Perception and classification of disfluent phenomena as fluent communicative devices in specific prosodic contexts

From HLT@INESC-ID

Revision as of 15:09, 2 April 2009 by Rdmr (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Helena Moniz
Helena Moniz

Date

  • 15:00, Friday, April 3rd, 2009
  • Room 336

Speaker

Abstract

This work explores prosodic cues of disfluent phenomena. We have conducted a perceptual experiment controlling tonal scaling, prosodic phrasing and pause length of the stimuli. Results suggest that some disfluencies may be considered felicitous by listeners. This is the case of filled pauses and prolongations, uttered at break indices 3 and 4, when they are regularly scaled relatively to their adjacent constituents and produced with silent pauses.

In an attempt to discriminate which linguistic features are more salient in the classification of disfluencies as either fluent or disfluent phenomena, we have used CART techniques on a corpus of 2 hours of spontaneous and prepared non-scripted speech. This sample was annotated with prosodic (break indices, contour shape and F0 restart), part of speech and syntactic information.

CART results pointed out 2 splits: break indices and contour shape. The first split indicates that events uttered at breaks 3 and 4 are considered felicitous. The second shows that these events must have flat or ascending contours to be considered as such; otherwise they are strongly penalized.

In spite of the early stage of our work, our preliminary results suggest that there are regular trends in the production of these events, namely, prosodic phrasing and contour shape. This work is a first step towards the discrimination of fluent and disfluent phenomena in spontaneous and prepared non-scripted speech, based on larger corpora for European Portuguese.