Researchers involved: Aleksandra Ćwiek (FLESH, ZAS Berlin), Pilar Prieto (MultIS, UPF Barcelona), Frank Kügler (MultIS, Goethe University Frankfurt) and Patrick Rohrer (external collaborator of MultIS)
Our aim is to combine the expertise of the respective parties with regard to speech prosody, with a focus on extending this to visual prosody, e.g., annotations of crucial gesture phases. We will test the correlation between top-down prosodic annotation systems (e.g., ToBI) and bottom-up automatic annotation. The time- and training-intensive top-down annotation is very exact but costly, whereas the bottom-up automatic annotation is very time-efficient but error-prone. There is currently no middle ground between these two methods, and this collaboration is designed to address this issue. We will begin by examining correlations between prosodic annotations and acoustic measurements, such as CoG, F0, and amplitude envelope peaks. The next step will be to extend the analyses to visual prosody using the annotated data from MultIS and automatic analysis scripts from FLESH.