Improving the efficiency of speech interface with the use of cognitive and linguistic knowledge

In the modern information society a huge amount of voice data is accumulated daily, and the automation of processing can provide fundamentally new opportunities for the analysis of management decision-making automation systems, therefore the development of methods and means of automating speech information recognition and analysis systems is of great scientific and state significance. Information technology plays an enormous role in modern society. Specifically, the most natural form of communication between a man and an information environment is a speech interface. Unfortunately, not all of the results of novell theoretical and applied research are up taken to design speech interfaces. This happens because there are no easily identifiable criterias, the solutions have strong context dependence, the problems itself are hard to formalize, the problems are very labor-intensive. At the same time, this research is of high demand since speech interfaces become increasingly widespread and the developers need practical guides. In this paper we investigate the problems of speech interfaces. Specifically, we demonstrate new methods to solve some problems that arise while designing man-machine speech technology-based interfaces. Finally, we formulate the rules, that are based on our research and experience in building speech-based applications, to design effective interfaces.