Discriminative training for speech recognition is compensating for statistical dependence in the HMM framework.

Dan Gillick, Steven Wegmann, Larry Gillick

Browse the full ICASSP paper archive.