HumanRobot Dialogue Learning (HuRDL) Corpus

HuRDL Corpus是由美国陆军研究实验室和塔夫茨大学合作创建的一个对话数据集，旨在研究智能代理如何在多模态不确定环境中通过提问学习。该数据集包含22个对话，总计13小时，主要用于分析和改进智能代理的问题生成能力。数据集中的对话是在线互动虚拟环境中收集的，参与者扮演机器人执行协作工具组织任务。数据集的创建过程涉及设计特定的任务环境和对话交互，旨在通过人类参与者的提问行为来指导智能代理的学习策略。该数据集适用于开发和测试能够处理高不确定性环境中的智能对话系统。

The HuRDL Corpus is a dialogue dataset co-created by the U.S. Army Research Laboratory and Tufts University, aiming to investigate how intelligent agents learn through questioning in multimodal uncertain environments. This dataset includes 22 dialogues with a total duration of 13 hours, and is primarily used for analyzing and improving the question-generation capabilities of intelligent agents. The dialogues in the dataset were collected in an online interactive virtual environment, where participants assumed the role of robots to perform collaborative tool organization tasks. The dataset creation process involved designing specific task environments and dialogue interactions, with the goal of guiding the learning strategies of intelligent agents via the questioning behaviors of human participants. This dataset is suitable for developing and testing intelligent dialogue systems capable of handling highly uncertain environments.