Using expert systems to implement a semantic data model of a large mass store system

Abstract

The successful development of large volume data storage systems will depend not only on the ability of the designers to store data, but on the ability to manage such data once it is in the system. Our hypothesis is that mass storage data management can only be implemented successfully based on highly “intelligent” meta data management services. Such services would allow database administrators and users to manipulate, update, and access data and related information and knowledge in a logical manner, and yet would be powerful enough to support the performance needs of a large mass store system. Historically, there have been attempts at building data management services for very large volume data systems; however, when the amount of data being managed got large the meta database itself failed as a consequence of its own size and complexity. However, if the standard were expanded to include the semantics and pragmatics of the data domain using a Semantic Data Model (SDM), the result would be an overall information system design that is organized and accessible in the context of the data's user and purpose. The implementation of an SDM requires the application of Artificial Intelligence (AI) and related computer science technologies such as object oriented representation, property inheritance, and rule-based decision making. Presently, there does not exist unique software for developing SDMs that addresses the complex representation of data meta data and related information and knowledge. This paper presents the results of a demonstration prototype SDM implemented using the expert system development tool NEXPERT OBJECT. In the prototype, a simple instance of a SDM was created to support a hypothetical application for the Earth Observing System, Data Information System (EOS-DIS). The massive amounts of data that EOSDIS will manage requires the definition and design of a powerful information management system to support even the most basic needs of the project. The application domain is characterized by a semantic-like network that represents the data content and the relationships between the data based on user views and on more generalized domain architectural views of the information world. The data in the domain are represented by objects that define classes, types, and instances of the data. In addition, data properties are selectively inherited between parent and daughter relationships in the domain. Based on the SDM, a simple information system design is developed from the low-level data storage media, through record management and meta data management, to the user interface.