Post-editing of machine translation

Abstract

This chapter analyses the evolutionary process that human post-editing of Machine Translation (MT) output has undergone in previous years. It is well known that post-editing can improve translating productivity as well as target-text quality relative to translation carried out 'from scratch', but postediting has many facets. Initially, it was akin to a step in the MT development pipeline where humans tidied the machine output to make it usable. More recently, post-editing is a professional service with its own international standard. In addition, the focus of translation tasks involving MT is shifting towards more integrated and human-centred environments. The chapter reviews previous postediting research and discusses emerging issues and future directions on the use of MT in professional translation. It outlines how post-editing has matured as a practice and service and adopts an encompassing understanding of post-editing tasks that reflects MT's changing role in professional translation processes.

An Exploration of Post-Editing Effectiveness in Text Summarization

“…Post-editing is a common technique in machine translation, where translators edit the translations produced by automatic methods (opposed to completing the translations manually). It has been shown to increase productivity and improve translation quality (Plitt and Masselot, 2010;Koponen, 2016;Vieira, 2019), particularly when initial translations are good. However, post-editing longer segments can require more cognitive effort to identify errors and plan corrections (Koponen, 2012).…”

“…Taking advantage of the complementary strengths of humans and AI, can they collaborate to improve summarization performance? In the area of machine translation, a common method of human-AI collaboration is human post-editing of AI-generated text, which increases human productivity and improves the quality of translation (Koponen, 2016;Vieira, 2019). However, in spite of its potential impact, studies of post-editing for summarization have been very limited, e.g.…”

The interaction effect between source text complexity and machine translation quality on the task difficulty of NMT post-editing from English to Chinese: A multi-method study

“…The recently emerged paradigm of Neural Machine Translation (NMT) has greatly advanced MT quality, especially in the aspects of fluency or readability of translation output, when compared to the once-dominant Statistical Machine Translation (SMT) (Sennrich et al, 2016;Junczys-Dowmunt et al, 2016). However, recent studies show that NMT also brings new challenges to post-editors by producing unpredictable errors hidden in its fluent texts, which make it more difficult to identify and correct translation errors during PE (Yamada, 2019;Vieira, 2019).…”

Cohesive connectivity in argumentative writing by EFL students

“…From a translation perspective, such texts are specifically inspiring, because they need a grounded domain understanding from the translator's part, who might need to refer to an expert specialized in the field of grounded studies. This is done to conduct the needed inferencing as a linking function and help her/him to practice the target language (TL, hereafter) register-restricted cohesive devices to reformulate target text (TT, hereafter) coherence or unity (Nunes, 2020). The significance of cohesion and coherence in the translational setting has been defined by many scholars and from diverse viewpoints (Cronin, 2020).…”