Edward Flemming
MIT

Edward Flemming

MIT
United State

Generating and parsing f0 contours using a functional model of f0 production

I will present ongoing work aimed at developing a framework for formulating phonetic grammars of tone realization that can derive complete f0 contours from phonological specifications.

The proposed framework consists of two main components: A set of weighted constraints that select and locate optimal tone targets for a given phonological representation, and a functional model of f0 production that maps time-aligned tone targets to f0 contours (based on Birkholz & Xu 2018). These two components work in parallel: the constraints that determine tone targets evaluate properties of the resulting f0 contour.

In addition to generating f0 contours, the model of f0 production can also be fitted to observed f0 contours to infer the underlying f0 targets, effectively parsing the f0 contour (cf. Öhman & Lindqvist 1965, Birkholz & Xu 2018). This proves particularly valuable when analyzing tones with targets that are difficult to locate because they do not correspond to a local maximum or minimum, such as tone targets that result in a transition from a falling f0 movement to a relatively level plateau, or vice versa. These kinds of f0 events are often referred to as ‘elbows’,

The framework will be illustrated through analyses of two tonal phenomena that involve elbow targets: (i) the timing of the L- phrase accent in English H*L- intonational melodies, and (ii) realization of the Mandarin rising tone.

References

Birkholz, Peter, & Xu, Yi (2018) Estimation of pitch targets from speech signals by joint regularized optimization. 26th European Signal Processing Conference, 2089-2093. Öhman, Sven, & Lindqvist, J. (1965) Analysis-by-synthesis of prosodic pitch contours. STL-QPSR 6(4), 1-6.

Biography

Edward Flemming is Associate Professor of Linguistics at the Massachusetts Institute of Technology. His research focuses on two topics: (i) the ways in which the sound systems of languages are shaped by the limitations of human speech production and speech perception, and (ii) the nature of the phonetic component of grammar.