2024 INTERSPEECH INTERSPEECH 2024

Impact of the tonal factor on diphthong realizations in Standard Mandarin with Generalized Additive Mixed Models

Abstract

This study examines the effects of lexical tones on diphthong realizations in Standard Mandarin. We investigated the two falling diphthongs /ai/ and /au/ from a Standard Mandarin reading text corpus. A set of GAMMs models was employed to test whether and how tones and f0 influence the diphthong realizations. The results show the vowel height (reflected by F1) to differ with respect to tones: with a high tone, the diphthongs tend to be realized as more closed, and as more open with a low tone; with a rising tone, they tend to have a typical diphthongized realization, with a dynamical pattern of F1 contour; and to be monophthongized with a falling tone. The interaction between f0 and F1 extensively confirmed in monophthongs across languages, is equally applicable to /ai/ and /au/: f0 is negatively correlated with F1. The results show the universality of the tonal effect on vowel realization in different diphthongs and imply physiological factors behind it.

🌉 Interdisciplinary Bridge — Machine Learning and Mathematics & Optimization
🧭 Keyword Pioneer — diphthong realization
🐝 Cross-Pollinator — Artificial Intelligence, Computer Science, Computer Vision, Data Science & Analytics, Deep Learning, Healthcare & Medicine, Interdisciplinary, Knowledge & Reasoning, Machine Learning, Mathematics & Optimization, Natural Language Processing, Reinforcement Learning, Speech & Audio