A Report On C-E MTPE Of Patent Abstract Translation

The advent of Neural Machine Translation in recent years signals the commercial era of machine translation(MT)is booming soon.Boasting advantages of agility and cost effectiveness,MT is accepted by an increasing number of professionals and companies in the translation business.The translation of patent texts fits in this category.Each year,millions of patents are expected to be translated for information retrieval.The MT is a solution for meeting the translation demand featured by time sensitivity and enormous quantity,especially true in the translation of patent abstracts.Meanwhile,due to the drawbacks of the machine,and the diversity,complexity and dynamism of to-be-translated materials,the quality of the raw output of the MT system is not always satisfying.For better translation quality,at present,a widely adopted format is the combination of machine translation plus post-editing(PE).The format has been verified to be feasible in improving the translation quality.This translation practice report,with the MT system designed by Nanjing Yunzaishangque Information Technology Co.,Ltd.as the case study,performs an in-depth analysis of the MT output,sums up the three major types of errors in the MT,and studies effective ways to improve the translation quality through post-editing under the guidance of the Skopostheorie.This report aims to generalize several effective PE approaches for improving the speed of translation output under the premise of ensuring that the translation quality meets the customer's requirements,and to provide reference for the post-editors in the field of machine translation post-editing(MTPE).