MARCS Research Seminar

Professor Mari Ostendorf, Department of Electrical Engineering, University of Washington will be presenting "Finding Information in Disfluencies"

Abstract: One of the characteristics of spontaneous speech that distinguishes it from written text is the presence of disfluencies, including filled pauses (um, uh), repetitions, and self corrections. In spoken language processing applications, disfluencies are typically thought of as "noise" in the speech signal. However, there are systematic patterns associated with where disfluencies occur that can be leveraged to automatically detect them and to improve natural language processing algorithms. Further, rates of different types of disfluencies (which are thought to be related to different levels of speech production) appear to vary depending on the individual speaker and the social/cognitive context. Through improved models of speaker variability, we aim to extract additional information from spoken language -- beyond the literal meaning of the words -- by detecting disfluencies and analyzing their structure. In this talk, we describe both computational models for disfluency detection and analyses of different corpora that provide some insights into inter- and intra-speaker variability in conversational vs. argumentative contexts.