As was made clear by Quebec’s Digital Action Plan for Education and Higher Education, it is no longer a question of if we should be “promot[ing] the use of digital technologies in education,” but rather how. 1 But what if those technologies form a barrier to the development of key competencies? Should translation students be using machine translation at all? Of course they should! And they should be taught to use it appropriately, effectively and responsibly, as professionals.
In 2016, neural machine translation (NMT) shifted the ground beneath our feet. Like most professionals in most industries faced with the disruptive effects of artificial intelligence, reactions ranged from fearing the worst (being replaced by machines) to harbouring unrealistic expectations of limitless productivity gains.
Since then, various industry stakeholders have studied the issue at length in order to shed light on this new reality, bringing much needed nuance to the conversation. For instance, the Ordre des traducteurs, terminologues et interprètes agréés du Québec (OTTIAQ) created a committee to study MT in order to, among other things, reinforce professional autonomy as the common thread throughout these changes. The Translation Bureau, after learning from the problematic rollout of its MT tool in 2016, mandated language solutions provider SDL to produce a translation management system, which will include an NMT module for use by professional translators. Meanwhile, universities have produced research on machine translation literacy and the evaluation of raw machine translation output, among other topics. Several companies have incorporated NMT and post-editing into their processes, and share their findings through various industry platforms, committees and publications. Finally, the Common Sense Advisory has conducted a number of studies, including but not limited to the ones referenced here.
It is now widely accepted that it would be impossible for human translators alone to meet the explosion in demand caused by globalization and big data,2 and that there is room for different levels of quality. In this context, NMT and post-editing are instrumental in making information accessible and meeting the world’s increasingly delineated and segmented translation needs.
Even though the dust has settled on many fronts and the profession in Canada has begun not only to accept this new reality, but even embrace it to capitalize on productivity gains, there is one area that remains somewhat clouded in uncertainty: Should translation students be taught to use MT? And if so, how?
Indeed, there is a very valid argument often made by course lecturers and professors that a too-early introduction of MT into the classroom creates a barrier to the development of the critical skills students need to climb that very steep learning curve that is translation. And the point is important: how can students learn the intricacies of accurate and idiomatic language transfer if an arguably readable or workable draft is already produced for them?
Certain ethical considerations must be raised when approaching the first question. Although universities play a broad societal role, one of their missions is to prepare students to enter the labour force. That is why universities publish placement rates, and why students equate a university education with better opportunities. Given that MT and post-editing play such a vital role in the new market, translation programs must teach students how and when to use these technologies appropriately. To fail to do so is tantamount to a breach in the implicit contract between the university and the student, as the latter will be ill-prepared to find employment and fully exercise their professional autonomy in this new context. However, the breach is also to the detriment of society, as students and graduates who learn to use MT on their own may unknowingly commit privacy violations by entering their clients’ information into free online platforms.
To answer the question of how, it is useful to come back to core skills. Brian Mossop’s Editing and Revising for Translators, 4th edition, offers some guiding principles. Mossop repeatedly posits the distinction between translation and revision as a question of core skills: translation is a writing activity, whereas revision is a reading activity… and working with raw MT output is essentially revision (albeit one that requires retranslation at times), hence the logic of teaching post-editing (defined as the revision of raw MT output) as part of a revision class. In sum, it is important for students to understand the distinction between the two skills, and why it is important to hone them both, and sometimes to hone them separately.
The role of MT and post-editing must also be explained in the context of the new market and evolving business processes. Students must also be helped to understand the limitations of the technologies used and the inherent security risks in using free platforms, not in an effort to make them wary, but rather to make sure they use technologies in an informed way. They must be empowered to act as autonomous professionals, even though algorithms have radically changed the profession.
While a plethora of challenges remain to teaching translation in this new context, the solution lies not in depriving students of this technology, but in teaching them to use it appropriately, effectively and responsibly, as professionals. At a time when translators everywhere are being called on to adapt by developing new skills, translation students must not be left behind.
Meaghan Girard, M. Sc., has been a full-time freelancer for 10 years. Never one to stick to the sidelines, Meaghan is dedicated to helping reshape the translation industry in this period of tremendous change, namely through her research, engagement initiatives and active participation on OTTIAQ’s board of directors as First Vice President. She founded OTTIAQ’s Youth Committee and teaches in the Translation Studies department of the McGill School of Continuing Studies.