In the article is discussed the problem of the aceessof broad masses of the population to information. The use of speech recognition in the interactive information systems is the most suitable way to make these systems more open and easily available for people. The purpose of our work was to create the functioning speech interfaces in Russian language to information and service systems and to analyze the results. Our efforts resulted in creation of the first voice servers for interactive information and services systems in Russia: computer reservation system Sirena, a system for dispatching of orders for taxi delivery, and a banking system. We developed several analytical models that allow dialogue optimization and control. One of our models, which enable the calculation of the mean number of retries and appropriate time losses in different conditions, is presented in this paper.