Post-Editing Metaphorical Expressions: Productivity, Quality, and Strategies

Abstract

This study aims to explore the impact of neural machine translation (NMT) post-editing on metaphorical expressions from English to Chinese in terms of productivity, translation quality, and the strategies employed. To this end, a comparative study was carried out with 30 student translators who post-edited or translated a text rich in metaphors. By triangulating data from keystroke logging, retrospective protocols, questionnaires, and translation quality evaluation, it was found that: (1) processing metaphorical expressions using NMT post-editing has significantly increased the translators’ productivity compared to translating them from scratch; (2) NMT was perceived to be useful in processing metaphorical expressions and post-editing produced fewer errors in the final output than translation from scratch; (3) different strategies were used to process metaphorical expressions in post-editing and from-scratch translation due to the inherent differences in the two tasks, with “direct transfer” used most frequently in post-editing as translators usually rely on the NMT output to produce the final translation but more balanced strategies adopted in from-scratch translation as they need to seek for different solutions to rendering the metaphorical expressions; the quality of NMT output played a major role in what strategies were adopted to process the metaphorical expressions and their final product quality in post-editing, rather than the conventionality of the metaphorical expressions in the source text. Practical and research implications are discussed.