Quality of Machine Translation and the Role of Post-Editing

Abstract

Among various extents that computer technology contributes towards translation, Machine Translation (MT) is the most convenient procedure possible, where a text can be automatically translated using a computer programme. However, despite research and experiments expanding over half a century, Machine Translation has not reached the expectations and work is still going on to make it more reliable and to make its applicability wider. The aim of this study has been evaluating the quality of machine translation output and the role of post-editing. The study was conducted taking an informative text in English as the source text, Sinhalese as the target language, and Google Translate as the Machine Translation system. A group of twenty students reading Translation Studies for their degree at the Sabaragamuwa University of Sri Lanka involved in the study. Their task was to engage in post-editing of the translation output produced by Google Translate, if required. Further, they were requested to note whether a complete Human Translation (HT) of the text is required. According to their response, the target text contained both translation and language blemishes, which need to be eluded by post-editing assigned to a translator. However, no one was of the opinion that the text needs to be retranslated by a human. A few instances had been identified where the programme had failed to grasp the underlying meaning of the source language terms. With regard to the language, errors related to spelling, concord, and word division had been identified. The response was different from student to student, suggesting that the competence level of the post-editor is also of importance. Accordingly, the conclusion is made here that post-editing has a significant role in improving the quality of Machine Translation output.