Fetching the paper…

Improved training of end-to-end attention models for speech recognition · Around