Getting beyond Boole

Although most computer-based information search systems in current use employ a Boolean search strategy, there is by no means a clear consensus throughout the information retrieval research community that the conventional Boolean approach is best. The well-known drawbacks of the Boolean design include an inhospitable request formalism, frequent null output and output overload, and lack of provision for differing emphasis on different facets of the search. Nontraditional design principles that overcome these problems are already known and available in the research literature. In this article several such alternative approaches are sketched and their advantages over the Boolean design indicated. In the 195Os, the era in which serious thought was first given to the possibility of computerized information searching, it was proposed that search requests might advantageously be formulated as Boolean combinations of document descriptors. This suggestion seemed to meet with the immediate approval of most mathematicians, computer scientists, and technically oriented information professionals. At that time only Bar-Hillel, a mathematical logician, objected strenuously [I]. A decade later, when the first large-scale bibliographic retrieval services were set up, the Boolean approach was adopted as the underlying retrieval strategy. Since then it has become the more-or-less standard search mode for almost all the commercial search services and in most automated library catalogs. It is also used in the command languages of many database management systems, office information systems, personnel search systems, and various other information access programs for scholarly, institutional, or personal use. In fact, insofar as search systems in actual operation today are concerned, the Boolean request form is quite ubiquitous. Thus it may come as a surprise to some readers to learn that specialists in information retrieval are by no means unanimous in their praise of the Boolean approach, that the research literature is full of alternative proposals, and that knowledgeable information scientists who think the standard Boolean design could be significantly improved on probably constitute an overwhelming majority. Admittedly, there is as yet no clear consensus among researchers as to which of the many available non-Boolean designs is best, and this lack of a single clear alternative candidate has doubtless been a factor tending to perpetuate the current monopoly of Boolean systems in the marketplace. Nevertheless, it is not difficult to point to well-researched retrieval strategies that are clearly superior to the Boolean in at least some important respects. This article sketches briefly a few of these possibilities for the benefit of system designers who might not otherwise be aware of them. A commercial implementation of any of them would be a practical advance and might help the information community to break through the Boolean barrier toward some of the more sophisticated designs that are already familiar to information retrieval researchers and experimenters. Several proposals will be sketched in order of increasing sophistication and decreasing conformity with the conventional Boolean design. Although some of them may be novel in detail, the general principles behind these designs are all to be found in the research literature. The design ideas will be presented in the form of certain problems inherent in the Boolean search logic, and the proposed post-Boolean solutions to these problems. PROBLEM 1: THE UNFRIENDLINESS OF BOOLEAN FORMULAS Those who were initially enthusiastic about Boolean retrieval in the 1950s and 1960s were presumably computer people and other mathematically minded folk who already