Abstract: 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(2015 -2017 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.

Key words: post-editing, neural machine translation, science and technology text, Google, corpus translation