INTERACTING WITH KNOWLEDGE BASED SYSTEMS THROUGH DIALOGUE GAMES
Paul H. Leng · 2014
Approaches to text generation have emphasised the need to plan dialogues between user and computer, but have paid insufficient attention to the highly conventional nature of much human-human interaction. It is possible to conceive of certain classes of dialogues as highly rule-governed activities, and to do so facilitates the planning of such dialogues, by decreasing the reliance on a model of the user. One such class of dialogue is interactive argument, in which a topic is discussed with the aim of reaching a conclusion as to the truth of some proposition. We contend that interactive argument provides a useful model of interaction with a KBS, could provide a novel means of knowledge elicitation, and would serve as a fruitful model for educational systems also. Details of a dialogue manager based on a well known dialogue game (Mackenzie’s DC) are given. Since KBS will exist independently of this way of interacting with them we give details of a Dialogue Abstract Machine (DAM), which could serve as a target interface between our dialogue manager and a KBS of any formalism.