Accuracy Improvement of Word Vector-Based Machine Translation Algorithm in the Communication of National Community Awareness

Abstract

In this paper, word vector alignment is introduced into the machine translation model to initialize the word embedding layer of the neural machine translation model for joint training. A monolingual corpus is used to train the model for denoising and self-encoding to improve its encoding ability, decoding ability and translation accuracy. The People's Daily is chosen as the data source, and the study shows that interactive metadiscourse is more frequently used in the news discourse of People's Daily, with 332,433 words, accounting for 81.61% of the total number of metadiscourse, so as to enhance the effect of the communication of national community consciousness. Through the questionnaire survey, it is found that the national community consciousness enhancement score of college students is 4.69, and through the method of this paper, it can provide intelligent solutions for the dissemination of the national community consciousness, which greatly improves the dissemination effect of the excellent traditional Chinese culture.