Abstract
Disfluencies (“um,” repeats, self-repairs) are prevalent in spontaneous speech, and are relevant to both human speech communication and speech processing by machine. Although disfluencies have commonly been viewed as `noisy’ events, results from a large descriptive study indicate that disfluencies show regularities in a number of dimensions. This paper reports selected results on Switchboard and two comparison corpora of spontaneous speech. Results illustrate the systematic distribution of disfluencies, and highlight differences as well as universals across corpora and speakers.
