Supporting the notion of context within a database environment for intelligent reporting and query optimisation

Information used for decision-making is often semi-quantitative, rather than quantitative in nature. For example, knowing that a student received the top mark in an examination may be more useful than knowing the precise mark achieved; similarly, to be given the information that a company has low market penetration in a given census district might be more important than to be given a numerical value derived from (say) sales divided by average take home pay. The purpose of this paper is to examine how database semantics can be enhanced to provide support for the interpretation of raw data, in order that the significance of the data can be understood. The significance of a reported result is usually dependent on its context, that is, the domain in which it appears, so it is clear that the notion of context must be supported. This paper describes extensions to the object-oriented model which involve equipping it with the means to handle contextual knowledge relating to the patterns of data values, and the means to perform analyses on the data held within the system. The authors hope to make good use of these extensions in two ways. (i) As data is retrieved from the database it can be placed in context, that is, some measure of the significance of reported values can be given. (ii) This new knowledge can be used as a guide to query optimisation, since it allows the optimiser to determine, with greater accuracy, the expected size of the intermediate stages when performing a search operation. This paper concentrates on the rules necessary for the reporting of significance information, along with a control strategy for firing the rules, and a mechanism for the inheritance of significance data.