Structural Sequence to Sequence Learning

AbstractWe propose a generative framework for simultaneous machine translation. Conventional approaches use a fixed number of source words to translate or learn dynamic policies for the number of source words …...