This project first summarizes the characteristics of ethnic multilingual languages and uses web crawling technology to obtain corpus data. The data are preprocessed and divided into training set, validation set and test set according to the ratio of 7:2:1, and the encoder, attention mechanism and Transformer model are used to complete the design of the machine translation model, and the optimal model parameters are determined through model training. Based on the traditional multilingual discourse system, the machine translation model is introduced, and the multilingual discourse system under the machine translation theory is explored and analyzed. The BLEU values of this paper’s model (0.904, 0.935, 0.945, 0.946 for the four ethnic discourse corpus) are better than the other four models (CNN, RNN, SVM, LSTM), and these results reflect the effectiveness of this paper’s model in the ethnic multilingual discourse translation task. In addition, it is found that there is no significant difference between the control group before and after the intervention (P=0.119>0.05), while there is a significant difference between the assessment indexes before and after the intervention of the control group in the experimental group (P=0.001<0.05), which fully verifies that the machine translation model has an excellent performance performance in constructing the multilingual discourse system of ethnic community consciousness.