Applying General Turn-taking Models to Conversational Human-Robot Interaction
Автор: Gabriel Skantze
Загружено: 2025-02-27
Просмотров: 272
Описание:
Demo of our work on applying our general turn-taking models (Voice Activity Projection and TurnGPT) to human-robot conversations. The models are trained on human-human conversational data and applied directly, without any fine-tuning. As can be seen in the video, this allows the robot to answer quickly when the user is yielding the turn, while avoiding to interrupt the user when they make longer pauses. It also allows the robot to predict whether the user is trying to barge in or not.
The paper can be found here:
https://arxiv.org/abs/2501.08946
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке: