Rethinking Logics of Action and Time

It is over thirty years since I developed interval temporal logic and the accompanying logic of action and time. Overall, these theories have held up well and, with some extensions over the years, have remained useful in our work on AI planning/reasoning systems and natural language understanding. Recently I have become interested in systems that can learn by reading, and specifically, that can learn necessary conditions for event occurrence from reading dictionary definitions. This task adds new constraints on the form of the temporal logic we need. In this talk, I will review our earlier work on temporal logic and then look at the problems that have forced a recent generalization of the formalism in order to allow compositional construction of event definitions from natural language definitions.