Translating Creative Texts through Machine Translation: Deepl vs. Google Translate

Abstract

Together with advancements in technology, there is an ongoing change in Translation Studies as more and more technological tools are being developed and implemented. As a result of this technological shift, machine translation (MT) gradually becomes an inseparable part of the industry just like CAT tools. Free- to- use MT engines as well as paid and more professional ones have increasingly become available. The role of the translator has also been changing with this shift; the new role human translators assume is not that of a “translator” but of a “post- editor”. However, the usage of MT in translating creative texts is still under question. Creative texts are defined as a broader term than literary texts, encompassing non- fictional works such as philosophical works, didactic books, and self- help books, performative works, and promotional texts (Hadley et al., 2022, p. 6). Within this context, this present study aims to explore the performance of DeepL and Google Translate, the market leader neural machine translation (NMT) engines, in terms of the translation of non- fictional creative texts. A philosophical work, a didactic book, and a self- help book were selected and translated from English into Turkish using DeepL and Google Translate. The raw MT outputs of creative texts were post- edited by five experts in accordance with Translation Automation User Society’s (TAUS) “Human Translation Quality” post- editing guidelines to identify their effects on the productivity of post- editors by measuring their words per hour (WPH) rates and edit- effort rates. Findings have shown that DeepL demonstrates a remarkable achievement with its raw MT output being usable with no or hardly any changes, outperforming Google Translate. The collected data have consistently indicated that in terms of the efficiency of non- fictional creative text translation, DeepL is much better when compared with Google Translate.