Post-editing Efficiency And Related Factors In English-Chinese Translation

Post-editing is a modern translation method based on machine translation (MT), in which post-editors revise and edit the translated text produced by MT. Different from the traditional translation method, post-editing has been used to meet the demand for rapid translation in this information age. An objective of its development is to save as much time and effort as possible. Researchers have started to investigate post-editing since the 1980s, and made solid progress in the recent two decades, and the wide use of post-editing in the language industry seems to attest to its effectiveness.This study explores the influence of both subjective and objective factors on improving the efficiency of post-editing in order to find out more about post-editing strategies and further improve the efficiency.For the objective part, the quality of MT decides to a large extent the effort post-editors need to make in the process. As MT shapes the process of post-editing, it becomes an essential factor in affecting efficiency. Source text difficulty and sentence length play an important role in the efficiency of translation, thus these two factors were also taken into account as variables in this experiment. MT quality, source text difficulty and sentence length were measured by different tools, and compared in the selection of experiment texts. In addition, post-editors were asked to give personal evaluations on MT quality and source text difficulty in order to test the reliability of both measurement instruments.For the subjective part, post-editors develop their own post-editing strategies based on MT just as what they do in traditional translation. Previous studies show that different post-editing strategies result in different post-editing efficiency. Therefore, this study used PET (Post-Editing Tool), a keystroke-logging program, to record participants' editing behavior, observe their strategies, and try to find out the correlation between strategies and efficiency. However, an efficient post-editing strategy is only useful when the translation output is satisfying. Thus, the quality of all the translation outputs were evaluated by experts. Only qualified texts were selected for further analysis on post-editing strategy and efficiency.In conclusion, it is found that MT quality has a large influence on post-editing efficiency, while the impact of source text difficulty is not significant. As for human evaluation, post-editors' evaluation on MT quality is more reliable than that on source text difficulty. In analyzing post-editing strategy, we find a strong correlation between the number of keystrokes and post-editing time. Also, the efficient and inefficient post-editors are mainly different in the extent to which they utilize the MT output, demonstrating different strategies in deletion, addition, and revision behavior.