Abstract:

In a programming by demonstration system, examples provided by the user are generalized to form abstractions that are used by the system. There are several reasons why the system and the user should collaborate to annotate examples with knowledge that guides generalization. These include dealing with inconsistencies among examples or the underlying domain theory, making queries in order to speed generalization (e.g., active learning), and making suggestions (perhaps based on data mining) to change the outcome of generalization. This paper presents the design of a system that interacts with users via a set of "guessers": algorithms for suggesting possible annotations to the user, in the context of learning hierarchical task models.