Enriching Mandarin speech recognition by incorporating a hierarchical prosody model | IEEE Conference Publication | IEEE Xplore