The study explores the concept of entrainment – the tendency of people to engage in synchronized behavior during interactions. Using Deep Neural Networks (DNN) embeddings like BERT and TRIpLet Loss network (TRILL) vectors, the researchers measured semantic and auditory similarities in dialogues. The findings revealed a higher tendency for people to entrain on semantic features compared to auditory ones. Additionally, a positive correlation was found between semantic and auditory linguistic entrainment. These insights could assist in improving human-machine interaction (HMI).
Publication date: 27 Dec 2023
Project Page: https://arxiv.org/abs/2312.16599
Paper: https://arxiv.org/pdf/2312.16599
