Translators’ perceptions of literary post-editing using statistical and neural machine translation

Abstract

In the context of recent improvements in the quality of machine translation (MT) output and new use cases being found for that output, this article reports on an experiment using statistical and neural MT systems to translate literature. Six professional translators with experience of literary translation produced English-to-Catalan translations under three conditions: translation from scratch, neural MT post-editing, and statistical MT post-editing. They provided feedback before and after the translation via questionnaires and interviews. While all participants prefer to translate from scratch, mostly due to the freedom to be creative without the constraints of segment-level segmentation, those with less experience find the MT suggestions useful.

The impact of post-editing and machine translation on creativity and reading experience

“…These results seem in line with the view often expressed by translators that they are constrained by MT proposals, which makes MTPE texts less fluent (Moorkens et al 2018). They are also consistent with previous research that indicates that the final quality of post-edited products (in terms of number of errors) is equal to or higher than that of translations produced without any aid (Guerberof-Arenas 2012).…”

“…Their results show that NMT outperforms SMT consistently and by a wide margin. NMT is also the preferred paradigm for the translators who participate in related experiments in which MT is used as a translation aid, although these same translators clearly prefer to translate without MT, as they feel more creative when working on their own (Moorkens et al 2018).…”

“…Research in Natural Language Processing (NLP) has tested the usability of MT for literary texts (Toral, Wieling, and Way 2018), showing that MT post-editing might help literary translators, for example when it comes to productivity. At the same time, however, the perception of translators is that the more creative the literary text, the less useful MT post-editing is (Moorkens et al 2018). In the opinion of many professional translators, MT output needs to be adequate for its purpose and this might not be the case when creativity is considered (Cadwell et al 2016, 245).…”

Apertium-fin-eng–Rule-based Shallow Machine Translation for WMT 2019 Shared Task

“…which is not fluent at all, but doesn't hallucinate gays there so it may be more usable for post-editing. For further research in the problems of NMT for real-world use, see for example Moorkens et al (2018).…”

Translating perceptions and managing expectations: an analysis of management and production perspectives on machine translation

“…Most previous work on attitudes to MT concentrates on translators' perspectives (e.g., Cadwell et al, 2017;Guerberof, 2013;Läubli & Orrego-Carmona, 2017;Moorkens & O'Brien, 2015;Moorkens, Toral, Castilho, & Way, 2018;Rossi & Chevrot, 2019). Some previous studies report translators' views on MT and MT post-editing to be predominantly negative, especially among more experienced professionals (see Läubli & Orrego-Carmona, 2017;Moorkens & O'Brien, 2015).…”