Speech understanding and spoken dialogue systems

Speech technology is maturing rapidly and attention is switching to the problems of using this technology in real applications, especially applications which allow a human to use voice to interact directly with a computer based information or control system. Apart from the simplest of cases, such systems involve a sequence of interactions between the person and the machine. They therefore involve dialogue and discourse management issues in addition to those associated with prompting the user and interpreting the responses. These systems are often referred to as spoken dialogue systems (SDS). Existing SDS are diverse in nature and there are many papers describing them (J. Siroux and D. Gillet, 1985; D. Luzzati, 1987; J. Peckham, 1991; D. Sadek et al., 1995). Recent overviews of current capability in spoken dialogue systems (SDS) are given by V. Zue (1997) and N.O. Bernsen et al. (1998). The article reviews current practice in spoken dialogue systems. It emphasises the techniques being used in deployed SDS rather than in laboratory research systems but it does include some references and pointers to the latter