Research On Indian English-Chinese Neural Machine Translation With Language Features

With the rapid development of the Internet and information technology,the field of artificial intelligence has attracted more and more attention of people,attracting a large number of researchers and developers.Machine translation is the focus of the field of artificial intelligence research,with important theoretical significance and great application value.Main work of this paper is to explore how to build India English/Chinese bilingual corpus.And how to train India English-Chinese machine translation model through transfer learning based on English-Chinese machine translation model.And how to use language features of English in India combined with neural network,and how to build neural India English-Chinese machine translation system.According to the above problem,this paper's main work is as follows:1.Acquisition of Indian English corpus on social media.Indian English is a kind of typical English variants,India history,geography,political,economic and cultural aspects of influence,especially social media of natural spoken style.So we choose some social media sites.We get text on the page accessed by the crawler programming.Then we can get corpus after processing the text.2.Researching of feature of Indian English language.Indian English is different of American English and British English.The Indian English has its own some linguistic features.Our task is combining linguistic features with neural network efficiently.So we must study the language features of India English.When getting language features of India English,we can probably use put language features of India English into neural network.And improving the performance of the neural machine translation model and improving the quality of translation.3.Application of transfer learning.Compared with American English and British English,Indian English is obvious regional,ethnic,cultural.Using of India English is limited.The focus is not high,it is difficult to get a lot of corpora.So we can only use a small amount of corpora to train model with transfer learning in order to solve the low resource(or language variants)the problem of machine translation model in the training corpus.4.Designing and building machine translation system.In this paper,an Indian English-Chinese neural translation system is designed and built by fusing neural network,transfer learning and language features.The experimental results show that the BLEU value of the translation is improved compared with the current mainstream neural translation system.