Subword segmentation is widely used to address the open vocabulary problem in machine translation .
Number of merges:
32000
Dropout:
0.1
Segment
Outputs