International Open Access Journal Platform

大语言模型驱动的人机交互机器翻译Prompt指令集设计

The Design of Translation Prompts for Human-Computer Interaction Machine Translation Driven by Large Language Models

Way to Translation / 2024,4(2): 31-37 / 2024-06-18 320 920

View Download PDF

• Authors: 陈婧希¹      蔡雨桦¹      ²     
• Information:

  1．中南财经政法大学外国语学院AI翻译产教融合创新实践基地，武汉；
  2．中南财经政法大学会计学院，武汉

• Keywords: Prompt; Machine translation; Large language models Prompt；机器翻译；大语言模型
• Abstract: The rapid surge of the large language models represented by ChatGPT has stricken the translation industry since 2023, providing new possibilities for the improvement of artificial intelligence translation technology. Designing the specific Prompts and clarifying the translation task based on the intractability of ChatGPT can effectively improve the translation quality and ChatGPT’s practicability by stimulating the its potential. The study introduces six design thinking of Prompts and selects sample texts for experiments to explore the effects of different Prompts on the translation effectiveness of ChatGPT. 自2023年以来，以ChatGPT为代表的大语言模型迅速崛起，给翻译行业带来了强烈冲击，为人工智能翻译技术的改进提供了新的可能思路。基于ChatGPT的可交互性，通过设计特定的翻译Prompt指令，明确翻译任务，能有效提升译文质量，激发ChatGPT的翻译潜能，提高其在翻译领域的实用性。本文介绍了六种Prompt的设计思路，并选取样例文本进行实验，以探寻不同Prompt对于ChatGPT翻译效能的影响。
• DOI: https://doi.org/10.35534/wtt.0402005
• Cite: 陈婧希，蔡雨桦．大语言模型驱动的人机交互机器翻译Prompt指令集设计［J］．译道，2024，4（2）：31-37．