Exploiting Plan Reasoning to Generate Effective Task Descriptions

Abstract

The task of generating effective textual descriptions of the plans produced by automatic planning algorithms is made more difficult by the complexity of the plans themselves. Describing every component of a complicated plan makes for an awkward and ineffective description. Previous work on the generation of instructional text from plans has not addressed the issue of the generation of concise and effective plan descriptions. The work I propose focuses on the development of domain-independent techniques to eliminate the complexity from these textual descriptions. I propose several techniques that exploit the plan reasoning capabilities of users and the task context in which plan descriptions occur to produce effective and concise descriptions.