As the Instructor for the Machine Translation Master Class from The Localization  Institute, I am very happy to continue to share with you my ideas about the relationship  between humans and machines. In this issue, we will focus on the role of humans in an  MT deployment process, in which there are three key players, namely, project  managers, linguists, and MT suppliers.

Project managers: how good can you leverage machine power?  

As we grow our company, we are also growing our linguistic assets such as terminology  and translation memory. When our language data have increased greatly in quantity  and quality, it is very natural that we are thinking of more ways to leverage our  resources, and of course, one of the best ways is to apply them to a machine translation  system. As a result, we are bringing a new player into the field, who is so different from  us.

As the manager of the whole team, it makes much more sense that you have a better  knowledge of this new player and also help other team members understand how to  work with it in the MT deployment process. This is not easy, as machines break the  existing balance of your workflow, including how and what you need to communicate  with your team members, your colleagues in other departments, your clients, your  LSPs, and last but not least, your MT suppliers. It is indeed challenging but interesting  to manage this dynamic situation. To this end, possessing the most relevant knowledge  of machine translation and data management is a must. It helps you make plans to  meet the needs and expectations from both humans and machines.

Linguists: a “translation system” to be compared with MT systems?  

As a translator, at some point, if you happen to find yourself being evaluated with some  MT systems, don’t panic or feel bitter. Behind machines, there is such a “collective

wisdom” obtained from millions or billions of language use cases by people. The  amount of data an MT engine can process within a short period of time might be greater  than what a human translator can process in his or her whole life. It is not surprising if  you win or lose this game. This cannot change the fact that each individual is a unique  existence of this world.

In a more practical sense, it might be a good idea that we start to think about how to  make use of what you are good at and which direction we shall move in. This situation  exists not only in the language industry, but for each and every profession. I envision  there will be a time when personalized MT engines become a feasible option for each  linguist, who will give purposes and meanings to these MT “translators”, guiding them to  better represent human experiences. Humans will be focusing on the coordination and  the leadership role in an MT implementation process. Ultimately it is humans who are  the end-users of technology, be it human-guided tools or machine learning tools.  Machines represent extended intelligence of humans. They are about humans, about  every one of us. That is the reason why we are drawn to them.

MT suppliers: how good can you meet client needs?  

For an MT supplier, it is not surprising that a client asks you many questions that you  might feel outside of the territory of machine translation, as MT systems integrated  aspects of other language technologies, such as segmenting, terminology, and  translation memory. While these could be exciting new business opportunities, an MT  supplier should be able to understand client needs and decide on your level of  involvement based on your own business priorities and resources, from just offering MT  interface, methodology and technological support, to MT customization, and to offering  the whole package including MT engines, linguists and QA services.

No matter what, a key foundation is to build trust with your collaborators. For many  people, an MT engine is like a black box and what usually matters to other human  players, including clients, project managers and linguists, is transparency and fairness.  Working in a machine learning environment often means the same resources will be  shared by both humans and machines. MT deployment can be machine oriented or  human oriented. Having a fair and transparent approach will help all stakeholders plan  for both human learning and machine learning.

Takeaways:  

1.  Traditional localization workflow will be updated with MT as a new player
2.  MT systems help human translators explore their internal language model
3.  MT suppliers need to build trust with other human players

If you want to know more about machine translation, sign up for our next Machine Translation Master Class.