Post-editese: an Exacerbated Translationese

TL;DR

This paper analyzes post-editing machine translation outputs, revealing that they are simpler, more normalized, and exhibit more source language interference compared to human translations, using computational measures across multiple datasets.

Contribution

It introduces a comprehensive computational analysis comparing post-editing and human translation, focusing on translation universals and laws.

Findings

PEs are simpler than HTs

PEs are more normalized than HTs

PEs show higher source language interference

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.