Improving End-to-end Speech Recognition with Pronunciation-assisted Sub-word Modeling | IEEE Conference Publication | IEEE Xplore