通过文本类型分类评估Google NMT的翻译质量

发布时间：2018-02-26 18:45

本文关键词： Google NMT 文本类型讲话错误分析人工测评　出处：《北京外国语大学》2017年硕士论文　论文类型：学位论文

【摘要】：目前,随着跨文化交际增加、国际竞争不断增强,需要进行翻译的内容越来越多。由于翻译会消耗大量的精力和时间,机器翻译应运而生。机器翻译经历了最初的IBM模型、基于规则(RBMT)的二代翻译方法、基于短语的机器学习(PBMT)等多个时期。2016年9月27日,Google Translate推出新版本的翻译程序GNMT(神经机器翻译系统)引起翻译界的大变革。GNMT基于端到端的自动翻译学习方法,英法、英西的互译准确度已经超过了 90%,中英互译准确率则在80%左右。本文探讨了 Google NMT在政治讲话翻译中的应用问题,发现了在政治讲话翻译中存在的一些主要问题,并提出了一些建议。本文先简要回顾机器翻译研究历史以及机器翻译本身存在的问题、Google Translate的演变以及使用谷歌Google NMT技术进行汉译英的可行性和实用性;之后探讨了文本分类类型及具体的翻译策略;对本论文测评对象《习近平谈治国理政》中的讲话语篇进行介绍和文本类型分类;之后从语篇层面的文内因素和文外因素角度对Google NMT译出的《习近平谈治国理政》译文进行对比评析和分析;最后做出结论,发现Google NMT在应用方面具有一定的普适性,但和真正人工翻译的译文仍有距离。
[Abstract]:At present, with the increase of cross-cultural communication and the increasing international competition, more and more content needs to be translated. Because translation consumes a lot of energy and time, machine translation emerges as the times require. Machine translation has experienced the initial IBM model. Second generation translation method based on RBMT. In September 27th 2016, Translate introduced a new version of the translation program GNMT (Neural Machine Translation system). The accuracy of translation between English and Spanish has exceeded 90%, while that between Chinese and English is about 80%. This paper discusses the application of Google NMT in political speech translation, and finds some main problems in political speech translation. Some suggestions are put forward. Firstly, this paper briefly reviews the history of machine translation research and the problems existing in machine translation. It also discusses the evolution of Translate and the feasibility and practicability of using Google Google NMT technology to translate from Chinese to English. Then it discusses the text classification types and translation strategies, introduces the discourse and text classification in Xi Jinping, the subject of this thesis. Then, from the perspective of intra-text factors and extrinsic factors, the author makes a contrastive analysis and analysis of the translation of "Xi Jinping on governing the country" translated by Google NMT, and finally draws a conclusion that Google NMT has a certain universality in application. However, there is still a distance from the true translation of manual translation.
【学位授予单位】：北京外国语大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：H085

【参考文献】