Downloads & Free Reading Options - Results

This project aimed to make significant advances towards enabling more adaptive systems involving human and machine components by characterizing system behavior as the result of a reasoning process. Rather than specifying every operation in advance, this approach only requires one to provide the system with its overall goals in addition to some knowledge of the environment, its dynamics and the effects of its actions. In unanticipated or troublesome situations, the system would adapt its behavior by reasoning about the appropriate actions to take to achieve its goals. During this work we helped enable a system that took vague and incomplete commands from a human user and performed the correct action. This was enabled by using a reasoning engine to resolve the ambiguities and infer missing information from the command. The technical challenges was that current inference engines did not scale to problems of this size. We made several advances that enabled them to be used on such problems and demonstrated them on working systems that displayed a notable increase in the abilities of computers to understand natural language commands.