Keywords:Discourse Markers, Information Structure, Prosody, Corpora
In this paper we propose a different way to face the notion of Discourse Markers (DM), based on prosodic features and using data from spontaneous speech corpora. We show how it is possible to predict DMs through prosodic parameters if we use corpora segmented into tone and information units. We will also show that DMs with different functions feature different prosodic profiles that are the formal counterpart of function in speech. Besides this, we will present the result of a first attempt of a statistical test, and discuss the limits and perspectives of the proposal. Finally, we will present a detailed description of four different DMs.