The main research topic for my PhD thesis is about Domain Adaptation for Statistical Machine Translation. Domain adaptation methods aim to utilize a small data sample exemplifying the target domain for the learning. We approach the problem with our proposed novel latent domain variable models for adaptation (see ).
Of course, in practice, the target domain may not be known at training time or it may change over time depending on user needs. This raises a very hard problem of Online Adaptation. Recently, we are close to deal with the challenge (see ). If you are interested in it, drop me an email!
In general, I am an MT nerd. I enjoy learning/thinking about how to improve translation using Machine Learning/Statistics all my day. That (kind of) stamina is perhaps the strongest ability I have!