How AdaptiveMT is learning?

I would like to know how AdaptiveMT is leaning and reflect translaors' feedback.

 

I already asked "Is AdaptiveMT using NMT system?" on another forum, and Mr. Paul Filkin (SDL Client Services) answered "No".

If not, how AdaptiveMT is learning?

 

Anyone knows the algorithms of AdaptiveMT?

I can't find any information on the Internet.