Multiversion information retrieval systems and feedback with mechanism of selection

This article initiates a study of multiversion information retrieval (IR) systems. The term “multiversion” refers to the existence of several different methods for realizing one (as a minimum) system's process. The article also provides a theoretical justification for the necessity of creating such systems to perform an optimal search for the satisfaction of the user's information need. The basic principles for designing multiversion IR systems are described. These principles are used in developing a method for selecting an optimal (the best of possible) state of the system for each individual user. We also indicate how the proposed method is used in conjunction with the feedback mechanism in the IR system. The results of an experiment illustrate the capabilities of multiversion systems and their advantages over traditional IR systems. It is shown that multiversion IR systems are superior in their abilities to adapt to an individual user in the process of satisfying user information need. © 1993 John Wiley & Sons, Inc.