Prosodic Focus Modeling in Persian: An Articulatory- Functional Approach

This paper is an attempt to model Persian prosodic focus using an articulatory-functional approach, i.e., parallel encoding and target approximation (PENTA). The modeling was done on 150 utterances produced in different focus conditions using PENTAtrainer2 in Praat software. PENTAtrainer2 is a trainable prosody synthesizer that optimizes categorical pitch targets each corresponding to multiple communicative functions. The appraisal was carried out through numerical and subjective evaluations by comparing the F0 trajectories generated by the extracted pitch targets to those of natural utterances. The numerical results showed that the synthesized F0 contours were close to the natural ones in terms of RMSE (= 1.94) and correlation coefficient (= 0.84). Subjective evaluation also showed that the rate of focus identification and naturalness judgment were highly similar between synthetic and natural F0 trajectories.