Addressee Estimation

Addressee estimation, the task of identifying who is being addressed in a conversation, is crucial for developing more natural and human-like interactions, particularly in human-robot interaction and multi-party conversations. Current research focuses on improving the accuracy of addressee recognition using deep learning models, often incorporating multimodal data (e.g., visual cues like gaze and body posture) and graph neural networks to capture complex conversational structures. These advancements aim to enhance the capabilities of social robots and conversational AI systems by enabling them to better understand and participate in dynamic, multi-person interactions.