Interface Design Concepts in the Development of ELSA, an Intelligent Electronic Library Search Assistant.

Several developments must occur before we are likely to see commercially available intelligent intermediary systems for information retrieval. The first, which has received a great deal of attention and research, is the design of appropriate inference engines and knowledge structures. The second is the implementation of computational processes that are sufficiently efficient to handle the potentially overwhelming computational demands of such systems when scaled up. The third is development of full-scale knowledge bases and databases. The fourth is the design of interfaces that support easy and effective use of the intelligent search functions. This paper briefly reviews issues and advances associated with the first concern, designing intelligent search functions. It then discusses in detail issues associated with the design of interfaces to such functions, and describes some general design principles to guide in the implementation of these interfaces. Finally, a functional prototype is described that is based on such principles to illustrate their application. Introduction Many bibliographic information-retrieval systems are now available to information seekers. Unfortunately, the mechanisms devised to access these databases are often inflexible and difficult to use. As a result, expert search intermediaries are sometimes needed to help information seekers achieve more successful results from these systems. Since the intermediary is an expensive resource, designers have attempted to incorporate the intermediary's expertise into knowledge-based systems. It is not, however, sufficient to capture such expertise in order to produce useful systems. Instead, potential functions must be evaluated for usability and an interface which facilitates access to them must be developed. The development of ELSA, an Electronic Library Search Assistant, is offered as a case study in such a design process. Current State of Information-Retrieval Systems Bibliographic databases have become commonplace in many library settings. Slightly over 12 percent of all academic libraries now have online bibliographic retrieval systems. Furthermore, 65 percent of those academic libraries without online retrieval systems have plans to implement such systems in the future (Yee 1991). These systems provide bibliographic information about published documents such as their authors, titles, journals, and abstracts. Most of these systems allow information seekers to retrieve various types of information. For instance, if the information seeker is interested in a specific author, a list of works published by that author can be retrieved. Problems Associated with Current Systems Often, information seekers encounter a wide range of problems when working with such information-retrieval systems. For instance, the information seeker may know the author he is interested in but still be unable to determine how to enter that name into the system. This may be because he has not been able to determine what syntax the system expects. If he is interested in the author William James, he might type in a command such as "Search James, William." If, however, the system expects the author's name to be entered using the form aut/James, william then the search will be unsuccessful. Even seemingly simple errors of this sort are often difficult for the information seeker to detect, diagnose, and correct on his own (Janosky, Smith, and Hildreth 1986). Another possible problem arises when the information seeker manages to execute a query but receives results he did not anticipate. Suppose the information seeker is interested in pollution generated by the use of automobiles. He decides that he would like to see all documents that contain the character strings "pollution" and "automobiles". Given this, he enters the following query: pollution and automobiles (This example assumes he is using a character-string-based search system. …