Yoon, Tae-Jin. 2021. GAMM-based modeling of dynamic phonetic trajectories. Studies in Phonetics, Phonology and Morphology 27.3. 463-481. Dynamic trajectories of F0s and formants influence the perception of speech sounds. For this paper, the dynamic trajectories of two calibration sentences read by all participants (more than 600) in the TIMIT database are preprocessed by the general-purpose Python programming language and the third-party packages Parselmouth, TextGridTools, and Pandas. Parselmouth and TextGridTools function as interfaces to the underlying Praat software, which is a de facto standard tool for phonetic research but lacks the functionality of general-purpose computer programs. The data preprocessed through the interfacing packages are formed into a data frame and then fed into the R environment, which still has superior capability to deal with statistical modeling in comparison to Python, to model the dynamic phonetic trajectories using the Generalized Additive Mixed Model (GAMM). The F0 trajectories over all calibration sentences were also inspected across dialects and different education levels. The approach taken in this paper will further our understanding of the phonetic trajectories in the shaping of phonetic and phonological patterns. 

Keywords: GAMM, Parselmouth, TextGridTools, Pandas, Python interfaces to Praat, Dynamic Phonetic Trajectories, Formants, F0, The TIMIT database

[pdf]