摘要
人机交互是当下翻译工作的主流趋势,但译前编辑作为其中的重要一环,却未能在译界引起足够的重视。本文通过分析机器翻译错误、译后编辑对象、受控语言规则和已有的译前编辑研究,探讨可行的译前编辑策略,并在此基础上开展译前编辑实验,以《习近平谈治国理政》第四卷的选段为实验对象,以官方译文为参考译文,将DeepL翻译结果作为候选译文,选取BLEU为翻译评估方法,对不同译前编辑策略下的机器译文进行评分,并结合编辑的工作负荷,权衡不同策略的编辑效率。本文根据实验结果,探讨适合人机交互的译前编辑策略,以期提高机器翻译质量。
Human-machine interaction has emerged as the predominant trend in contemporary translation practices,yet pre-editing,a pivotal process within this framework,has not garnered adequate attention within the translation community.This paper delves into the realm of feasible pre-editing strategies by examining machine translation errors,post-editing targets,controlled language rules,and existing research on pre-editing.An experiment is conducted to evaluate pre-editing efficiency,using random samples from Xi Jinping:The Governance of China(Volume 4)for its case study,the official translation serving as the benchmark and DeepL-generated translation as candidate translation.Employing BLEU as the assessment metric for machine translation quality,machine-translated outputs under different pre-editing strategies are evaluated,factoring in the workload imposed on editors to balance the efficiency of each strategy.Drawing upon experimental outcomes,the study explores optimal pre-editing strategies tailored to human-machine interaction,aspiring to enhance the quality of machine translation output.
出处
《翻译研究与教学》
2024年第2期112-121,共10页
Translation Research and Teaching
基金
2022年度国家社会科学基金项目“神经网络机器翻译质量提升研究”(项目编号:22BYY042)的阶段性成果
关键词
人机交互
译前编辑策略
BLEU
效率
human-machine interaction
pre-editing strategies
BLEU
efficiency
作者简介
戴光荣,博士,广东外语外贸大学高级翻译学院教授、博士生导师.研究方向:语料库翻译学、机器翻译人机互动研究、翻译技术教学与研究;罗萱格,广东外语外贸大学高级翻译学院硕士生.研究方向:翻译与人工智能。
