Non-sentential utterances in dialogue: a corpus-based study

Dialogue is full of intuitively complete utterances that are not sentential in their outward form, most prototypically the "short answers" used to respond to queries. As is well known, processing such non-sentential utterances (NSUs) is a difficult problem on both theoretical and computational grounds. In this paper we present a corpus-based study of NSUs. We propose a comprehensive, theoretically grounded classification of NSUs in dialogue based on a sub-portion of the British National Corpus (BNC). The study suggests that the interpretation of NSUs is amenable to resolution using a relatively intricate grammar combined with an utterance dynamics approach. That is, a strategy that keeps track of a highly structured dialogue record of entities that get introduced into context as a result of utterances. Complex, domain-based reasoning is not, on the whole, very much in evidence.