Machine translation post-editing: why training is the bridge to linguistic AI enablement

Andrea Stevens 09 Nov 2022 6 mins
SDL Campus
Knowledge is power. 
 
Admittedly, that’s a bit of a cliché but bear with me. 
 
I’ve been working in machine translation and post-editing for almost 20 years. I have witnessed firsthand the many changes that have culminated in the neural machine translation technology we know today.
 
From implementing rules-based machine translation for a leading manufacturer of agricultural machinery to outsourcing post-editing for travel and automotive clients with statistical machine translation, it’s been a long and fascinating journey.
 
One thing I’ve learned during my years working with MT and linguists is that change is the only constant. It is vitally important to support the people working with machine translation through this change journey by sharing knowledge and expertise.

Knowledge is power. Knowledge shared is power multiplied. (Robert Boyce)

RWS has just been recognized as the world’s best translation company for 2022. Our purpose is unlocking global understanding, and our core business is the transfer of information and, therefore, knowledge from one language and culture to another.
 
Because of the content explosion we’ve been witnessing for several years now, MT and post-editing play a significant role in how knowledge is transferred. Sharing our expertise in MT, post-editing, and linguistic AI is vital to educating translators and project managers working with MT daily.
 
The idea of empowering translators through knowledge sharing is not necessarily a new one – we created our first certification program back in 2014. Since then, the translation ecosystem has continued to evolve and transform, making it more critical than ever to offer training and share experiences.
 
Earlier this year, we released our new eLearning-based certification program for machine translation, post-editing, and linguistic AI. The certification is an entirely free-of-charge program designed to give learners a clear understanding of how MT technology and post-editing services are used in practice, together with a first insight into the linguistic AI features that will shape the future.

Bridging the linguistic AI gap through training

Now more than ever, these linguistic AI features are in focus. With machine translation moving rapidly beyond the traditional boundaries of text-in-text-out automated translation, translator training is once again key to staying ahead of technology developments. Our recent blog 'Empowering localization project managers with linguistic AI' explores what linguistic AI can do for project managers and their customers. Equally, we now need to empower translators to work confidently with linguistic AI.
 
Linguistic AI is the future of localization. If we don’t train translators on what this future looks like, we risk alienating the very group of people that make this industry work. In our blog 'Translators and machine translation – the next chapter,' we explain that technology and innovation require an agile and data-driven approach, something that is very much evident in linguistic AI features such as quality estimation, where a machine-translated segment is automatically attributed a quality level, directing the effort a translator needs to spend. Other capabilities revolve around providing content summarization and insights, the greater adaptation of existing NMT models, and, further in the future, the intelligent routing of work based on content characteristics.
 
Linguistic AI features surround the classic MTPE use case with additional information, context, and insights. Translators have always been at the forefront of working with new technologies, and they now need the space and guidance to explore linguistic AI and the benefits it can bring.

How certification can help

We want to build lasting partnerships with translators, whether they use Language Weaver technology or simply like to learn more about neural machine translation and post-editing, the ecosystem that surrounds the use of MT, and of course, linguistic AI innovation.
 
Our certification course mixes traditional aspects of post-editing with relevant and emerging topics such as:
  • Defining distinct post-editing levels with associated quality matrices
  • Developing an MT content strategy
  • Monitoring the health of an MT or post-editing project through its lifecycle
  • Dealing with gender bias in MT models
  • And finally, exploring linguistic AI capabilities

Not bridging the skills gap carries risks

The primary risk is for linguists, who must ensure their businesses and livelihoods are sustained in a competitive market by keeping up with technological developments. 
 
But there are also risks for the LSP and the enterprise. LSPs that invest heavily in technology could see a slower uptake of new technological solutions with an impact on ROI. And consequently, enterprises and their customer base see an extended time to market and a lower content footprint. It would be negligent if we didn’t cement the bond between people and technology if we are to reap the rewards of innovation and expertise.

Today knowledge has power. It controls access to opportunity and advancement. (Peter Drucker)

Despite the inevitable focus on technology, we know that human experts are at the core of what we do. That is why linguistic AI enablement matters.
 
The certification course is an investment in people, opening an easy and accessible path for growth and development and helping to shape a future that combines the best of people and technology.
Andrea Stevens RWS SDL
Author

Andrea Stevens

Principal Linguistic AI Consultant
Andrea has over 20 years of experience in the localization industry and is now focused on providing consultancy for machine translation, post-editing, and linguistic AI enablement, with an emphasis on training and development.
All from Andrea Stevens