Recent Advances in End-to-End Automatic Speech Recognition