What are machine training tasks and how do I work on them?

As part of our commitment to producing excellent quality translations for all of our customers, we’re always looking to improve our machine translation engines. The best way to do that is with the help of skilled language professionals, which is why we’re launching a new type of task for editors to work on. 

These tasks are called machine training tasks and help us to produce better translations for live chat which, unlike the email and FAQ translations you’ll be used to, doesn’t get seen by a human at any point before delivery.

Here’s an example of one of these tasks. You’ll notice that it’s split into two parts: what the agent has written (which is what you’d normally translate most of the time at Unbabel) and what the customer has written. You only need to work on the agent’s segments; the client’s are just there for context. 

Please remember that rules around High and Medium Confidence segments still apply (you can read about these here). As with these features, you won’t be evaluated on segments that can’t be edited.