Deep Contextualized Acoustic Representations for Semi-Supervised Speech Recognition | IEEE Conference Publication | IEEE Xplore