Multi-Modal and Modality Specific Error Handling in the GEMINI Project

GEMINI (Generic Environment for Multilingual Interactive Natural Interfaces) is an EC-funded research project which started in April 2002. The goal of GEMINI is to provide a flexible platform for the generation of applications, able to produce multi-modal and multilingual dialogue interfaces to databases with a minimum of human effort. All description models which are generated by the tools of the platform are based on our newly designed XML-based description language called GDialogXML (Gemini Dialog XML). It is an abstract dialogue description language, which is flexible enough to account for both modality independent and modality specific features of the target language, i.e. VoiceXML or xHTML. The platform will generate high-quality, state-of-the-art applications, e.g. a speech application that includes mixed-initiative dialogues and sophisticated error handling. In this paper, we focus on the treatment of error handling for speech. Error handling for the web will be done in a future stage of the project.