學生口譯員對 ChatGPT 輔助遠距英進中逐步口譯之感知

Abstract

本實徵研究旨在探討學生口譯員對 ChatGPT 輔助遠距英進中逐步口譯的感知。三十三位實驗參與者來自全臺六所翻譯所，遠端在 Google Meet 上觀看模擬影片並進行逐步口譯。影片模擬結合自動語音辨識（Voice Control for ChatGPT）和機器翻譯（ChatGPT）所進行的逐步口譯。實驗結合問卷和訪談，量性和質性分析可從幾個面向來看。首先，參與者對於六項使用者經驗分數有正面感知，即，認知易用、認知有用、介面友善、工作流暢度、壓力，以及足夠時間使用科技工具。然而，使用者經驗分數與修課時數之間無統計相關性，綜合訪談內容，這意味著在該模式下所需的口譯能力可能與傳統模式不同。此外，參與者對 ChatGPT 翻譯的修改內容，也反映出對翻譯品質的感知。在後續的半結構式訪談中，受訪者分享對於該口譯模式的想法，優點像是抓得到重要細節、節省氣力，缺點則涵蓋分心、版面呈現、翻譯品質、過度依賴和自我懷疑；不同階段使用的口譯策略，也有助於理解認知過程；而關於可能性和普及性、課程訓練建議，以及個人因應的訪談內容，也進一步豐富了實務意涵。

The empirical research aims to investigate how student interpreters perceive the mode of ChatGPT-assisted remote English-to-Chinese consecutive interpreting. The experiment, conducted via Google Meet, involved 33 participants performing consecutive interpreting (CI) with a simulation video. The video demonstrated the application of automatic speech recognition and machine translation. The experiment combined a questionnaire and a semi-structured interview and the analyses revealed several aspects of student interpreters’ perceptions. First, participants showed positive perception in six user experience scores, namely perceived ease of use, perceived usefulness, interface friendliness, workflow, stress, and time for technology tools. However, no correlations were found between user experience scores and class hours. This quantitative analysis, taken together with a qualitative analysis of the interviewees’ responses suggests that the interpreting abilities required in ChatGPT-assisted CI might be different from conventional CI. Additionally, participants’ modifications to ChatGPT-generated Chinese translations reflected their perceptions of the translation quality.In the semi-structured interview, participants offered valuable feedback about the interpreting mode. They highlighted advantages such as capturing critical details and saving effort, while noting disadvantages associated with distraction, display format, translation quality, overreliance, and self-doubt. In addition, interviewees’ interpreting strategies also shed light on the cognitive process in the interpreting mode. The interviews also discussed the potential for adoption and popularity, suggestions for courses and training, and individual responses, which enrich the practical implications of the research. In conclusion, the results of student interpreters’ perceptions of the mode and the subsequent practical implications not only enhance our understanding of the perception of computer-assisted consecutive interpreting (CACI) but serve to identify the strengths and limitations of specific computer-assisted interpreting (CAI) tools.