Split Utterances in Dialogue: a Corpus Study
Purver Matt, Howes Christine, Healey Pat and Gregoromichelaki Eleni
SIGDIAL Workshop on Discourse and Dialogue (SIGDIAL 2009)
Queen Mary University of London, September 11-12, 2009
This paper presents a preliminary English corpus study of split utterances (SUs), single utterances split between two or more dialogue turns or speakers. It has been suggested that SUs are a key phenomenon of dialogue, which this study confirms: almost 20% of utterances were found to fit this general definition, with nearly 3% being the between-speaker case most often studied. Other claims/assumptions in the literature about SUs' form and distribution are investigated, with preliminary results showing: splits can occur within syntactic constituents, apparently at any point in the string; it is unusual for the separate parts to be complete units in their own right; explicit repair of the antecedent does not occur very often. The theoretical consequences of these results for claims in the literature are pointed out. The practical implications for dialogue systems are mentioned too.