PUB - Publikationen an der Universität Bielefeld

In spite of their important quantitative role, discourse particles have so far been neglected in automatic speech processing for two reasons: Firstly it is not clear what they may contribute to the aims of automatic speech processing, and secondly their functions seem to vary so much that it seems difficult to identify the information relevant to such aims. The approach presented here therefore attempts to provide automatic means to distinguishing the different readings of discourse particles and to filtering out the information which can be useful for speech understanding systems, employing positional information and their role within a dialogue model of the respective domain, two types of information which are especially easy to obtain. First results indicate that discourse particles can indeed be automatically disambiguated on the basis of the model proposed.