生成式人工智能与神经网络机器翻译人工译后编辑效率对比研究
An Investigation into the Post-editing Efficiency Between AI-Generated Content and Neural Machine Translation
仲文明 1王迪 1田莎1
作者信息
- 1. 中南大学 外国语学院,长沙 410083
- 折叠
摘要
ChatGPT等生成式人工智能在翻译质量及人机交互方面展现了优势,为机器翻译译后编辑提供了新的机遇.本文以30名译后编辑人员的译后编辑为研究对象,通过眼动追踪、键盘记录和译文质量评估,考察对比生成式人工智能与神经网络机器翻译人工译后编辑效率.研究发现,两者在原文和译文的注意力分配模式方面存在显著差异.相较而言,生成式人工智能译后编辑工作速度更快,技术操作、认知负荷方面的工作付出更少,工作质量更高,但漏译问题突出,术语翻译欠佳.本研究初步验证了基于生成式人工智能的人工译后编辑模式在汉英翻译中的效率优势,为人工智能时代语言服务质量和效率提升提供了实证支持.
Abstract
Artificial Intelligence-Generated Content(AIGC)such as ChatGPT exhibits advantages in translation quality and human-computer interaction,offering opportunities for post-editing in machine translation.This paper investigates the efficiency of post-editing with AIGC compared to neural machine translation,employing eye-tracking,keystroke logging,and translation quality assessment methodologies with a group of 30 post-editors.The study identifies significant differences in attention allocation patterns between the original text and translated text.In comparison,post-editing with AIGC demonstrates faster work pace,reduced demands in technical operations and cognitive load,and higher work quality.However,issues pertaining to omissions and terminology translation are noted.This research provides preliminary validation of the efficiency advantage of AIGC-based post-editing in Chinese to English translation,offering empirical support for the enhancement of language service quality and efficiency in the era of artificial intelligence.
关键词
生成式人工智能/ChatGPT/眼动追踪/认知负荷/译后编辑效率Key words
AIGC/ChatGPT/eye-tracking/cognitive effort/post-editing efficiency引用本文复制引用
出版年
2024
NETL
NSTL
万方数据