The current study initially presents a teaching methodology incorporating artificial intelligence in French translation education. Then, we build an unsupervised NMT framework based on pre-training by adopting the Transformer network along with natural language processing tools (particularly BERT). Furthermore, BERT and word embeddings are integrated to generate a two-level representation for the Chinese-to-French NMT framework, thereby facilitating smart French translation. Based on the results, the proposed model successfully utilizes the knowledge learned from pre-trained language models to extract domain-specific characteristics and improve translation performance in various domains, while its utilization proves effective in handling multi-domain translation issues. Besides, after applying the pedagogic model and the proposed method, there is substantial enhancement in the participants’ translation skills, as the mean difference in translation scores between the experimental group and the control group in the pre- and post-tests is significantly larger in the former (P = 0.000, < 0.05).
