We describe our system by dividing it into two processes, training and decoding. These processes are assumed to be Japanese-English translation.
