Data-Driven Critical Tract Variable Determination for European Portuguese

Abstract

Technologies, such as real-time magnetic resonance (RT-MRI), can provide valuable information to evolve our understanding of the static and dynamic aspects of speech by contributing to the determination of which articulators are essential (critical) in producing specific sounds and how (gestures). While a visual analysis and comparison of imaging data or vocal tract profiles can already provide relevant findings, the sheer amount of available data demands and can strongly profit from unsupervised data-driven approaches. Recent work, in this regard, has asserted the possibility of determining critical articulators from RT-MRI data by considering a representation of vocal tract configurations based on landmarks placed on the tongue, lips, and velum, yielding meaningful results for European Portuguese (EP). Advancing this previous work to obtain a characterization of EP sounds grounded on Articulatory Phonology, important to explore critical gestures and advance, for example, articulatory speech synthesis, entails the consideration of a novel set of tract variables. To this end, this article explores critical variable determination considering a vocal tract representation aligned with Articulatory Phonology and the Task Dynamics framework. The overall results, obtained considering data for three EP speakers, show the applicability of this approach and are consistent with existing descriptions of EP sounds.

Data-Driven Analysis of European Portuguese Nasal Vowel Dynamics in Bilabial Contexts

“…Further work to improve these aspects is required. From previous experience (e.g., [4]), a first approach with potential would be an exploration of the methods adopted in this paper with data extracted not directly from images, but from tract segmentations-techniques that would allow one to consider direct measures of the velopharyngeal passage and interlip distance, as we considered in [4], from RT-MRI.…”

“…Considering articulatory data extracted from RT-MRI, the authors have explored datadriven approaches to determine critical articulators from vocal tract data extracted from RT-MRI [4,37], expanding an approach proposed for EMA [38]. One notable aspect of the methods conveyed in Silva et al [4] is that, even though the method considers statistical modeling for the different sounds, the consideration of tract variables aligned with the articulatory phonology framework [39], as grounds for the method, yields results that maintain a connection to the tract anatomy (e.g., constrictions on the tongue tip and body) and are, hence, more interpretable towards a critical discussion of the outcomes and an improved knowledge of speech production. Nevertheless, while this provides valuable information for a wide range of sounds in an unsupervised manner, it is still not able to provide it over time, since it relies on a single vocal tract configuration to represent each sound, often taken at the middle of its production.…”

“…The study of such complex sounds, exhibiting characteristic dynamic patterns, contributes to improving our knowledge on speech production, supporting a range of applications from speech therapy to articulatory speech synthesis. To tackle these aspects, recent advances in real-time MRI (e.g., [2]) include a considerable step-up in the possible time resolution (e.g., 50 frames per second), and our team has been working on acquiring a novel RT-MRI dataset that encompasses a large number of speakers to study European Portuguese nasal vowels, and addressing how to tackle these data to attain a systematic processing and quantitative analysis of the articulatory data (e.g., [3,4]). In this context, the large amount of available data also fosters the consideration of novel approaches based on data-driven methods (e.g., [4]).…”

“…To tackle these aspects, recent advances in real-time MRI (e.g., [2]) include a considerable step-up in the possible time resolution (e.g., 50 frames per second), and our team has been working on acquiring a novel RT-MRI dataset that encompasses a large number of speakers to study European Portuguese nasal vowels, and addressing how to tackle these data to attain a systematic processing and quantitative analysis of the articulatory data (e.g., [3,4]). In this context, the large amount of available data also fosters the consideration of novel approaches based on data-driven methods (e.g., [4]). These can have an impact on how we improve our knowledge on nasal vowels, by making it possible to computationally model their dynamic behavior and providing a means to deal with the variability of the data (e.g., due to the idiosyncratic nature of each speaker's anatomy and articulation).…”