A Translation Method And System Implementation For English Poem To Chinese Poem Based On Unsupervised Machine

As China has gradually become an important force in the world,the communication of our nation with different cultures becomes more and more important,and the strategy of going global of our culture is full of opportunities and challenges.When culture goes out,it is necessary to introduce and learn first.Poetry,as a treasure in literary works,is deeply loved by people and can build a bridge for cultural learning and communication.The realization of automatic translation of poetry is conducive to selectively learning and absorbing foreign cultures,a comprehensive understanding of "the Moon is not all round abroad ",promoting traditional national culture,and strengthening national cultural self-confidence.Poetry translation is an extremely challenging task,which is faced with such problems as topic drift,the lack of data,loss of metrical patterns,etc.This paper proposes a method for automatic translation of English poetry into Chinese poetry based on unsupervised machine translation and applies it in practice.The main work is as follows:(1)This paper proposes a method of automatic translation of English poetry into Chinese poetry.Based on the previous research on natural language processing,a brand-new process method for Chinese translation of English poems is designed and verified by experiments by selecting valuable poetry research data.The task of creatively translating English poetry into Chinese classical poetry is decomposed into the machine translation stage and the poetry generation stage.We use unsupervised machine translation to solve the problem of the lack of parallel corpus,apply neural network for task of translating English poetry into classical Chinese poetry for the first time,in addition,the recurrent neural network and the attention mechanism applied in Machine Translation are introduced.The experiment effectively verified the feasibility and applicability of the iterative back-translation training strategy in the field of low-resource language translation,and generate a considerable number of Chinese poem translation texts of different training times for English poems,and achieve good evaluation in BLEU and manual evaluation,which provides data support to follow-up research.(2)On the basis of the proposed poetry translation model,the B/S system architecture was adopted to set up the preliminary development and construction the system for poetry translation,and we test system with practical application.After completing the requirements analysis,feasibility analysis and functional design,we further designed the overall architecture of the system as well as translation,evaluation,management and other major functional modules.Finally,the automatic translation system from English poetry to sevencharacter quatrains is realized.This paper proposes an unsupervised poetry translation method and verifies its feasibility.The translation system provides a certain reference value in the field of poetry translation,and provides ideas and references for further research on automatic translation of Song poetry in the future.