Search for a command to run...
Improved training of end-to-end attention models for speech recognition