Exploration of Language Translation Problems under the Background of Artificial Intelligence Technology

With the deepening of globalized cooperation among countries around the world, the human society's demand for machine translation has increased rapidly, and the advancement of artificial intelligence technology has also put forward new requirements for the quality of machine translation.In this paper, we propose a neural machine translation model with adversarial training algorithm, using the Transformer model as the baseline model, by incorporating prior knowledge in order to strengthen the correlation between the intermediate hidden layers.In addition, besides introducing lexical knowledge into the neural machine translation approach, a neural machine translation model PhraseNet with phrase memory capable of storing bilingual phrases in the form of symbols is also incorporated.Finally, the effectiveness of the language translation model is tested using the NIST corpus.The results show that compared with the baseline model, the BLEU values of this paper's method in the NIST 2010, 2012, and 2016 test sets are improved by 1.81, 2.63, and 2.48, respectively, which significantly outperforms the baseline translation system.It shows that the method is able to bring a large improvement for both language translation accuracy and training execution cycle.Meanwhile, it also has a good guiding significance for the research and development of language translation in the context of the whole artificial intelligence technology.