A Model-Driven Question-Answering System for Mixed-Initiative Computer-Assisted Construction

This paper describes a pragmatic question-answering system which was developed for experimental use in a computer-assisted instruction environment. The two primary design goals of this system are 1) the representation of dynamic processes along with their causal interrelationships, and 2) the generation of responses in units of discourse greater than simple sentences. A method of modeling processes with augmented finite-state automata is discussed and is shown to permit efficient inferencing as well as provide a satisfactory deep structure for paragraph generation. Coupled with the dynamic process model is a semantic/conceptual network which contains static or factual information. Examples of the inferencing techniques using both the automaton model and the semantic network are given.