Porozumeni telefonickemu dotazu pro automatickou dopravni informacni sluzbu.

The presented thesis concerns the spoken language understanding in an information retrieval dialogue system. There are described methods of use the finite state transducers for conceptual semantic parsing and meaning extraction from speaker's utterances. In this case, the main aim of understanding is the identification of important semantic constituents and their application dependent on the interpretation within the supposed frame structure. The key problem is to create an appropriate mapping between a sequence of the recognized words and a concept based on meaning that represents the real-world entities. We preferred for the given task to use a hierarchical stochastic language model principle rather than a linguistic rule based on parsing methods used in the natural language processing. The used approach is convenient particularly when the treated spoken language doesn't follow a standard grammar due to spontaneous speech phenomena and recognition errors.Available from STL Prague, CZ / NTK - National Technical LibrarySIGLECZCzech Republi