Impact Of Real-time Speech Recognition On Interpreting Quality Of E-C Simultaneous Interpretation Of Impromptu Speech:An Empirical Study

Technological advances in the era of artificial intelligence have benefited humanity in all aspects of their lives,including interpretation.As one of the most frequently used tools in interpreting practices,speech recognition technology can provide live captioning for interpreters’ reference.However,the tendency of speakers to stutter,repeat,backtrack,and utter incomplete or illogical sentences makes it harder for speech recognition technology to transcribe the original speech accurately.This study aims to examine whether speech recognition can assist interpreters in interpreting impromptu speech so that interpreters can perform better.This study has chosen Elon Musk’s address at the 2021 Tesla Annual Shareholder Meeting as the experiment material and invited 8 second-year graduate students from the same class at GSTI,BFSU,who demonstrate the same level of simultaneous interpretation and are divided into the experiment group（aided by speech recognition）and the control group（without any assistance）for the purpose of the experiment.The utilized speech recognition technology is iFLYREC,with a transcription accuracy of 92%.After the interpreters have finished their interpretation,4 professional interpreters and 4 pure listeners who do not speak English are invited to evaluate the interpreters’ performance.The results show that the experiment group performs better than the control group from the quantitative perspective,with the interpreters getting higher scores in information accuracy,expression＆delivery,and audience expectations.From a qualitative perspective,recognition errors might not necessarily mislead interpreters.Also,the experiment group yields more accurate renditions when interpreting numbers,jokes and metaphors.However,the monolingual terminology and wrong punctuations in the transcription might backfire by confusing and misleading the interpreters.Speech recognition can,to some extent,improve interpreting quality.However,there is still room for improvement,such as providing bilingual terminology in the realtime transcription,and improving the accuracy of chunking and punctuation.