Chinese variant character restoration based on multidimensional features and local attention

Lingjun Meng; Hongchang Chen; Gengrun Wang

doi:10.1117/12.3036899

21 July 2024 Chinese variant character restoration based on multidimensional features and local attention

Lingjun Meng, Hongchang Chen, Gengrun Wang

Proceedings Volume 13219, Fourth International Conference on Applied Mathematics, Modelling, and Intelligent Computing (CAMMIC 2024); 132191U (2024) https://doi.org/10.1117/12.3036899
Event: 4th International Conference on Applied Mathematics, Modelling and Intelligent Computing (CAMMIC 2024), 2024, Kaifeng, China

Abstract

The restoration of Chinese variant characters refers to the process of converting variant characters in a text to the standard form of Chinese characters. Existing methods mainly treat the variant character restoration as a text error correction or machine translation task. Due to the diversity of variant character forms and the fact that many Chinese characters are visually or phonetically similar but have distinct meanings, Chinese variant character restoration presents challenges. However, traditional text correction models are used to handle aligned sequences, which cannot be directly applied to the variant character restoration. Moreover, using machine translation models for restoration results in low efficiency of restoration and restoration errors. To address these issues, the variant-T model based on multi-dimensional features and local attention is proposed in this paper. The multidimensional word embedding based on the RoCBERT is used to extract multi-dimensional features of text. Further, a model based on local attention is employed, and a local masking mechanism is proposed, which effectively the problem of incorrect restoration. Experimental results show that the word error rates are reduced to 0.51% and 0.62%, and the F1 scores reach 85.2% and 85.7% on the SIGHAN14 and SIGHAN15 datasets. Furthermore, the word error rate is reduced to 0.2% on the variant SMS datasets, which outperforms other models.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Lingjun Meng, Hongchang Chen, and Gengrun Wang "Chinese variant character restoration based on multidimensional features and local attention", Proc. SPIE 13219, Fourth International Conference on Applied Mathematics, Modelling, and Intelligent Computing (CAMMIC 2024), 132191U (21 July 2024); https://doi.org/10.1117/12.3036899

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
10 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Transformers

Data modeling

Error control coding

Performance modeling

Education and training

Semantics

Feature extraction

Show All Keywords

Keywords/Phrases

Search In:

Publication Years