Introduction

While promising technological advances have been made in the areas of speech recognition, generation and understanding, developing usable dialogue systems is still difficult as researchers find themselves in need for models of spoken discourse. The models should cover such phenomena as entrainment, turn-taking or dynamic adaptation by providing insight into both human-human and human-machine interaction and their similarities and differences. The work in the field of human-computer interaction (HCI) aids the creation of human-machine interfaces, allowing for spoken communication which is as close as possible to human-human interaction in natural language. The systems developed are designed to assist people with little or no technological insight: a doctor in his medical decisions, a tourist in a foreign city, a customer taken through a display of products, or a machine operator at his place of work. They are also designed to aid a learner in their attempts at mastering a foreign language by way of interaction with a virtual agent. Besides natural language understanding and generation, such agents are being endowed today with other human-like features in an attempt to make them resemble communication between humans as closely as possible. If systems like these are to be viable, they must be able to process spoken language in its many layers – phonetic, semantic and pragmatic – and, at the same time, perform some useful tasks like giving advice, providing information or teaching a language.