1 Introduction

The discussion focused on the insufficient efficiency of traditional manual translation in the context of globalization, while the development of AI assisted translation technology has made machine translation and post translation editing a research hotspot. Machine translation post editing (MTPE) has received empirical support in improving translation efficiency and ensuring translation quality, especially in general and technical text translation. Scholars have begun to explore the application of MTPE in the field of literary translation, but literary translation is considered the most challenging form due to its particularity. The language and cultural differences in the translation of Chinese literature have increased the difficulty of translation. At present, the application research of MTPE in literary translation mainly focuses on European languages, and there is relatively little empirical research on Chinese English translation. This study adopts a multimodal approach, combined with eye tracking and keystroke recording techniques, to compare and analyze the performance differences of student translators in Chinese novel English translation using manual translation and post editing modes from two dimensions: cognitive process and translation quality. The aim is to answer whether post editing can reduce cognitive load and achieve a quality level equivalent to or better than manual translation, provide empirical evidence for the application of MTPE in the "going global" strategy of Chinese literature, and explore new paths for the integration of translation technology and literary translation.

2 Research Background

Explored the process and translation quality of manual translation and machine translation post editing (MTPE). Research has found that MTPE can significantly shorten task duration in the time dimension, and in the technical dimension, manual translation involves more keyboard input operations, while deletion operations are complex. In terms of cognitive processing, more cognitive resources are invested in the process of manual translation. The comparison of translation quality shows that MTPE has better quality in technical texts, but there is no significant difference in translation quality between manual translation by student translators and MTPE. The application of MTPE in literary translation has received attention, and research involves empirical comparison and ethical exploration. Professional literary translators subjectively prefer manual translation, while MTPE translations score higher in the dimension of reader enjoyment. The influence of MTPE on translator style and translation ethics has also received attention. There are shortcomings in existing research, including a lack of objective empirical studies on translators' cognitive processes, a focus on language pairs within the Indo European language family, and a lack of research on translating from the mother tongue to a foreign language.

3 Research Methods

This chapter introduces an experimental study aimed at comparing the cognitive processes and translation quality under two translation modes: manual translation and machine translation post editing. The study adopts an in subject design, where each participant is required to complete two modes of tasks, with the main independent variable being the translation mode, including manual translation and machine translation post editing. To control text differences, divide the two source texts into four parts as a balancing factor. The dependent variables include process indicators and translation indicators, with the former measuring the level of effort and attention allocation in the translation process, and the latter measured through translation quality ratings. 31 master's students majoring in translation were recruited for the experiment. All participants had Chinese as their first language and English as their second language, passed the English Proficiency Test Band 8, and received training in English Chinese/Chinese English translation. The experimental materials are selected from the opening sections of two contemporary Chinese novels, divided into four consecutive text segments, aiming to maintain similarity in word count, word count, vocabulary richness, and text line count. The experiment was conducted in the laboratory, using an eye tracker to record eye activity and Translog II software to record keyboard activity. Before the experiment, participants were familiar with the blind typing mode and software interface, and underwent eye tracking calibration. The average duration of the experiment was 1.09 hours. Data analysis follows eye movement data filtering standards to eliminate abnormal data. Keyboard activity data was processed using Python, eye movement and keyboard activity data were fitted with a linear mixed effects regression model using R software, and translation quality score data were analyzed using non parametric testing methods.

4 Results

We compared the level of effort and attention to resource allocation in the process of manual translation and post editing through three dimensions: time, technology, and cognitive processing. The results showed that manual translation significantly exceeded post editing in terms of time, input operations, and deletion operations, indicating that manual translation requires more technical effort. In terms of cognitive processing, the average fixation duration and fixation frequency of manual translation are significantly higher than those of post translation editing, especially in the translation area. The evaluation of translation quality found that post editing scored significantly higher than manual translation in terms of accuracy, appropriate selection, and reproduction of the original style, but there was no significant difference between the two in terms of coherence, cohesion, and discourse organization.

5 Discussions

This chapter discusses the differences in cognitive processing and resource allocation between literary manual translation and post translation editing. Research has found that post translation editing mode has significant advantages over manual translation mode in terms of time and cognitive investment, which is consistent with previous research results. The post translation editing mode saves the translator's cognitive resources in the stage of understanding the original text, while in the stage of producing and modifying the translation, manual translation requires more cognitive resources. Screen recording observation shows that in manual translation mode, translators are more inclined to read and modify, while in post editing mode, translators tend to use quick alternation to view machine translated sentences and the original text for modification. In terms of technical effort, manual translation requires more keyboard input and deletion operations than post translation editing. In terms of translation quality, machine translation systems based on neural networks can effectively assist student translators in conveying the meaning of the original text more accurately, and post edited translations score higher in the dimension of original style reproduction. However, the scores of manually translated translations in terms of coherence, cohesion, and discourse organization are slightly higher than those of post edited translations, which may be related to the characteristic of machine translation systems that mainly process translation tasks on a sentence by sentence basis. This study suggests that it is still necessary to conduct a detailed analysis of the translation quality between manual translation and post editing in the context of translating Chinese novels into English, especially at the discourse level.

6 Conclusion

This chapter summarizes an empirical comparative study of two modes of Chinese novel English translation: manual translation and post translation editing. Research has found that post editing can significantly shorten translation time, reduce cognitive load, and outperform manual translation in terms of semantic accuracy, word selection, and original text style reproduction. However, there was no significant difference between the two modes in terms of coherence, consistency, and discourse organization. Research suggests that post editing has potential applications in the field of literary translation, especially for translators in the learning stage. However, there are limitations to the research, such as the sample being student translators, only using Google Translate system, and limited literary text fragments. Future research should recruit professional translators, incorporate diverse texts, and evaluate the potential of large language models in literary translation. Combining reader acceptance research will provide important basis for evaluating the effectiveness of different translation modes.