Application of BERT Model in Chinese Language and Literature Text Generation and Evaluation

For traditional bidirectional encoder representations from Transformers (BERT) models, their main function is for text understanding tasks; however, as for how to apply them to text generation, especially in the generation of Chinese language and literature texts, BERT models still face many difficulties and challenges. A method for generating Chinese language and literature texts based on the BERT model was proposed, aiming to improve the quality and literary quality of the generated texts. Firstly, the BERT model was adaptively improved by introducing domain knowledge and customized fine-tuning strategies to enhance its expressive power in literary creation. Secondly, an experimental framework was designed to evaluate the performance of the improved BERT model in literary text generation tasks, such as poetry and novels. The experimental results showed that the optimized BERT model performs well in generating smooth, emotionally expressive, and literary quality text. Research has shown that BERT-based text generation methods, especially in the field of Chinese language and literature, have strong potential for application.