The Research On Neural Machine Translation Based On Generative Adversarial Nets

In the information age,cross-language exchanges between people of different languages have become more frequent.Traditional human translation has gradually failed to meet the current translation requirements due to its translation efficiency.Because machine translation is highly efficient,many researchers have turned their attention to the translation of machine translation.In recent years,with the rapid development of machine learning and deep learning related technologies,deep learning has gradually combined with natural language processing.How to improve the accuracy of neural machine translation through related deep learning techniques is also a problem that researchers have been studying.This thesis studied the Neural Machine Translation model based on the generative adversarial nets.This thesis proposed a method for applying the generative adversarial nets to neural machine translation,to further improve the accuracy of neural machine translation.Building a conditional sequence generative advensarial nets which comprises of two adversarial sub models:a generator and a discriminator.The generator aims to generate sentences which are hard to be discriminated from human-translated sentences,the discriminator uses a convolutional neural network to discriminate,and the goal is to try to distinguish the machine-generated sentences from the humantranslated sentences.Additionally,the static sentence-level BLEU is utilized as the reinforced objective for the generator.During training,both the dynamic discriminator and the static BLEU objective are employed to evaluate the generated sentences and feedback the evaluations to guide the learning of the generator.Experimental results surface,on the English-German and Chinese-English datasets,compared to a single cyclic neural network machine translation model and a BiRNN machine translation model that introduces attention mechanisms,and the state-of-the-art of the translation Model Transformer based entirely on attention mechanisms,after the introduction of the Generative Adversarial Networks,the evaluation index of the translation effect is higher than the previous BLEU point.