Computational modelling of double focus in American English
F Liu, Santitham Prom–on, Yi Xu, DH Whalen · UCL Discovery (University College London) · 2015
This study investigated how double focus in English statements and questions can be computationally modelled. PENTAtrainer2 was used to learn syllable-sized multi-functional targets from a corpus of 1960 English utterances, with controlled variations in lexical stress, focus, modality, and sentence length. The results showed that the learned targets could generate F0 contours close to the original. In particular, the asymmetry in the interaction between focus and modality was effectively simulated.