Abstract
Various models have been proposed to account for judgments of the relative prominence of pitch accents in relation to FO variation. Two topics are addressed in this paper. The first topic is how pitch accents need to be realized in order to obtain the appropriate prominence patterns. In order to answer this question, relevant data and models for prominence perception are summarized. It is tentatively concluded that the prominence associated with FO peaks is judged relative to the local FO range, as signalled by the pitch at utterance onset. No model for prominence perception proposed so
far can account for the available data, and more insight is needed into the issue of pitch range estimation before real progress can be made. The second topic concerns the assumption of free gradient variability underlying models of prominence perception: it is assumed that the prominence associated with pitch accents may vary freely and in a gradient way from accent to accent within the phrase. Prominence ratings collected for fragments of spontaneous speech provide no evidence of a constraint prohibiting such variation. Some implications of these findings are considered.
| Original language | English |
|---|---|
| Title of host publication | Computing prosody: computional models for processing spontaneous speech |
| Editors | Y. Sagisaka, N. Campbell |
| Pages | 95-116 |
| Publication status | Published - 1997 |