A Research on the Generation Model and Evaluation Model of Chinese Wu-Qing Couplets

The Wu-Qing couplet is a unique form of expressions in couplets and an important part of Chinese traditional culture. It shows the profoundness and interesting features of Chinese. However, there is still no research on the automatic generation of Wu-Qing couplets because the corpus of Wu-Qing couplets is scarce and it cannot support the training process of deep learning. This paper proposes a sequence-to-sequence Wu-Qing couplet generation model based on the idea of transfer learning. At the same time, in order to further improve the effectiveness of the model, based on the characteristics of Wu-Qing couplets, such as coherence, rhythm change, and semantic separation, this paper proposes an evaluation model, which can reorder the output of the generation model for better results. Finally, a complete Chinese Wu-Qing couplet automatic generation system is constructed based on the generation model and the evaluation model.