no code implementations • AACL (NLP-TEA) 2020 • Yikang Luo, Zuyi Bao, Chen Li, Rui Wang
For the correction subtask, we utilize the masked language model, the seq2seq model and the spelling check model to generate corrections based on the detection results.