Plain language as a pre-editing tool in machine translation

Abstract

  This paper examines plain language as a framework for pre-editing intended to enhance the quality of draft translations generated by neural machine translation systems. It is based on the premise that textual clarity is not an absolute property but a contextual one, determined by discourse genre and the profile of the intended audience. Based on a recent literature review, plain-language recommendations are identified across four levels of intervention (discursive, orthotypographic, morphosyntactic, and lexico-semantic). Guidelines are assessed in terms of relevance for translation-oriented pre-editing, and compatibility with the preservation of the source text’s fluency and genre conventions. The study results in a proposal comprising 20 rules, ordered according to their sequence of application and defined through parameters that allow their scope to be adjusted to the communicative context. It is argued that the implementation of these rules can enhance the interpretability of the text for the translation system and reduce post-editing effort by shifting part of the cognitive load to intralingual revision in the source language.