The Structure of Tonal Representations - Evidence from Tune-Text Synchronisation
Final Report Abstract
In this project we have integrated information from both the time domain (tonal alignment) and frequency domain (tonal scaling) in the further development of our theoretical model of intonation structure. We propose that a number of pitch accents should be regarded as derived from an underlying representation which is not fully specified as to the association properties of certain tones. Operations such as downstep (and possibly upstep) are applied to this underlying representation, and are thus unaffected by the association properties of the H tones they apply to. The surface association of the tones is affected by factors such as speed of delivery and the number of metrically strong syllables available as potential docking sites for pitch accent tones. We investigated the alignment of fO peaks in rising LH p'ltch accents In two varieties of German (Vienna and Düsseldorf), both in relation to acoustically defined segments, referred to as segmental anchors, as well as to dynamically defined speech gestures, referred to as articulatory anchors. Although there was a tendency for Southern varieties to have later peaks than Northern ones, as also found in previous studies, alignment latencies of individual speakers in the two dialectal groups overlapped. These results support a gradient view of dialectal variation in tonal alignment. When comparing prenuclear and nuclear accents, by contrast, we found differences which could be interpreted as discrete. We observed an anchor shift, in that nuclear accent peaks were anchored to the intervocalic consonant, while prenuclear accent peaks were anchored to the following unstressed vowel. This was clearly the case both in the acoustic and articulatory records, reflecting a difference at the symbolic level, possibly in terms of an additional tone following the LH complex. Furthermore, we have developed a smoothing algorithm for removal of microprosodic influence, based on fO weighting functions depending on vowel height and voicing. This algorithm considerably reduces the influence of microprosody while maintaining macroprosodic aspects of the fO shape. In the context of speech synthesis our smoothing procedure can contribute to an improvement in fO generation, since it provides the basis for separate modelling of micro- and macroprosodic influences on fO. In terms of macroprosodic influences on fO, we have further developed and updated the annotation scheme for German intonation, GToBI, and have adapted the training materials accordingly including new interactive exercises.
