The ultimate goal of the proposed research is to provide intelligent and integrated computerized assistance so that searching of computer databases, especially bibliographic and textually-oriented files in biomedical and related areas, can be made more effective and efficient for inexperienced end users as well as for information specialists. Our hypothesis is that a computer assistant can distill and extend the expertise now residing in expert human searchers. Models of the search process will be further developed, regularized, and incorporated into the search assistant so that decison making in the search process can be elevated from an intuitive art to a rational, quantifiable procedure.
The specific aims of the reserch are to extend the advanced techniques for computerized search assistance that we have been investigating; to incorporate these extensions into an enhanced expert experimental computer assistant; to test the techniques by having the computer system employed by a variety of users in hospital, laboratory, library and office locations; and to evaluate the results of these experiments in terms of potential further developments in operational computerized search systems as well as in the models and theories on which developments in information science and technology ultimately depend. Particular features of search assistance to be provided include simplified, user-friendly access to a multiplicity of heterogeneous bibliographic retrieval systems and their databases. Over 300 databases from the major online retrieval systems are thus made easily accessible. Features unique to our proposed search assistance include certain techniques for aiding users in identifying relevant databases, formulating their search topic in a formalized structure, automatically transforming the resulting search topic structure into effective search statements and executing them, evaluating the effectiveness of searches, and suggesting search reformulations. Extensions to our current experimental assistance systems of special significance include techniques for making quantified estimates of recall and cost parameters for searches both before and after they are executed. Recall is estimated based on search strategy comprehensiveness and an analysis of users' relevance judgments.
