Faster, but not less demanding

Abstract Recent efforts in debiasing Machine Translation (MT) concentrate on gender-inclusive or neutral language for the translation of sentences containing ambiguous gender entities. Such studies, however, ignore cases that require a specific gender beyond masculine and feminine, i.e. non-binary. By comparing translation with post-editing, the present contribution investigates whether MT can be a useful tool to produce gender-fair translations despite its biased outputs. Twelve language professionals had to either translate or post-edit three English-language texts mentioning non-binary actors into German. For each text, they had to use a different gender-fair language (GFL) approach, i.e. gender-neutral rewording, gender-inclusive characters, and neosystems. Results from screen recordings, retrospective interviews, and target text analysis show that, while post-editing is usually faster than translation, the perceived cognitive effort is generally high with no significant differences emerging in the translation process and, partially, the number of mistakes in the use of GFL.