In this guest blog post, ITI West Midlands Group member Robin Humphrey writes about our Boost Day event on machine translation post-editing that was held on 26 September 2020.
It nearly didn’t happen, but the Northwest Translators’ Network (NWTN) made it and it was so well run by Mike Hanson at the controls. I’d been looking forward to this since March when it was postponed at the last minute. The day’s topic generates much debate — for PEMT (post-editing machine translation) or MTPE, is part and parcel of the translation business and has been for years now.
I’m pushed to do justice here to the level of detail each speaker brought to their own address and what follows is a summary of those presentations.
Before the first talk from Dr Maeve Olohan of The University of Manchester, the topic of PEMT was introduced as representing a current challenge to professional translators. Do we wish to meet the challenge and what happens if we don’t?
From RBMT (rules-based machine translation) and early advances in MT and the Cold War, to today’s AI, neural MT and learning algorithms that help drive MT tools, Maeve presented an authoritative overview of the status of today’s machine translation and how we as a race got here — and what it means for our profession. MT progress was not always brisk. Funding dried up in the 1970s as MT had just not delivered on its promises, but then came SMT — statistical machine translation, which used numerical modelling, vast corpora of both source and target text, statistical analysis and a model of what a translation could or should be. It had little to do with language and everything to do with applied mathematics. In 2007, Google Translate, Bing and Microsoft Translator appeared. 2016 saw the first Artificial neural networks and a free online NMT tool. In 2016, the world was introduced to Google NMT.
MT is fascinating and its developmental history is too, but I can’t go into that here. I encourage anyone who’s interested to read more about it, for what NMT represents in terms of human ability to communicate across borders and cultures is truly captivating. Dr Olohan ended her presentation by saying that MT and PEMT do indeed drive translation’s commodification — that alone is another topic for another day. For an exceptionally detailed look at productivity and quantity in post editing of output from translation memories and MT, I suggest Anna Gerberof Arenas’ PhD thesis. Here’s the link:
I’d also recommend this paper by Anthony Pymin The International Journal for Translation & Interpreting Research entitled What technology does to translating:
Dr Silke Lührmann is an English to German translator, copywriter and researcher. Her presentation focused on NMT Workflow Management & Control and a translator’s need to keep on top of MT workflows. Some agencies will drive relentlessly for greater productivity in order to see a return on their investment. We owe it to ourselves to remember that we’re not machines.
In an attempt to demystify the hype, Silke presented findings of her own research and the results of an NWTN survey. Part of Silke’s message was that PEMT needs to be proofread, as with any text and machines can’t do it all, yet, can they? The problem is therefore that while humans cannot cater for the daily global translation requirement and machines can’t check their own output, expectations need to be managed. Silke’s experience is that many wish to continue to work as translators, that researchers and promoters of NMT tell us that human translation will be phased out one day — we’ve been told that since the mid-1950s — and that the future viability and enjoyability of human translation is under scrutiny. Does that give us a competitive edge or put us on a cliff edge? Well, it places linguists at the centre of the equation and I believe we owe it to ourselves to be as well-informed as Silke is in terms of the role and potential pitfalls of PEMT. We can draw on her positivity and be technically adept linguists, but trainee translators cannot expect to excel if their first port of call is Google Translate.
Silke’s presentation was extremely well-informed and many questions arose from it, ranging from student learning to PEMT quality assurance processes for experienced translators. It was interesting that there was little or no mention of the “avoid PEMT” notion backed by some and in many ways, that doesn’t matter. Silke’s talk did underline skills commodification and translator transformation into machine editors. She reminded us always to keep in mind that translation is a slow process and it is right to take one’s time.
Nick Rosenthal summed up today’s presentations and spoke of a clash of cultures. He also exposed what for me is a common question in our profession — how are MT and PEMT going to affect me? Are they threats or sources of new income? If they are the latter, can I accept that post editing is not translating? Well, there are those who feel that to do PEMT work is to translate twice. The relationship between tech and translating informs the interrogation of technology and language which itself is the subject of much research. During Nick’s discussion I considered my own preparedness to engage in PEMT workflow and accept PEMT assignments.
My circumstances mean I don’t have to chase agencies or clients. I’m fortunate in that currently I spend my working hours on CPD, writing short articles and other projects. So, in effect, I don’t need to take on PEMT to put food on the table and can dispel present concerns about the effect of PEMT on my professional standing — it’s just not part of my professional world. But I do use Linguee (a TM tool) and other online tools when I translate. I therefore engage with computer-aided translation, which is not PEMT. I digress.
Essentially, I recognise the unlikelihood for some of engaging with PEMT. It’s just not their cup of tea. I know translators who still translate from scratch without MT. Conversely, there are many colleagues who embrace it. There are those who heed experienced practitioner advice and who turn to specialisation, because they’re skilled translators and not mediocre ones. They don’t entertain distraction and retain their own sense of linguistic prowess (skills in writing).
I found the NWTN Boost Day on PEMT/MTPE — whichever you prefer — massively stimulating. As with some other colleagues, I just don’t take on PEMT work. PEMT pure is not for me right now. Thank you to NWTN for a wonderfully stimulating webinar. Special thanks to Mike Hanson and his team for avoiding the cancellation of the event altogether back in March, for managing technical operations and for making the event open to non-NWT and members. I loved it!
© ROBIN HUMPHREY
Robin Humphrey, a Master’s in Translation in a European Context from Aston University, MITI and MCIL, is a professional, independent French to English translator and public service interpreter. He focuses on legal, financial, environmental and business texts and manages to write articles for The Bulletin. Robin lives in Solihull, West Midlands and enjoys long countryside walks in Warwickshire (when he is not at his desk), baking and anything daft that makes him laugh.