Design of an intelligent assistant system from several applications

Often a user that solves a problem with the support of a system (knowledge-based system, software, etc.), must intervene to overcome the limited capabilities of the system. We aim to introduce a type of intelligent assistant system (IAS) between the user and the system. The IAS will acquire the knowledge that permits the user to solve problems and will after use it for solving similar problems to relieve the user's interventions. Thus, the IAS builds in an incremental manner a "personal" knowledge base that is a kind of 'user-in-action' model. IASs must accomplish a number of tasks, in one hand for managing knowledge (e.g., acquisition, assimilation, validation), and, in the other hand, for managing interaction with the user (e.g., dialogue management, explanation, cooperation). It is not realistic to tackle the design of IASs in one shot. The keystone of the approach lies on the observation that tasks have not the same importance in each application and on the fact that tasks are used incrementally. We present an approach to design IASs from several applications. This permits one to: generalize the results obtained in each application; and share results among applications.The author has been following such an approach on the basis of four applications in different domains (power systems, banking, genetics, alarm control monitoring).<
>