A Humanoid Robot Dialogue System Architecture Targeting Patient Interview Tasks
Published in ROMAN '24, 2024
The project focuses on enhancing humanoid robots' performance in conducting patient interviews by developing a dialogue system architecture. Humanoid robots offer a unique advantage in utilizing verbal and behavioral cues during interviews. However, the human-like appearance can create expectations of human-level performance, leading to interaction quality degradation if not met. The project introduces a dialogue system architecture that includes a nested inner real-time control loop to enhance the robot's responsiveness based on the concept of "stance." This architecture also expands the dialogue state to monitor task progress and human engagement. Experiments with the proposed architecture demonstrate improved performance in terms of response timeliness and user impressions during patient interviews.
Recommended citation: Yifan SHEN*, Dingdong Liu, Yejin Bang, Ho Shu Chan, Rita Frieske, Hoo Choun CHUNG, Jay Patrick Monton Nieles, Tianjia ZHANG, Trung Kien Pham, Wai Yi Rosita CHENG, YINI FANG, Qifeng Chen, Pascale FUNG, Xiaojuan Ma, Bertram Emil Shi
Recommended citation: Yifan SHEN*, Dingdong Liu, Yejin Bang, Ho Shu Chan, Rita Frieske, Hoo Choun CHUNG, Jay Patrick Monton Nieles, Tianjia ZHANG, Trung Kien Pham, Wai Yi Rosita CHENG, YINI FANG, Qifeng Chen, Pascale FUNG, Xiaojuan Ma, Bertram Emil Shi
Download Paper