Machine Translation Training (MTT)

Professional translators face profound challenges brought on by contemporary business practices and digital technologies. They are increasingly asked to correct, or ‘post-edit’, machine translated texts rather than translate from scratch. But despite the productivity gains that often come with post-editing, the process can result in fatigue, boredom and demoralisation (Moorkens and O’Brien 2017). The growing use of cloud-based translation platforms, for both computer-assisted (human) translation, and the post-editing of machine translations, means that the agencies for whom freelancers work, or direct employers, now have a convenient means of tracking all actions performed by translators, leading to what can be perceived as unwelcome surveillance (Kenny 2016). The automated harvesting of human translations on such platforms can lead to data dispossession, despite the human translator’s apparent intellectual property rights (European Commission 2014), and the particular technological configurations with which translators work can lead to ergonomic problems (Ehrensberger-Dow et al. 2016). Yet despite all these challenges, translation is still viewed as a ‘bright outlook’ profession, less prone to technological unemployment than other areas, and set to grow substantially over the next decade (Frey and Osborne 2016; Bureau of Labor Statistics 2016).
Against this background, the current research aims to investigate how translation can be sustainable as a profession in the face of both rapid growth and increasing technologization. In particular we wish to explore models of translator training that prepare students for technologized workflows, providing them with appropriate technical skills and knowledge, but also with the intellectual tools that will enable them to face the ethical, legal and economic challenges ahead. This involves working out how to integrate state-of-the-art neural machine translation (NMT) into translation workflows (a concern within the wider machine translation community) and developing a corresponding pedagogy, as well as understanding the factors that help professional translators accept translation technologies and work with them in ways that do not compromise their well-being, ergonomically or otherwise.
We believe that the models we will explore could also be used to train language teachers and help the general user of MT.

This project started with a PHC Ulysses partnership funded by the French and Irish embassies .