This chapter presents a prototype expert interface (IR-NLI, Information Retrieval Natural Language Interface) to an information retrieval system, developed at the University of Udine in the frame of a broader research effort concerning the topics of cooperative man-machine interaction and expert systems. After a discussion of the novel notion of expert interface, attention is focused on the IR-NLI system for the access to online information services by non-technical users. General specifications, design criteria, and architecture of IR-NLI are presented first. Knowledge representation methods and reasoning mechanisms adopted are then illustrated in detail. In this context, a new mechanism, called task, for representing and using meta-knowledge in rule-based systems is proposed. The internal operation of the system, together with two examples of interaction with the user, are illustrated. The paper concludes with the discussion of some preliminary ideas on how learning capabilities could be introduced in IR-NLI through the task mechanism.
