Integrating Source-Channel and Attention-Based Sequence-to-Sequence Models for Speech Recognition | IEEE Conference Publication | IEEE Xplore