基于GOOGLE神经网络汉英翻译的译后编辑研究——以科技文本为例

A Study on Post-editing in Chinese-English Translation of Science and Technology Texts by Google’s Neural Machine Translation System

近年来,以Google神经网络为代表的机器翻译发展迅猛,且对于某些正式文本的翻译质量高达百分之六七十,但机器译文仍需要人工进行审校与修改,因此以机器翻译为基础的译后编辑需要进一步深入研究。从中国知网(CNKI)数据库选取200篇来自江西理工大学强势学科有色金属方面的SCI、EI科技论文(2015-2017年)的中文摘要进行Google汉英机器翻译,并对译文进行人工分析,同时参考其原有英文摘要进行比较研究。结果显示:Google科技文本汉英机器翻译的错误主要在于词汇、句法、逻辑等方面,且出现错误频率依次递减。针对以上4类错误,笔者进行详细的案例分析,在词汇、句法、逻辑、标点使用方面提出应对措施,为科技文本机器翻译的译后编辑提供实践参考。

In recent years,machine translation led by Google’s Neural Machine Translation System has developed rapidly,with the accuracy of translation being greatly improved to 60 percent to 70 percent.However,human reviewing and revising are still needed in machine translation for greater accuracy,and post-editing in machine translation requires further researches.The author selected 200 Chinese abstracts of SCI and EI scientific papers on non-ferrous metals(20152017年),and translated them into English with Google’s Neural Machine Translation System.The machine translated English texts were compared with the original English abstracts manually,and the results showed that the errors in the machine translated English abstracts existed in four aspects:vocabulary,syntax,logic and others,and the frequency of errors decreased in turn.Based on the case analysis,the author puts forward the resolutions in terms of vocabulary,syntax,logic and others,providing practical reference for post-editing in machine translation of science and technology texts.