| Subject searches in a bibliographical database system (BDS) generally have been considered as difficult to conduct successfully. The difficulties stem from uncertainty that underlies classification and browsing activities. We present a model to deal with uncertainty in the classification and browsing processes in a BDS. The model is based on the Dempster-Shafer theory of evidence.;The semantics of the Dempster-Shafer theory is investigated in the context of a BDS. Different schools of thought in statistics are based on different representations of uncertainty and different modes of inference. The questionable superiority of the Dempster-Shafer theory over the other schools in improving document classification and browsing processes is explored.;Concepts like hierarchy of classes, ambiguous relevance, inexact classification, and pooled classification, are discussed using the nomenclature of belief functions and Dempster's rule. The model thus gives a normative framework in which one can describe and address many problematic phenomena which characterize the way people classify and retrieve documents.;This model is incorporated in a prototype computer system. Our system is tested against an alternative system in a laboratory experiment. The test results showed that our system outperformed the alternative system in user acceptance, effectiveness, and external validity. Aside from its practical merit, our model is interesting on theoretical grounds because it gives a canonical example in which the Dempster-Shafer theory "makes sense". |
