View article

Over the last 30 years, building dialog systems for spontaneous speech has been one of the main focuses of the Speech Processing and Understanding Group of the Chair of Pattern Recognition at the University of Erlangen-Nuremberg. By now, many telephone services based on automatic systems exist and provide cost-efficient services for many scenarios. The available services differ in the design of the dialog between the calling user and the system. Simple systems like the telephone answering machine service of a major German telephone company accept as input modality only input from the keypad. As a consequence, the design of the dialog is rather fixed. The disadvantages are quite obvious: as the system can only react to the entered digits and the user cannot be asked to remember the various meanings of a certain digit at a particular dialog step, the system has to inform the user about the possible …