Friday, October 16, 2015 - 12:00pm to 1:00pm
Location:Traffic 21 Classroom 6501 Gates & Hillman Centers
Speaker:VITTORIO PERERA, Ph.D. Student /VITTORIO%20PERERA
As robots move out of labs and into the real world, it is critical that humans will be able to specify task requirements in an intuitive and flexible way. In this talk, we will see how to enable untrained people to instruct robots via dialog. In the first part of the talk I am going to present a joint probabilistic model over speech, the resulting semantic parse and the mapping from each element of the parse to a physical entity in the building. This model allows the robot to understand spoken commands while being flexible to the ways that untrained people interact with it, being robust to speech to text errors, and learning referring expressions for physical locations in a map. The second part of the talk focuses on complex commands, a natural language sentence consisting of sensing-based conditionals, conjunctions, and disjunctions. In this second part I am going to present a template based algorithm able to recover the correct structure of a complex command, two dialogue models that enable the user to confirm or correct the extracted command structure and how the structure extracted can be used for planning and execution by the service robot. Joint work with Manuela Veloso. In partial fulfillment of the speaking requirement.