In this talk, I report on a research framework developed for handling miscommunication between people and robots in task-oriented spoken dialogue. This involved crowdsourcing a dataset of human-robot language and developing a conversational interface to robots called TeamTalk. I'll describe my recent experiments and a resulting novel learning method that improves how a robot uses dialogue to recover from misunderstandings about its surroundings. When the robot encounters a new problem, it looks back on its interaction history to consider how it resolved similar situations. This type of recovery allows spoken dialogue systems for robots to be more resilient to miscommunication and operate in a wide range of scenarios with people.
In the larger context, this research addresses the emerging field of intelligent agents (like robots) enabling people to accomplish collaborative tasks. Issues I will address include referential ambiguities or impossible instructions along with possible responses. The contribution of this research is to incorporate additional information from situated context (a robot's path planner and surroundings) to detect and recover from misunderstandings.
Matthew Marge is a Computer Scientist in the Multilingual Computing and Analytics Branch of the U.S. Army Research Laboratory. He received the Ph.D. and M.S. degrees in Language and Information Technologies from the School of Computer Science at Carnegie Mellon University, and the M.S. degree in Artificial Intelligence from the University of Edinburgh. Dr. Marge's research focuses on improving how robots and other artificial agents can establish common ground with people through dialogue. His current interests lie at the intersection of computational linguistics and human-robot interaction, specializing in dialogue systems. Dr. Marge is a recipient of the National Science Foundation Graduate Research Fellowship, the St. Andrew's Society of the State of New York Scholarship, and the Barry Goldwater Scholarship.