| This study evaluates and compares the output quality of AI and human simultaneous interpretations for speeches of various registers,analyzes the general gap and register-related gaps,and provides suggestions for both groups.This study,based on Sun Haiqin and Zhang Ailing(2015)’s classification,assigns three Chinese speeches of low to high registers for Chinese-English interpreting,with eight students in their second year of a Master of Translation and Interpreting program working as human interpreters,and XiaoduPods as the AI interpreter.Based on Yang Chengshu(2005)’s interpretation quality scoring table,this study proposes an adapted quantitative assessment framework to evaluate and compare the output of human interpreters and the AI interpreter,supplemented by qualitative comparisons.Quantitative comparison reveals that the AI interpreter generally underperforms human interpreters with a relatively narrow gap in the high register speech,and the AI interpreter outperforms certain human interpreters in terms of delivery and language.Qualitatively comparison reveals that speech recognition poses the biggest challenge for the AI interpreter,with challenges differing in various registers:(a)Titles and proper nouns in a formal speech;(b)Loose sentence structure and underlying ideas in a less formal speech;(c)Colloquial style in an informal speech;human interpreters,despite better performances in various aspects,shows a greater propensity of over-translation and backtrack.This study suggests that interpreters study grammar and avoid grammatical mistakes,curb the tendency to over-translate,get a good command of background knowledge,cultivate transferable capabilities,and stay objective towards the development of machine interpretation. |
