The machine is blind: Bottom-up feedback on machine translation

What's the impact of machine translation on human translation performance? Check our debrief from the AMTA Machine Translation Conference.

SUBSCRIBE TO OUR BLOG

At the recent AMTA 2020 machine translation (MT) conference, I had the pleasure of representing Amplexor as an invited speaker for the iMpacT workshop.

Amplexor at the AMTA 2020 Machine Translation Conference

With daily keynote speeches delivered by representatives from Google, Facebook AI & Washington University, Microsoft, the ADAPT Centre & Dublin City University, and the U.S. Department of Defense, the conference had no shortage of engaging information on the newest developments from leaders in the world of machine translation.

In addition to these keynote speeches, the event featured a wide range of software demos from leading language solution providers, various expert-led workshops, and a number of practical tutorials. There were also numerous presentations by thought leaders in MT which were generally divided into commercial, research and public sector tracks.

 

The iMpacT workshop

Unlike most of  such conferences, usually focusing on the  technical aspects of machine translation development and implementation, the organizers of the iMpacT workshop aimed for something new. They wanted to give some perspective on the impacts of machine translation implementation on the professional lives of human translators who increasingly work alongside MT engines. Having served in various roles over the last years that allowed me to understand the objectives and motivations of translators and decision-makers in the context of MT, I was glad to accept.

 

The machine is blind: bottom-up feedback on the impact of MT on human translation performance

My presentation focused on certain aspects of machine translation implementation that often remain difficult for language service providers (LSPs) to document and analyze in order to obtain actionable information.

There is no shortage of research and experimental rigor in regard to cost-effectiveness, accuracy, and other technical considerations in the context of machine translation (MT). What is often absent from these discussions, however, are the unfiltered opinions and experiences of the human translators who work with – and are increasingly at risk of being displaced by – machine translation technology.

In my experience, there are real-world consequences to those attitudes, which remain elusive when gathering hard data and evaluating KPIs. In short, translators seem to have varied opinions on the technology itself, but there is a common thread among them when it comes to implementation: creating and updating MT systems and processes without the input of the translators that ultimately work with those systems is a recipe for trouble.

These linguists’ attitudes and the corresponding outcomes have potential to aid in decision-making within companies that choose to implement MT. There are ethical as well as financial considerations involved in this arena, and while at times they appear to be at odds with one another at first glance, closer inspection can reveal that they are often aligned.

 

Adoption of machine translation: key success factors

Some of the possible negative consequences of moving forward with MT in the absence of translator involvement include declining translation quality, delayed or failed adoption of the technology by translators, large-scale brain drain and capacity shortages in translator pools.

More conversely, when translators are included in the aforementioned systems and process development, it can result in rising quality, improved translator loyalty, and an expanding knowledge base.

 

Looking ahead: the future of machine translation

Ultimately, my message to the participants was that any LSP with an interest in ensuring continuity and profitability in their organization should consider the perspective of translators when deciding when and how to implement MT as part of their service offering. These considerations apply to aspects as diverse as collecting and evaluating translator feedback, the specifics of dealing with pricing adjustments for MT post-editing tasks, and how to approach initial MT engine configuration and ongoing retraining.

The time has come to take a broader view of where things are heading for human translators in the face of steadily improving MT quality and availability. Understanding and reacting to the hidden effects of translators’ attitudes toward MT could prove to be of critical importance for language service providers in the future.

Published on    Last updated on 31/12/2020

#Machine Translation, #Translation & Localization

About the author

Rhett Whitaker is a Senior Translator at Amplexor and is based in Berlin, Germany. Rhett specializes in technical translation and translation process optimization and has been part of the team since 2019.

SUBSCRIBE TO OUR BLOG