Search for a command to run...
On the Choice of Modeling Unit for Sequence-to-Sequence Speech Recognition