Predicting end-of-turn in multiparty conversations is crucial to increase the usability and natural flow of spoken dialogue systems, offering substantial enhancements to conversational agents. We present a novel window-based method to predict end-of-turn moments in real-time in multiparty conversations, by leveraging the capabilities of cutting-edge pre-trained language models (PLMs) and recurrent neural networks (RNN). Our method fuses the distilBERT language model with a Gated Recurrent Unit (GRU) to accurately predict end-of-turn points in an online fashion. Our approach can significantly outperform conventional Inter-Pausal Unit (IPU)-based prediction methods that often overlook the nuances of overlap and interruption during dynamic conversations. Potential applications of this study are significant, particularly in the domains of virtual agents and human-robot interactions. Our accurate online end-of-turn prediction model can be facilitated to enhance the user experience in these applications, making them more natural and seamlessly integrated into real-world conversations.

In this paper we present a computational study to quantitatively examine the task of predicting the next speaker in multi-party conversations using machine learning models. To accomplish this, we create features that accurately represent information relevant to speaker changes in such conversations. We utilize sentence-based models, rather than the widely-used InterPausal Unit (IPU)-based models, and extend the definition of verbal backchanneling to include additional reactions that signify listeners’ attention or interest. Through extensive experiments with various machine learning models and inputs, we show that our sentence-based models outperform existing IPU-based models, with the best model achieving 61.39% accuracy. Our study provides design implications and recommendations for the development of virtual agents or humanoid robots with interactive social interaction capabilities.