A Domain Specific Modeling Language for...

Designing a voice user interface (VUI) can become more challenging than designing a graphical user interface (GUI). Without visual interaction elements, a user can be less bound to predefined interaction regulations and restrictions. Concrete user requests and system responses in a dialog strongly depend on the initial intention of the user and the user’s utterances during the dialog, to which the voice-based system has to respond. The aim of this paper is to present an intention-oriented approach to the design of VUIs. This is achieved in particular by defining and applying RIML, a domain specific modeling language, which enables VUI designers to create platform-independent inten-tion models. A RIML model includes to which requests a VUI should respond to, what intentions are involved, how the system handles user requests and responses, and what to do in case of misunderstanding or failure. Based on a RIML model, a voice-based system with a model driven architecture is able to communicate flexibly via its VUI with the user.