Does syntactic context play a role?
Josh Tauberer, Department of Linguistics Understanding where and why pauses occur in spontaneous speech, and providing a computational model for predicting them, could be useful both in speech recognition applications, which e.g. must distinguish a pause within speech from the silence following the completion of an utterance, and in speech production. The present paper investigates whether pauses between words within sentences from the Switchboard corpus could be predicted on the basis of two types of features: word-level features like POS and syntactic structure features like the category labels surrounding a pause.
