A discriminative model for continuous speech recognition based on Weighted Finite State Transducers | IEEE Conference Publication | IEEE Xplore