Post-editese: an Exacerbated Translationese

Abstract

Post-editing (PE) machine translation (MT) is widely used for dissemination because it leads to higher productivity than human translation from scratch (HT). In addition, PE translations are found to be of equal or better quality than HTs. However, most such studies measure quality solely as the number of errors. We conduct a set of computational analyses in which we compare PE against HT on three different datasets that cover five translation directions with measures that address different translation universals and laws of translation: simplification, normalisation and interference. We find out that PEs are simpler and more normalised and have a higher degree of interference from the source language than HTs.

Are post-editese features really universal?

“…[7] was unable to prove the existence of post-editese, whether based on human judgement or on automatic classification. On the other hand, [3,4,11,13] have observed certain significant differences between HT and PEMT. In [11], the investigated PEMT corpora exhibit lower lexical richness and lexical density than their HT counterparts, as well as a sentence length that is closer to the source (with the exception of one corpus).…”

“…While these studies have yielded promising results in terms of developing our understanding of post-edited language, it remains difficult to generalise on the characteristics of post-edited texts, due the variety of study designs and features under investigation. In an attempt to establish a stronger theoretical and methodological framework for the study of PEMT in contrast with HT, some researchers [3,4,7,11] have decided to examine the question through the prism of corpus-based translation studies (CBTS) and the well-known concepts of translationese and translation universals (also sometimes called "universal features of translation"), as introduced by [1]. In 2017, [7] put forward the notion of post-editese and defined it as "the expected unique characteristics of a post-edited text that set it apart from a [human] translated text".…”

“…In 2017, [7] put forward the notion of post-editese and defined it as "the expected unique characteristics of a post-edited text that set it apart from a [human] translated text". Following this work, several researchers took part in the quest to identify the universal features of post-edited texts (see [3,4,11,13]).…”

Quality assessment of machine-translated post-edited subtitles: an analysis of Brazilian translators’ perceptions

“…In a study conducted on the PE of literary texts, Castilho and Resende (2022) found out that the PE proximity to the MT output may result in the writers' style being distorted and the final product being influenced. In addition to that, an investigation conducted by Toral (2019) about the manifestation of translationese and post-editese found that post-edited texts tend to be simpler and have a higher degree of interference from the source language when compared to human translation.…”

La traducción automática y la posedición en el ámbito médico

“…También se hace hincapié en que hay mucho más peligro de calcar las estructuras del idioma origen, lo que resta naturalidad al texto meta. La TA plantea, pues, el reto de no caer en lo que actualmente se conoce como post-editese (Toral, 2019), es decir, la pérdida de naturalidad y el aumento de las interferencias del original (Sánchez-Gijón y Piqué, 2020).…”