This study investigates the linguistic features of generative artificial intelligence translation (GenAIT) and human translation (HT) of scientific texts from English to Chinese. GenAIT was generated by ChatGPT 3.5, while HTs were performed by 19 Master-of-Translation-and-Interpreting students. Results show GenAIT and HT exhibit distinct linguistic features at both lexical and syntactic levels. HT produced longer texts with lower average word diversity, while GenAIT showed higher accuracy in translating terminology. HT had a greater average sentence count but shorter average sentence length. Human translators more frequently transformed passive voice into active voice than ChatGPT 3.5. The study reveals complementary capabilities in both, suggesting optimization strategies for future translator training, language service providers, and GenAIT development.

Linling Fu, Lei Liu