A Chinese Spelling Check Framework Based on Reverse Contrastive Learning

Chinese spelling check is a task to detect and correct spelling mistakes in Chinese texts. Existing research aims to enhance the text representation and exploit multi-source information to improve the detection and correction capabilities of models, with little attention to improving the ability to distinguish between confusable words. Contrastive learning, aiming to minimize the distance in the representation space between similar sample pairs, has recently become a dominant technique in natural language processing. Inspired by contrastive learning, we present a novel method for Chinese spelling checking, RCL-CSC, which consists of three modules: language representation, spelling check, and reverse contrastive learning. Specifically, we propose a reverse contrastive learning method, which explicitly forces the model to minimize the agreement between similar examples, namely, the phonetically and visually confusable characters. Experimental results show that our method is model-agnostic, thus being combined with existing Chinese spelling check models to yield state-of-the-art performance.