报告时间：2021年5月21日（周五）上午9点

报告地点：王选所大楼106报告厅

报 告 人：刘昌松

报告题目：Towards Mediating Shared Perceptual Basis in Situated Dialogue

摘要：In situated dialogue, although an artificial agent (e.g., robot) and its human partner are co-present in a shared environment, they have significantly mismatched perceptual capabilities (e.g., recognizing objects in the surroundings). When a shared perceptual basis is missing, it becomes difficult for the agent and the human to establish common ground and communicate effectively. The work presented in this talk focuses on computational approaches to enable robust and adaptive human-agent communication in situated settings, including graph-based approaches for referential communication, learning methods for mediating perceptual differences and mapping language to videos of real life tasks, and software architecture for building situated dialogue agents. All these approaches contribute to the ultimate goal of building collaborative dialogue agents for situated interaction, so that the next generation of intelligent machines/devices can better serve human users in daily work and life.

简介：刘昌松，本科和硕士毕业于电子科技大学，2009年赴美国密歇根州立大学语言交互研究实验室攻读博士研究生，2014年获计算机科学与工程专业博士学位。2015至2017年先后在密歇根州立大学语言交互研究实验室和加州大学洛杉矶分校计算机视觉、认知、学习与自主机器人中心从事博士后研究，后加入初创企业暗物智能科技担任首席科学家。2020年回国参与北京通用人工智能研究院筹建及科研规划工作。在美期间长期从事多模态人机交互、场景中的语言理解与对话、智能教学系统等方向的研究与应用开发工作。曾作为主要科研人员参与美国自然科学基金多模态语言处理与场景对话架构、美国国防高级研究计划局跨模态知识学习与自主推理、暗物智能科技基于认知智能的多模态交互教学系统等项目的科研攻关。在自然语言处理和人机对话等领域发表论文15篇，申请美国技术专利5项。