基于Transformer增强架构的中文文本纠错研究

Research on Chinese Text Error Correction Based on Transformer Enhanced Architecture

摘要: 本文将Transformer模型应用于中文文本自动校正领域，并将Transformer模型中不同神经模块的输出动态结合，同时在模型训练时引入课程学习策略，以加快模型收敛速度。实验结果表明，本文所提出的增强模型及在训练中引入的课程学习策略对校正结果的准确率、召回率、纠错F0.5值有较大提升。

Abstract: In this paper, the transformer model is applied to the field of Chinese text automatic correction, and the outputs of different neural modules in the transformer model are dynamically combined. At the same time, the curriculum learning strategy is introduced in the model training to speed up the convergence speed of the model. The experimental results show that the proposed enhancement model and the curriculum learning strategy introduced in the training can greatly improve the accuracy, recall rate and error correction F0.5 value of the correction results.