BIT-ACT: An Ancient Chinese Translation System Using Data Augmentation

Li Zeng; Yanzhi Tian; Yingyu Shan; Yuhang Guo

BIT-ACT: An Ancient Chinese Translation System Using Data Augmentation

Li Zeng
, Yanzhi Tian
, Yingyu Shan
, Yuhang Guo^*

^*Corresponding author for this work

School of Computer Science and Technology

Beijing Institute of Technology

Research output: Contribution to conference › Paper › peer-review

2 Citations (Scopus)

Abstract

This paper describes a translation model for ancient Chinese to modern Chinese and English for the Evahan 2023 competition, a subtask of the Ancient Language Translation 2023 challenge. During the training of our model, we applied various data augmentation techniques and used SiKu-RoBERTa as part of our model architecture. The results indicate that back translation improves the model's performance, but double back translation introduces noise and harms the model's performance. Fine-tuning on the original dataset can be helpful in solving the issue.

Original language	English
Pages	43-47
Number of pages	5
Publication status	Published - 2023
Event	1st Workshop on Ancient Language Translation, ALT 2023, co-located with 19th Machine Translation Summit, MTS 2023 - Macau, China Duration: 5 Sept 2023 → …

Conference

Conference	1st Workshop on Ancient Language Translation, ALT 2023, co-located with 19th Machine Translation Summit, MTS 2023
Country/Territory	China
City	Macau
Period	5/09/23 → …

Cite this

@conference{f2a4f1922a714b66a2eac3e939e9b4e9,

title = "BIT-ACT: An Ancient Chinese Translation System Using Data Augmentation",

abstract = "This paper describes a translation model for ancient Chinese to modern Chinese and English for the Evahan 2023 competition, a subtask of the Ancient Language Translation 2023 challenge. During the training of our model, we applied various data augmentation techniques and used SiKu-RoBERTa as part of our model architecture. The results indicate that back translation improves the model's performance, but double back translation introduces noise and harms the model's performance. Fine-tuning on the original dataset can be helpful in solving the issue.",

author = "Li Zeng and Yanzhi Tian and Yingyu Shan and Yuhang Guo",

note = "Publisher Copyright: {\textcopyright} 2023 The Authors.; 1st Workshop on Ancient Language Translation, ALT 2023, co-located with 19th Machine Translation Summit, MTS 2023 ; Conference date: 05-09-2023",

year = "2023",

language = "English",

pages = "43--47",

}

TY - CONF

T1 - BIT-ACT

T2 - 1st Workshop on Ancient Language Translation, ALT 2023, co-located with 19th Machine Translation Summit, MTS 2023

AU - Zeng, Li

AU - Tian, Yanzhi

AU - Shan, Yingyu

AU - Guo, Yuhang

PY - 2023

Y1 - 2023

N2 - This paper describes a translation model for ancient Chinese to modern Chinese and English for the Evahan 2023 competition, a subtask of the Ancient Language Translation 2023 challenge. During the training of our model, we applied various data augmentation techniques and used SiKu-RoBERTa as part of our model architecture. The results indicate that back translation improves the model's performance, but double back translation introduces noise and harms the model's performance. Fine-tuning on the original dataset can be helpful in solving the issue.

AB - This paper describes a translation model for ancient Chinese to modern Chinese and English for the Evahan 2023 competition, a subtask of the Ancient Language Translation 2023 challenge. During the training of our model, we applied various data augmentation techniques and used SiKu-RoBERTa as part of our model architecture. The results indicate that back translation improves the model's performance, but double back translation introduces noise and harms the model's performance. Fine-tuning on the original dataset can be helpful in solving the issue.

UR - https://www.scopus.com/pages/publications/85185218119

M3 - Paper

AN - SCOPUS:85185218119

SP - 43

EP - 47

Y2 - 5 September 2023

ER -

BIT-ACT: An Ancient Chinese Translation System Using Data Augmentation

Abstract

Conference

Other files and links

Fingerprint

Cite this