A Comparative Study Of Human Translation And Machine Translation Post-editing In Terms Of MTI's Productivity,Quality And Attitude

Machine translation post-editing is a heated topic in translation research nowadays.The research on post-editing can date back to 1980 s,and progressive results had been made in recent 20 years.Still a large amount of researches is being conducted to investigate whether machine translation post-editing is indeed faster than translation from scratch and how is the quality of the machine translated output.While the benefits of using post-editing had been more or less acknowledged by many researchers,especially for the technical text type,and post-editing is often regarded as a viable alternative to human translation for technical text types in translation workflow.However,it still remains uncertain whether this result hold true for English to Chinese language pair,since there is scarce researches on English-Chinese nationwide or outbound.In addition,we are interested in how is the translation quality of post-editing in contrast with human translation,whether the quality of post-edited output is in comparable with the one of human translated.What's more,there is still a lack of understanding of translators' attitude towards human translation and post-editing,which may exert a great impact upon the performance of translators.This thesis includes a comparative analysis of post-editing and human translation by students translators for technical text types from English-Chinese from the aspect of their productivity,quality,and translators' attitudes.For the productivity analysis,we looked at translation duration time,processing speed and productivity gain.For the quality analysis,we looked at overall quality of human translation and post-editing,their differences and commons in error types,as well as the correlation between productivity and quality.The attitude analysis contained questionnaires about translator's perception and attitude towards machine translation,post-editing and human translation.We setup an experiment with a group of 31 student translators,using a combination of Translog,BBflashback and questionnaires.The translators were asked to complete a human translation and post-editing task from English to Chinese,and to complete three questionnaires in different session.A mixed-methods approach was adopted for the qualitative and quantitative analysis of the data.This thesis intends to answer the following research questions:(1)what are the MTI's overall performance of HT and PE? What are the difference between HT and PE productivity and final quality and what is the correlation between productivity and quality?(2)How do the MTI students' attitudes affect overall performance of HT and PE? What are the MTI student's overall attitudes? How do the MTI student's attitudes affect their productivity and quality?The findings of the study are listed as follows:(1)In terms of productivity,post-editing is found to yield a more productive translation in contrast to human translation,with a productivity of approximately 35.82% on average.Translog data shows post-editing leads to somewhat less total duration time on the whole,with only four exceptions and there is a statistically significant difference in duration time of these two methods after paired sample test.The processing speed of post-editing is indeed faster than that of human translation in general and a statistically significant difference was also found between them.As far as translation quality concerned,it is found that post-editing tends to give rise to a somewhat less errors than human translation,which indicates that PE is more likely to yield a somewhat more convincing translation quality in this study.We found that the top three error categories of HT are addom(addition and omission),extra and mistranslation.Meanwhile,PE quality suffers most from addom,mistranslation and critical mistranslation.Addom is the most common error type both in HT and PE output.The more detailed error analysis indicates that post-editing leads to somewhat better results regarding accuracy,and human translation leads to better results regarding fluency.It is found that there is no correlation between translation productivity and quality in this study by conducting a Pearson correlation test.(2)In terms of translator's attitude,the results show that there is a mixed attitude towards HT and PE.Generally speaking,participants hold a more positive attitude towards PE.However,81% of the participants does not necessary think that PE will replace HT in the future.We found that 65% of the participants agree that the pattern of PE will be adopted in the translation workflow of near future.As for their view upon translation speed,68% of participants hold the view that PE can largely increase translation speed,and 13% of participants think these two methods are of comparable speed.As for the translation quality of these two methods,52% of participants consider that the quality of HT is higher than that of PE,13% of them think another way,and other 35% of participants think they are of comparable quality.