A adaptive interface KNOWBOT® was designed to solve some of the problems that face the users of large centralized data bases. The interface applies the neural network approach to information retrieval from a data base. The data base is a subset of the Nuclear Plant Reliability Data System. The interface KNOWBOT preempts an existing data base interface and works in conjunction with it. By design, KNOWBOT starts as a tabula rasa but acquires knowledge through its interactions with the user and the data base. The interface uses its gained knowledge to personalize the data base retrieval process and to induce new queries. The interface also forgets the information that is no longer needed by the user. These self-organizing features of the interface reduce the scope of the data base to the subsets that are highly relevant to the user needs. A proof-of-principal version of this interface has been implemented in Common LISP on a Texas Instruments Explorer I workstation. Experiments with KNOWBOT have been successful in demonstrating the robustness of the model especially with induction and self-organization. This paper describes the design of KNOWBOT and presents some of the experimental results.