The ART of Conversation: Measuring Phonetic Convergence and Deliberate Imitation in L2-Speech with a Siamese RNN

• DOI: 10.21437/interspeech.2023-2283
• Publication Date: 2023-08-14T08:22:20Z
• Authors: Zheng Yuan; Aldo Pastore; Dorina de Jong; Hao Xu; Luciano Fadiga; Alessandro D'Ausilio
• Keywords: Slovak

Phonetic convergence describes the automatic and unconscious speech adaptation of two interlocutors in a conversation.This paper proposes Siamese recurrent neural network (RNN) architecture to measure holistic spectral characteristics sounds an L2-L2 interaction.We extend alternating reading task (the ART) dataset by adding 20 native Slovak L2 English speakers.We train test RNN model phonetic from three different language groups: Italian (9 dyads), French (10 dyads) dyads).Our results indicate that effectively captures dynamics speaker's imitation ability.Moreover, this text-independent is scalable capable handling L1-induced speaker variability.