Chinese Grammatical Correction Using BERT-based Pre-trained Model

Hongfei Wang, Michiki Kurosawa, Satoru Katsumata, Mamoru Komachi

Abstract: In recent years, pre-trained models have beenextensively studied, and several downstreamtasks have benefited from their utilization.In this study, we verify the effectiveness oftwo methods that incorporate a BERT-basedpre-trained model developed by Cui et al.(2020) into an encoder-decoder model on Chi-nese grammatical error correction tasks. Wealso analyze the error type and conclude thatsentence-level errors are yet to be addressed.