A Practical Study of Translation Wisdom Teaching Based on Semantic Association Network Modeling

Abstract

The traditional translation teaching mode is obsolete and inefficient. This paper designs a model for assessing translation quality based on a semantic association network and innovates the translation teaching mode accordingly. Firstly, a special semantic association processing layer is introduced for the RNN translation quality assessment system based on a cross-language pre-training model with a self-attention mechanism. Then, the semantic correlation between the original text and the translated text is enhanced by the similarity-enhanced splicing mechanism to improve the accuracy of quality assessment so as to construct a translation quality assessment model based on a semantic correlation network. The model is employed in a university’s translation classroom. It is found that the semantic association network similarity enhancement splicing mechanism designed in this paper can improve more than 10 percentage points on the basis of the rest of the algorithms, and the Pearson correlation coefficients are 72.81% and 70.79% compared with the competing models, which are significantly better than the optimal models UNBA (EN-DE)/UNBA (EN-RU). After applying the model, the average score of the experimental class in the semester post-test was 69.19, which was 22.7 points higher than the pre-test. There is an obvious tendency for the distribution of scores to move to the high score range, and the densest distribution of scores becomes 68-70, with 22 students scoring 70 or more and 5 students scoring 80 or more, which is an obvious improvement compared with the semester pre-test. After practicing in a translation classroom, the model in this paper realizes intelligent teaching, saves teachers’ time, and improves the quality of translation teaching.

Keywords

• self-attention mechanism
• recurrent neural network
• semantic association
• translation teaching
• 97c70