[EN] In this paper, a study of different approaches to the problem of speech understanding in restricted semantic domains is presented. Two systems based on generative models are proposed and they are compared with a ...