Loss-Aware Curriculum Learning for Chinese Grammatical Error Correction
This work addresses Chinese grammatical error correction for language processing applications, representing an incremental improvement over existing methods.
The paper tackles the problem of Chinese grammatical error correction by addressing how varying correction difficulty across samples challenges model learning, proposing a multi-granularity curriculum learning framework that feeds samples from easy to hard and regulates the loss function, resulting in improved performance as demonstrated on various datasets.
Chinese grammatical error correction (CGEC) aims to detect and correct errors in the input Chinese sentences. Recently, Pre-trained Language Models (PLMS) have been employed to improve the performance. However, current approaches ignore that correction difficulty varies across different instances and treat these samples equally, enhancing the challenge of model learning. To address this problem, we propose a multi-granularity Curriculum Learning (CL) framework. Specifically, we first calculate the correction difficulty of these samples and feed them into the model from easy to hard batch by batch. Then Instance-Level CL is employed to help the model optimize in the appropriate direction automatically by regulating the loss function. Extensive experimental results and comprehensive analyses of various datasets prove the effectiveness of our method.