Conversational program model
Daniel D. Corkhill · ACM SIGART Bulletin · 1977
A practical approach to natural language recognition is taken in the development of a transition graph model for predetermined conversational program dialogues. The model incorporates context and prediction to reduce the number of user entries likely at any given point in a dialogue. In the design of the model, four major questions are addressed: (1) what types of conversational applications are suitable for practical natural language interaction, (2) what are the important characteristics of the dialogues used in those applications, (3) how can those dialogues be represented to improve program performance, and (4) how can those dialogues be represented to aid in the program design process. Implementation of an interactive computer science course advisor program provides evidence that the model can be a significant aid in the design of practical conversational programs accepting restricted natural language input via a teletype-like terminal.