In the process of machine translation, in order to improve the translation effect, a parallel corpus set that does not participate in training will be introduced into the machine translation model as the verification set. In this way, in the process of training, update every certain step, and a "checkpoint" will be saved in the parameters of the machine translation model. In order to improve the effect of machine translation, the performance of the checkpoint in the whole verification set is often evaluated, so as to select the best checkpoint as the final result of training.
正在翻译中..