Abstract
We present Flipper, a natural language interface for describing high leveltask specifications for robots that are compiled into robot actions. Flipperstarts with a formal core language for task planning that allows expressingrich temporal specifications and uses a semantic parser to provide a naturallanguage interface. Flipper provides immediate visual feedback by executing anautomatically constructed plan of the task in a graphical user interface. Thisallows the user to resolve potentially ambiguous interpretations. Flipperextends itself via naturalization: users of Flipper can define new commands,which are generalized and added as new rules to the core language, graduallygrowing a more and more natural task specification language. Unlike othertask-specification systems, Flipper enables natural language interactions whilemaintaining the expressive power and formal precision of a programminglanguage. We show through an initial user study that natural languageinteractions and generalization can considerably ease the description of tasks.Moreover, over time, users employ more and more concepts outside of the initialcore language. Such extensions are available to the Flipper community, andusers can use concepts that others have defined.