Research On Neural Machine Translation Based On Hierarchical Analysis Of Syntactic Rules

In recent years,the application of deep learning not only has outstanding performance in image recognition,text generation and other fields,but also has significant effects in natural language processing.Natural language processing systems abound in life.For example,the smart assistant on mobile phone is a natural language processing system.After understanding the user’s request,it feeds back the user’s information through text or voice.Human beings communicate with each other mainly through language to achieve the purpose of sharing knowledge and expanding interpersonal relationships.However,there are currently more than 5,000 languages spoken in the world,and proficient in multiple languages is a difficult task.At this time,Machine translation,as an important branch in the field of natural language processing,can achieve equivalent conversion between different languages while retaining the original semantics,and is an important tool for communication with other parts of the world.At present,Neural machine translation is superior to traditional machine translation methods in various performances,which makes Machine Translation technology reach a new height.However,neural machine translation models take less account of the importance of linguistic knowledge.If sentence structure and semantic information of the language can be incorporated into the translation process,it can play a role in assisting the neural machine translation model,thereby improving the accuracy of translated sentences.This paper proposes a neural machine translation model based on hierarchical analysis of syntactic rules,and improves translation quality by combining improved syntactic analysis method.This paper starts with syntactic analysis.At present,most researches on syntactic analysis are considered from the perspective of characters and words,and there are certain limitations.Therefore,this paper uses the influence of the grammatical structure relationship between sentence components on the part of speech and word order,and proposes a hierarchical analysis algorithm of syntactic rules based on binary and ternary models.The algorithm realizes the hierarchical analysis of sentences.The results show that,the combined model of the syntactic rule hierarchical analysis algorithm is better than the independent binary and ternary word models,and its accuracy and recall rates have reached 82.04% and 80.83%respectively;Compared with the existing sequential labeling model algorithm and lexicalization model algorithm,the accuracy and recall rate are significantly improved.The experimental results show that the hierarchical analysis algorithm of syntactic rules based on binary and ternary model is feasible.Syntactic analysis plays an important role in machine translation.Based on the idea of divide and conquer,this paper improves the hierarchical analysis algorithm of syntactic rules and integrates the algorithm into the neural machine translation model.Through the hierarchical analysis method of syntactic rules,the main frame and the longest phrase of the sentence are identified and separated,and then the Transformer model is trained,and the longest phrase and the main sentence frame are respectively translated and combined to obtain the final translation.the results show that,compared with the baseline system,the proposed method can significantly improve the translation performance on the Chinese-English translation task.The BLEU value of the translation is improved by 0.95 points,and the translation effect is better for long sentences with complex sentence structures.