Abstract This paper constructs a diversified Japanese language course recommendation model based on the autoencoder network model using deep learning, which includes stack autoencoder, self-attention mechanism encoder, and relevance decoder. A system for evaluating the quality of Japanese language education was constructed using hierarchical analysis. The correlation between the Japanese listening test and the degree of innovation of diversified Japanese language education courses and the influence of the innovation of diversified Japanese language education courses on students’ Japanese language performance were analyzed, respectively. The study showed that the correlation coefficient between the Japanese listening test score and the diversified Japanese language education program was 0.865, and the students who received the diversified Japanese language education program scored 7 points higher than the students who received the traditional Japanese language education program.
Read full abstract