Acoustic and language modelling for spoken dialog systems

In the paper, we present acoustic and language modelling procedures developed for Slovenian speech recognition. The continuous speech recognition system is part of the Slovenian spoken dialog system. The Slovenian spoken dialog system was developed within the joint project in multilingual speech recognition and understanding, spoken queries in European languages (SQEL-Copernicus-1634). The developed system can handle spontaneous speech and provide the user with correct information in the domain of air flight information retrieval. Recognition results of Slovenian read and spontaneous speech are presented.