Acoustically discriminative training for language models | IEEE Conference Publication | IEEE Xplore