The Xiaomi AI Lab’s Speech Translation Systems for IWSLT 2023 Offline Task, Simultaneous Task and Speech-to-Speech Task

Wuwei Huang; Mengge Liu; Xiang Li; Yanzhi Tian; Fengyu Yang; Wen Zhang; Yuhang Guo; Jinsong Su; Jian Luan; Bin Wang

The Xiaomi AI Lab’s Speech Translation Systems for IWSLT 2023 Offline Task, Simultaneous Task and Speech-to-Speech Task

Wuwei Huang^*
, Mengge Liu
, Xiang Li
, Yanzhi Tian
, Fengyu Yang
, Wen Zhang
, Yuhang Guo
, Jinsong Su
, Jian Luan
, Bin Wang

^*Corresponding author for this work

School of Computer Science and Technology

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

9 Citations (Scopus)

Abstract

This system description paper introduces the systems submitted by Xiaomi AI Lab to the three tracks of the IWSLT 2023 Evaluation Campaign, namely the offline speech translation (Offline-ST) track, the offline speech-to-speech translation (Offline-S2ST) track, and the simultaneous speech translation (Simul-ST) track. All our submissions for these three tracks only involve the English-Chinese language direction. Our English-Chinese speech translation systems are constructed using large-scale pre-trained models as the foundation. Specifically, we fine-tune these models’ corresponding components for various downstream speech translation tasks. Moreover, we implement several popular techniques, such as data filtering, data augmentation, speech segmentation, and model ensemble, to improve the system’s overall performance. Extensive experiments show that our systems achieve a significant improvement over the strong baseline systems in terms of the automatic evaluation metric.

Original language	English
Title of host publication	20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference
Editors	Elizabeth Salesky, Marcello Federico, Marine Carpuat
Publisher	Association for Computational Linguistics
Pages	411-419
Number of pages	9
ISBN (Electronic)	9781959429845
Publication status	Published - 2023
Event	20th International Conference on Spoken Language Translation, IWSLT 2023 - Hybrid, Toronto, Canada Duration: 13 Jul 2023 → 14 Jul 2023

Publication series

Name	20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference

Conference

Conference	20th International Conference on Spoken Language Translation, IWSLT 2023
Country/Territory	Canada
City	Hybrid, Toronto
Period	13/07/23 → 14/07/23

Cite this

Huang, W., Liu, M., Li, X., Tian, Y., Yang, F., Zhang, W., Guo, Y., Su, J., Luan, J., & Wang, B. (2023). The Xiaomi AI Lab’s Speech Translation Systems for IWSLT 2023 Offline Task, Simultaneous Task and Speech-to-Speech Task. In E. Salesky, M. Federico, & M. Carpuat (Eds.), 20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference (pp. 411-419). (20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference). Association for Computational Linguistics.

Huang, Wuwei ; Liu, Mengge ; Li, Xiang et al. / The Xiaomi AI Lab’s Speech Translation Systems for IWSLT 2023 Offline Task, Simultaneous Task and Speech-to-Speech Task. 20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference. editor / Elizabeth Salesky ; Marcello Federico ; Marine Carpuat. Association for Computational Linguistics, 2023. pp. 411-419 (20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference).

@inproceedings{b38281ed169e4146932e8d27cbc7e4af,

title = "The Xiaomi AI Lab{\textquoteright}s Speech Translation Systems for IWSLT 2023 Offline Task, Simultaneous Task and Speech-to-Speech Task",

abstract = "This system description paper introduces the systems submitted by Xiaomi AI Lab to the three tracks of the IWSLT 2023 Evaluation Campaign, namely the offline speech translation (Offline-ST) track, the offline speech-to-speech translation (Offline-S2ST) track, and the simultaneous speech translation (Simul-ST) track. All our submissions for these three tracks only involve the English-Chinese language direction. Our English-Chinese speech translation systems are constructed using large-scale pre-trained models as the foundation. Specifically, we fine-tune these models{\textquoteright} corresponding components for various downstream speech translation tasks. Moreover, we implement several popular techniques, such as data filtering, data augmentation, speech segmentation, and model ensemble, to improve the system{\textquoteright}s overall performance. Extensive experiments show that our systems achieve a significant improvement over the strong baseline systems in terms of the automatic evaluation metric.",

author = "Wuwei Huang and Mengge Liu and Xiang Li and Yanzhi Tian and Fengyu Yang and Wen Zhang and Yuhang Guo and Jinsong Su and Jian Luan and Bin Wang",

note = "Publisher Copyright: {\textcopyright} IWSLT 2023.All rights reserved.; 20th International Conference on Spoken Language Translation, IWSLT 2023 ; Conference date: 13-07-2023 Through 14-07-2023",

year = "2023",

language = "English",

series = "20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference",

publisher = "Association for Computational Linguistics",

pages = "411--419",

editor = "Elizabeth Salesky and Marcello Federico and Marine Carpuat",

booktitle = "20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference",

}

Huang, W, Liu, M, Li, X, Tian, Y, Yang, F, Zhang, W, Guo, Y, Su, J, Luan, J & Wang, B 2023, The Xiaomi AI Lab’s Speech Translation Systems for IWSLT 2023 Offline Task, Simultaneous Task and Speech-to-Speech Task. in E Salesky, M Federico & M Carpuat (eds), 20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference. 20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference, Association for Computational Linguistics, pp. 411-419, 20th International Conference on Spoken Language Translation, IWSLT 2023, Hybrid, Toronto, Canada, 13/07/23.

The Xiaomi AI Lab’s Speech Translation Systems for IWSLT 2023 Offline Task, Simultaneous Task and Speech-to-Speech Task. / Huang, Wuwei; Liu, Mengge; Li, Xiang et al.
20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference. ed. / Elizabeth Salesky; Marcello Federico; Marine Carpuat. Association for Computational Linguistics, 2023. p. 411-419 (20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - The Xiaomi AI Lab’s Speech Translation Systems for IWSLT 2023 Offline Task, Simultaneous Task and Speech-to-Speech Task

AU - Huang, Wuwei

AU - Liu, Mengge

AU - Li, Xiang

AU - Tian, Yanzhi

AU - Yang, Fengyu

AU - Zhang, Wen

AU - Guo, Yuhang

AU - Su, Jinsong

AU - Luan, Jian

AU - Wang, Bin

PY - 2023

Y1 - 2023

N2 - This system description paper introduces the systems submitted by Xiaomi AI Lab to the three tracks of the IWSLT 2023 Evaluation Campaign, namely the offline speech translation (Offline-ST) track, the offline speech-to-speech translation (Offline-S2ST) track, and the simultaneous speech translation (Simul-ST) track. All our submissions for these three tracks only involve the English-Chinese language direction. Our English-Chinese speech translation systems are constructed using large-scale pre-trained models as the foundation. Specifically, we fine-tune these models’ corresponding components for various downstream speech translation tasks. Moreover, we implement several popular techniques, such as data filtering, data augmentation, speech segmentation, and model ensemble, to improve the system’s overall performance. Extensive experiments show that our systems achieve a significant improvement over the strong baseline systems in terms of the automatic evaluation metric.

AB - This system description paper introduces the systems submitted by Xiaomi AI Lab to the three tracks of the IWSLT 2023 Evaluation Campaign, namely the offline speech translation (Offline-ST) track, the offline speech-to-speech translation (Offline-S2ST) track, and the simultaneous speech translation (Simul-ST) track. All our submissions for these three tracks only involve the English-Chinese language direction. Our English-Chinese speech translation systems are constructed using large-scale pre-trained models as the foundation. Specifically, we fine-tune these models’ corresponding components for various downstream speech translation tasks. Moreover, we implement several popular techniques, such as data filtering, data augmentation, speech segmentation, and model ensemble, to improve the system’s overall performance. Extensive experiments show that our systems achieve a significant improvement over the strong baseline systems in terms of the automatic evaluation metric.

UR - https://www.scopus.com/pages/publications/85174902668

M3 - Conference contribution

AN - SCOPUS:85174902668

T3 - 20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference

SP - 411

EP - 419

BT - 20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference

A2 - Salesky, Elizabeth

A2 - Federico, Marcello

A2 - Carpuat, Marine

PB - Association for Computational Linguistics

T2 - 20th International Conference on Spoken Language Translation, IWSLT 2023

Y2 - 13 July 2023 through 14 July 2023

ER -

Huang W, Liu M, Li X, Tian Y, Yang F, Zhang W et al. The Xiaomi AI Lab’s Speech Translation Systems for IWSLT 2023 Offline Task, Simultaneous Task and Speech-to-Speech Task. In Salesky E, Federico M, Carpuat M, editors, 20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference. Association for Computational Linguistics. 2023. p. 411-419. (20th International Conference on Spoken Language Translation, IWSLT 2023 - Proceedings of the Conference).

The Xiaomi AI Lab’s Speech Translation Systems for IWSLT 2023 Offline Task, Simultaneous Task and Speech-to-Speech Task

Abstract

Publication series

Conference

Other files and links

Fingerprint

Cite this