A multi-dimensional representation of context in a speech translation system : a practical approach

Abstract

In this paper we show how the notion of context has been refined in order to fulfill the requirements posed by a natural language processing system. We describe the context model of the speech translation system Verbmobil which has been derived from constraints given by various system components. In Verbmobil context is stored in the dialogue memory which is incrementally constructed by the dialogue processing component. We describe our context model and give sample representations. We also relate our model to state of the art approaches in linguistics.