Speech Recognition Algorithms Using Weighted Finite-State Transducers