A system for reducing imprecision in speech interfaces to generalised text input devices

This paper reports on ongoing research at Hatfield, where the Speech and Language Technology Group are developing a series of demonstrator systems in order to investigate human-computer interface design for systems where the input medium is speech. It discusses the ways in which ambiguities in input to speech systems can be avoided, how errors can be corrected, and describes the strategies which are being developed to deal with these problems at Hatfield and how these strategies are being evaulated.