Title: The impact of deep learning in Spoken Language Translation

Abstract:
Automatic Speech Recognition (ASR) and Machine Translation (MT) have been 
around since the second half of the past century. We have been predicting 
their technological maturity for a long time and finally it looks like it is 
happening. Recent claims of human parity in MT, with the usual reservations, 
can be added up to the super-human performance claims for ASR. Undoubtedly, 
deep learning has played a role in bringing both technologies to their current 
performance standpoint. The impact of deep learning has not only brought both 
technologies closer to human parity, it has also brought ASR and MT closer in 
terms of models and frameworks. This makes the field of Spoken Language 
Translation (SLT), also a child of the XX century, particularly interesting at 
this point in time. This talk will review the impact of deep learning in SLT 
with emphasis both in ASR and MT components. We will review recent works in 
end-to-end SLT, lattice integration and multi-task integration of ASR and MT 
and discuss potential future directions in the context of human-in-the-loop SLT.