Microsoft Word - Prom-on_Xu_TAL2012_final.docx

This paper presents a quantitative representation of Thai tones in the form of pitch targets. The optimal pitch targets of Thai tones were estimated by using a combination of the quantitative Target Approximation (qTA) model and a stochastic learning algorithm. With the advent of this technique, the representation of communicative functions can be systematically and quantitatively studied. The statistical analysis of estimated parameters shows clear contrastive differences between tone categories, indicating unique representation of Thai tones. The results of the modeling experiment demonstrate the adequacy of unitary (as opposed to compositional) underlying targets for generating continuous tonal contours in Thai.