1 Introduction

ChatGPT is an NLP system based on pre trained corpora and deep learning, with powerful semantic analysis capabilities that can handle multiple languages and adapt to complex contexts. It completes translation and proofreading tasks through question and answer format, providing instant feedback to translation learners and helping to identify and correct errors. This study takes technology translation as the material, constructs an interactive translation practice environment, and explores the impact of ChatGPT feedback on the translation quality of translation majors. Evaluate the quality of manual translation and revised translation based on ChatGPT feedback through comparative experiments.

ChatGPT and Translation Quality Assessment

Explored in depth the application of ChatGPT in translation practice and its impact on translation quality assessment. Firstly, the translation ability of ChatGPT was reviewed, including its performance in different language pairs and professional fields, particularly in handling terminology, proprietary terms, proprietary concepts, and the reliability of Chinese characteristic discourse. Secondly, the role of feedback provided by ChatGPT in the translation learning process was discussed and compared with the effects of traditional teacher feedback and self feedback. Finally, the main methods for evaluating translation quality, including manual evaluation and automatic evaluation indicators, were introduced, laying the foundation for subsequent experimental design and result analysis. Research has shown that ChatGPT performs better than Google Translate and DeepL in translating Chinese literary works, but there are still limitations in specific fields such as maritime translation. The computer feedback provided by ChatGPT demonstrates powerful functionality in foreign language learning due to its interactivity, providing students with targeted modification suggestions and effectively improving the quality of translations. Translation quality evaluation can be divided into two categories: qualitative evaluation and quantitative evaluation. This study mainly adopts quantitative evaluation methods, including manual evaluation of BlackJack translation error types and machine evaluation of BLEU, TER, and METEOR indicators. By constructing a human-computer interactive translation practice environment, this study compared and analyzed the changes in translation quality of students before and after receiving ChatGPT feedback, and explored the impact of this feedback on the quality of students' technology translation. This provides important reference for translation majors to use large language models to assist translation practice and improve translation abilities.

3 Research Design

A research design was introduced on the impact of ChatGPT feedback on the quality of technology translation. The research subjects are 32 third year students majoring in translation, of whom 24 have passed the TEM-4 exam. The evaluators include one professor from the Department of Translation and four MTI Master's students in Translation. The experimental materials are selected from the "Science and Technology Translation Tutorial", including 4 science and technology texts on different topics, 2 Chinese to English translations, and 2 English to Chinese translations. The experiment adopts a quasi experimental design and is divided into five steps: 1) ChatGPT usage training; 2) Manual translation, students can complete the translation task within 60 minutes; 3) Revise the translation based on ChatGPT feedback, and students will optimize the translation within 40 minutes based on the feedback; 4) Machine evaluation, using the Trial Translation Treasure evaluation tool to calculate BLEU, TER, and METEOR values; 5) Manual evaluation, 4 MTI students marked and scored translation errors based on BlackJack error types.

Translation quality evaluation and impact analysis before and after ChatGPT feedback

This chapter analyzed the translation quality of 32 students before and after adopting ChatGPT feedback through machine evaluation and manual evaluation. The machine evaluation uses BLEU, TER, and METEOR indicators, and the results show that ChatGPT feedback can improve student translation quality in most cases. Based on manual evaluation of BlackJack translation error types, it was found that adopting ChatGPT feedback significantly reduced translation errors, especially in spelling, numerical mismatches, and punctuation errors. However, ChatGPT has limited effectiveness in dealing with issues such as proper nouns, mistranslations, and inconsistent styles. The impact of ChatGPT on students' technology translation practice is mainly reflected in four aspects: standardization, correction, optimization, and potential misleading. Its standardization effect helps to enhance students' professional translation abilities, corrective feedback helps students to deeply understand the reasons for errors, and optimization effects help improve the quality and language sensitivity of translations. However, the potential misleading nature of ChatGPT also reminds us that when using AI tools for translation teaching, we need to cultivate students' critical thinking and independent judgment abilities, and cannot rely solely on AI.

5 Conclusion and Prospect

This study constructed an interactive translation practice environment and found that ChatGPT feedback can improve the quality of students' technology translation, especially in reducing spelling, grammar errors, and mistranslations. The study also revealed the multifaceted effects of ChatGPT feedback, including normativity, correctiveness, and optimizability, but also pointed out that it may be misleading when dealing with highly specialized content. The theoretical contribution lies in expanding the scope of translation technology research, deepening the understanding of interactive translation environments, and enriching the methodology of translation quality assessment. On a practical level, research suggests that translation education should integrate AI tools, cultivate students' critical thinking and interdisciplinary knowledge integration abilities, and innovate the translation teaching evaluation system. The limitations of the study lie in the sample size and diversity of text types. Future research can expand the sample size to include texts from other professional fields, and focus on the long-term impact of ChatGPT assisted learning on students' translation abilities. Future research can also explore the potential applications of ChatGPT in areas such as interpreting training and multimodal translation, as well as how to optimize the combination of artificial intelligence and human expertise to establish more efficient translation teaching and practice models.