1. Introduction

Generative artificial intelligence technologies such as ChatGPT have made significant progress in the field of artificial intelligence since the end of 2022, enabling natural human-machine dialogue and completing tasks such as translation, which have had an impact on scientific research, language education, and practice. The United Nations Educational, Scientific and Cultural Organization has released guidelines recommending the establishment of empirical teaching research and methodological standards to validate the effectiveness of generative artificial intelligence in education and research. Domestic and foreign scholars are exploring the application of generative artificial intelligence in translation practice, such as multilingual translation, machine translation, post translation editing, etc. With the iteration and upgrading of GPT technology, ChatGPT-4 has improved in terms of text length, language processing, and generation capabilities, and its performance in translation tasks deserves further research. This article analyzes the application potential of integrating generative artificial intelligence technology into online translation project management platforms through empirical experiments, explores ways to cultivate the artificial intelligence information literacy of student translators, and studies the task process characteristics, speed, task difficulty experience, and translation quality of student translators using this model for manual translation and post editing.

2. The potential of integrating generative artificial intelligence technology with professional translation technology

Explored the potential application of generative artificial intelligence technology, particularly ChatGPT, in the field of translation. Research has shown that ChatGPT exhibits differences in translation quality under different instructions in multilingual translation, especially performing well in resource rich language pairs, but performing poorly in resource scarce or non Latin language families. ChatGPT performs excellently in translation revision, correcting vocabulary, syntax, and semantic errors, and possessing contextual comprehension abilities. Although ChatGPT is helpful for translation tasks, it still requires manual refinement. ChatGPT-3.5 needs improvement in low resource language processing and logical reasoning, while ChatGPT-4 and ChatGPT-4o provide more accurate results. ChatGPT-4 demonstrates advantages in lexical diversity and syntactic complexity in Chinese English news translation. Translators rely on search engines and electronic dictionaries for translation and editing, and the search function of ChatGPT needs further research. At present, ChatGPT is mainly used in system dialog boxes and lacks an integrated translation platform environment. Machine assisted translation technology has become an industry standard configuration. Modern machine assisted translation software and platforms have powerful data storage and project management functions, support multiple file formats and languages, and utilize translation memory and machine translation. Integrating ChatGPT into a machine assisted translation platform can complement each other's strengths and better serve translators. This study utilizes an online translation project management platform embedded with ChatGPT-4 to explore the process, speed, difficulty experience, and translation quality of translators using ChatGPT to complete translation and editing tasks in a professional environment.

3. Experimental design

An experimental design was introduced for the research on the human-machine dialogue translation process of a translation project management platform that integrates generative artificial intelligence technology. The experimental subjects were 58 master's students majoring in translation, who received 8 hours of training to learn the operation methods of the online translation project management platform and the use of ChatGPT. The experimental materials include two equally difficult academic texts used to test the subjects' human-computer interaction and query needs in different translation tasks. The experimental task operation platform is YICAT, which includes original and translated text display, editing function area, function display area, and ChatGPT interface function. The experimental process involves three types of tasks: conventional manual translation, ChatGPT assisted manual translation, and ChatGPT assisted post translation editing. After completing the task, the subjects rated the difficulty of the task and filled out a user experience questionnaire, and the task process was recorded by a screen recording software. Data analysis was conducted using R software to investigate the impact of task patterns on dependent variables (number of human-computer interaction instructions, task speed, task difficulty, and number of translation errors), and professional translators evaluated the quality of translations based on MQM.

4. Experimental results

Analyzed the experimental results of ChatGPT assisted manual translation and post translation editing. In Task 2, the subjects issued 372 instructions, of which 344 were ChatGPT instructions, and the shortcut instructions mainly focused on translation polishing. Task three, the subjects issued 242 instructions, of which 241 were ChatGPT instructions, mainly for translation polishing. Experiments have shown that ChatGPT assisted post editing is significantly faster than conventional manual translation and ChatGPT assisted manual translation, with the latter slightly faster than conventional translation. The subjects subjectively believe that the task difficulty of ChatGPT assisted post editing is the lowest, and the difficulty of ChatGPT assisted manual translation is comparable to conventional translation. In terms of translation quality, the error rate of ChatGPT assisted post editing is significantly lower than that of conventional manual translation and ChatGPT assisted manual translation, with the latter two having comparable quality.

5. Discussion

This chapter explores the application effects of generative artificial intelligence technology, especially ChatGPT, in assisting human translation and post translation editing tasks. Research has found that ChatGPT can significantly improve translation speed while maintaining translation quality comparable to traditional manual translation. In manual translation, participants frequently use ChatGPT for polishing and querying synonyms, while in post translation editing, ChatGPT is mainly used to polish machine translation results. In addition, the study also pointed out the potential of integrating generative artificial intelligence technology with online translation project management platforms, as well as the optimization space for this integration in improving translation efficiency and quality. Specifically, the presetting and optimization of shortcut instructions, the improvement of instruction writing capabilities, and the adjustment of parameters for large language models are all key directions for future technological development. Finally, the chapter emphasizes the demand for big language modeling technology in the field of translation, as well as the importance of strengthening related technology teaching in translation education.

6. Conclusion

This chapter summarizes the research results of integrating generative artificial intelligence technology ChatGPT into an online translation project management platform. Research has found that human-machine dialogue has become a significant feature of translation tasks, and ChatGPT assisted post editing has significant effects in improving speed, reducing difficulty, and ensuring quality. Meanwhile, ChatGPT assisted manual translation has comparable quality to conventional manual translation. The study identified limitations and suggested that future research could be expanded to other major language models, different text types and language directions, as well as using cognitive psychology tools to deeply analyze translators' processing and interaction trajectories.