Post-editing human translations and revising machine translations : impact on efficiency and quality

Publication type
Publication status
Daems, J, & Macken, L.
Maarit Koponen, Brian Mossop, Isabelle Robert and Giovanna Scocchera
Translation revision and/or post-editing : industry practices and cognitive processes
Routledge (London; New York)
View in Biblio
(externe link)


Post-editing of machine-translation output is generally considered to be a distinct process from the revision of a human-translated text. The main reasons for this assumption are the quality of machine-translation output and the fact that it might be easier to criticize the work of a machine than the work of a fellow human translator. With the global shift of statistical machine-translation systems to neural machine-translation systems, however, the quality of machine-translation output has improved. What was true about differences between revision and post-editing in 2010 might therefore no longer be true today. In addition, translators hired to revise a text are not always aware of the origin of the text. This chapter compares revision and post-editing products made by professional translation agencies when the actual origin of a text corresponded to the instructions they were given (revision of a human translation, post-editing of machine translation) and when the origin did not match the instructions (post-editing of a human translation and revision of machine translation). We look at the number of edits made, the quality of the revision and the optimality of the intervention.