In this paper, we present our system for statistical machine translation that is based on weighted finite-state transducers. We describe the construction of the transducer, the estimation of the weights, acquisition of phrases (locally ordered tokens) and the mechanism we use for global reordering. We also present a novel approach to machine translation that uses a maximum entropy model for parameter estimation and contrast its performance to the finite-state translation model on the IWSLT Chinese-English data sets.