IMPRECISE INFORMATION AND UNCERTAINTY IN INFORMATION-SYSTEMS

Information systems exist to model, store, and retrieve all types of data. Problems arise when some of the data are missing or imprecisely known or when an attribute is not applicable to a particular object. A consistent and useful treatment of such exceptions is necessary. The approach taken here is to allow any attribute value to be a regular precise value, a string denoting that the value is missing, a string denoting that the attribute is not applicable, or an imprecise value. The imprecise values introduce uncertainty into query evaluation, since it is no longer obvious which objects should be retrieved. To handle the uncertainty, two set of objects are retrieved in response to every query: the set of objects that are known to satisfy with complete certainty and the set that possibly satisfies the query with various degrees of uncertainty. Two methods of estimating this uncertainty, based on information theory, are proposed. The measure of uncertainty is used to rank objects for presentation to a user. © 1990, ACM. All rights reserved.