2026/07/24 by Yingyi Zhuang, Q Y Liu, Qianyu Liu
Arts and Humanities · Computer Science · Medicine · #Artificial Intelligence in Healthcare and Education #Natural Language Processing Techniques #Translation Studies and Practices
paper · doi:10.1080/0907676x.2026.2696986
openalex publication_date 2026/07/24 · openalex created_date 2026/07/25 · openalex updated_date 2026/07/25
Rapid advancements in generative artificial intelligence (GAI) have sparked new possibilities for AI translation across diverse domains. Yet, questions persist as to whether these powerful tools can generate high-quality translations, especially in specialized domains involving linguistic and ontological asymmetries, such as legal translation. Collaborative efforts incorporating the expertise of human translators with systematic translation quality assessment and prompt engineering seem to be promising ways of addressing this challenge. This study adopts an integrated approach for enhancing Chinese-to-English legal translation by GAI. Using China’s Civil Code as a case, the study assesses ChatGPT’s English translation through automatic evaluation and error analysis based on the Multidimensional Quality Metrics. Informed by these analyses, the study develops prompt templates based on the RISEN framework to enhance ChatGPT’s translation performance. The results indicate that ChatGPT’s performance improved after the implementation of the optimization prompts, as evidenced by BERT, BLEU, and TER metrics, and further validated through human evaluation. The insights from this study contribute to effective prompt engineering techniques for GAI legal translation while demonstrating the synergy between human expertise and GAI in achieving higher-quality semi-automated translations.