Human-Computer Dialogue Systems.

This is a growing area of NLP concerned with the modelling and simulation of human-dialogues, usually with the computer modelling a human dialogue participant. The field thus spans all levels of dialogue analysis and generation: from the processing and understanding of the input to the generation of dialogue response based on stored knowledge bases, pragmatic functions concerning the overall goal or function of the conversation etc

Most research and development has been carried out on task-oriented dialogues in areas where there is some concrete application scenario in view, such as automating call centres that provide information on transportation services or handle travel bookings.

More challenging applications are where the task is less well defined or the information source available to computer is open ended or unstructured. For example, the role of the machine participant may simply be to chat on a range of topics; or it might be to help a user refine their understanding of a topic.

Our contribution

Our emphasis is on the application of machine learning techniques to dialogue management, dialogue segmentation and speech act identification. We have explored the use of information extraction techniques for shallow content extraction from the noisy output of automated speech recognition systems.


Rob Gaizauskas, Mark Hepple, Yorick Wilks


  • Companions - Intelligent, Persistent, Personalised Multimodal Interfaces to the Internet
  • SERA - Social Engagagement with Robots and Agents
  • AMITIES - AMITIES is a new system for multilingual human-computer interaction developed by a consortium of EU and US-based partners.
  • CA4NLP - Engineering Natural Language Interfaces: can CA help?
  • COMIC - COnversational Multimodal Interaction with Computers
  • Converse - Conversation program - Loebner Prize Winner
  • FASiL - Flexible and Adaptive Spoken Language and Multimodal Interface