-End and Simultaneous Speech Translation Challenge Tasks at IWSLT 2020
|
|
|
|
In: Proceedings of the 17th International Conference on Spoken Language Translation ; https://hal.archives-ouvertes.fr/hal-02895893 ; Proceedings of the 17th International Conference on Spoken Language Translation, Jul 2020, Seattle, WA, United States. pp.35-43, ⟨10.18653/v1/2020.iwslt-1.2⟩ (2020)
|
|
BASE
|
|
Show details
|
|
|
|