laming CtS

A number of researchers have investigated the use of planbased approaches to generate textual explanations (e.g., Appelt 1985; Hovy 1988; Moore 1989; Maybury 1990b). This paper extends this approach to generate multimedia explanations by defining three types of communicative acts: linguistic acts (illocutionary and locutionary speech acts), visual acts (e.g., deictic acts), and mediaindependent rhetorical acts (e.g., identify, describe). This paper formalizes several of these communicative acts as operators in the library of a hierarchical planner. A computational implementation is described which uses these plan operators to compose route plans in coordinated natural language and graphics in the context of a cartographic information system.