Semi-supervised learning and bidirectional decoding for effective grammar correction in low-resource scenarios

Zeinab Mahmoud; Chunlin Li; Marco Zappatore; Aiman Solyman; Ali Alfatemi; Ashraf Osman Ibrahim; Abdelzahir Abdelmaboud

doi:10.7717/peerj-cs.1639

Semi-supervised learning and bidirectional decoding for effective grammar correction in low-resource scenarios

PeerJ Comput Sci. 2023 Oct 24:9:e1639. doi: 10.7717/peerj-cs.1639. eCollection 2023.

Authors

Zeinab Mahmoud¹, Chunlin Li¹, Marco Zappatore², Aiman Solyman³, Ali Alfatemi⁴, Ashraf Osman Ibrahim⁵, Abdelzahir Abdelmaboud⁶

Affiliations

¹ School of Computer Science and Technology, Wuhan University of Technology, Wuhan, Hubei, China.
² Department of Engineering for Innovation, University of Salento, Lecce, Lecce, Italy.
³ School of Software Engineering, South China University of Technology, Guangzhou, China.
⁴ Computer Science, Graduate School of Arts and Sciences (GSAS), Fordham University, New York, United States.
⁵ Advanced Machine Intelligence Research Group, Universiti Malaysia Sabah, Kota Kinabalu, Malaysia.
⁶ Department of Information Systems, King Khalid University, Muhayel Aseer, Saudi Arabia.

Abstract

The correction of grammatical errors in natural language processing is a crucial task as it aims to enhance the accuracy and intelligibility of written language. However, developing a grammatical error correction (GEC) framework for low-resource languages presents significant challenges due to the lack of available training data. This article proposes a novel GEC framework for low-resource languages, using Arabic as a case study. To generate more training data, we propose a semi-supervised confusion method called the equal distribution of synthetic errors (EDSE), which generates a wide range of parallel training data. Additionally, this article addresses two limitations of the classical seq2seq GEC model, which are unbalanced outputs due to the unidirectional decoder and exposure bias during inference. To overcome these limitations, we apply a knowledge distillation technique from neural machine translation. This method utilizes two decoders, a forward decoder right-to-left and a backward decoder left-to-right, and measures their agreement using Kullback-Leibler divergence as a regularization term. The experimental results on two benchmarks demonstrate that our proposed framework outperforms the Transformer baseline and two widely used bidirectional decoding techniques, namely asynchronous and synchronous bidirectional decoding. Furthermore, the proposed framework reported the highest F1 score, and generating synthetic data using the equal distribution technique for syntactic errors resulted in a significant improvement in performance. These findings demonstrate the effectiveness of the proposed framework for improving grammatical error correction for low-resource languages, particularly for the Arabic language.

Keywords: Automatic correction of grammar; Deep learning; Multi-head attention network.

Grants and funding

This work was funded by the Deanship of Scientific Research at King Khalid University through large group Research Project under grant number (RGP.2/175/44). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.