A biologically inspired neural network for modeling phrase-final lengthening

Abstract

This work proposes a central-pattern-generator-inspired neural network model for the interaction between phrase-final lengthening and stress. Recent work in the area of speech prosody has been concerned with the mechanisms involved in phrase-final lengthening, and specifically how phrase-final lengthening interacts with stress or prosodic prominence. The current study investigates the interaction of stress and lengthening at the end of English phrases. Adult American English speakers were recorded reading aloud sentences in which phrase boundaries had been manipulated so that the target words were either phrase-final or phrase-medial, and the durations of syllables in the target words were compared between the two conditions. Results so far support previous findings that phrase-final lengthening in English affects stressed syllables near phrase boundaries and phrase-final syllables while leaving unstressed syllables between the two unaffected [Turk and Shattuck-Hufnagel, 2007]. Domain-based models of prosodic lengthening have so far been unable to provide a unified account of this phenomenon. A biologically plausible artificial neural network is shown which provides a model of the mechanism behind this interaction using three oscillators with differing periods which input to three interconnected thresholded integrate-and-fire artificial neurons, the output of which determines the timing of the syllables, stress feet, and phrase.