Plan recognition is an active research area in automatic reasoning, as well as a promising approach to engineering better interfaces in the area of natural language understanding as well as in a wide variety of other application areas. However, exactly what kind of interface tasks plan recognition can be used to support across domains and communication modalities, and how such tasks in turn constrain plan recognition algorithms, are questions that have typically remained unexamined.
In my work in plan-based natural language understanding, I developed a knowledge-based approach to the understanding of subdialogues such as clarifications and corrections. In particular, I developed a plan recognition system that could infer discourse plans (which represented how utterances could relate to a topic) as well as the traditional domain plans (which represented the topic). Incorporating knowledge of discourse into the plan-based framework enabled the understanding of a wide variety subdialogues, while maintaining the computational advantages of the plan-based approach. In other work I generalized the use of plan recognition, by developing a graphical plan-based interface for design. I identified how plan recognition made interfaces more intelligent and interactive (by supporting advice generation, task completion, context-sensitive responses, error handling), and showed how interface tasks constrained plan recognition.