Though there are now numerous examples of multimedia systems in the commercial market, these systems have been developed primarily on a case-by-case basis. The largescale development of such systems requires a principled characterization of multimedia systems which is independent of any single application. It requires a unified query language framework to access these different structures in a variety of ways. It requires algorithms that are provably correct in processing such queries and whose efficiency can be appropriately evaluated. In this paper, we develop a framework for characterizing multimedia information systems which builds on top of the implementations of individual media, and provides a logical query language that integrates such diverse media. We develop indexing structures and algorithms to process such queries and show that these algorithms are sound and complete and relatively efficient (polynomial-time). We show that the generation of media-events (i.e. generating di...

Knowledge discovery in databases (or data mining) , which extracts interesting knowledge from large databases, represents an important direction in the development of data- and knowledge- base systems. With fruitful research results on knowledge discovery in relational databases and the emerging trend in the development of object-oriented and active database systems, it is natural to investigate knowledge discovery in object-oriented and active databases. This paper overviews the mechanisms for knowledge discovery in object-oriented and active database systems, with an emphasis on the techniques for generalization of complex data objects, methods, class hierarchies and dynamically evolving data, and on the integration of knowledge discovery mechanisms with production control processes. 1 Introduction With rapid growth in the amount of information stored in databases, the development of effective and efficient tools for knowledge discovery in databases (KDD, or data/knowledge mining) ...

. We describe some preliminary exploratory work in the use of the GRAIL [2] description logic and GALEN terminology server architecture [24] to support classification of art images, based on our experiences of using GRAIL to support medical applications using clinical terminology [24]. We hope to apply lessons learnt in the medical domain and test whether the decisions made in the design and implementation of GRAIL and the terminology server architecture help or hinder efforts in a different domain. 1 Terminology and Art The use of terminologies to describe images is not new and predates computers. The principle of descriptive analysis of fine art was expounded by the Prague School in the 1920s [31] in the creation and application of a structured semiotic terminology. The terminology could be used to describe and automatically classify works of art by their content, and moreover identify patterns of change with reference to the social and environmental contexts of the artists and iden...

This thesis examines the potential for reduction in software complexity of applications developed under the Single Address Space Operating System Sombrero versus those developed under a conventional process-oriented operating system, Windows 2000. To test Sombrero`s impact on development costs, a set of sample database applications with varying functionality was developed under Windows 2000 and Sombrero. Using accepted software engineering metrics, the software complexity of these systems was compared. The results obtained indicate a substantial reduction of software development costs and defects in a single address space environment such as that provided by Sombrero. iv DEDICATION To my family for all their support. In memory of my Mother, to whom this day was so important. v ACKNOWLEDGEMENTS Many thanks for the effort and support of my committee members Dr. James Collofello, and Dr. Susan Urban. Special thanks to my committee chair, Dr. Donald Miller. In addition I would like to thank the members of the 64-Bit Operating System Group, particularly Alan Skousen, for their support and also Power Software Limited for the evaluation license of Krakatau Professional (C++) Metrics. vi TABLE OF CONTENTS Page LIST OF TABLES ........................................... ix LIST OF FIGURES ........................................... x CHAPTER 1 -

The presentation layer of hypermedia systems could benefit from standard object querying functionality and this is most effective if strong typing is enforced. By strong typing we mean the direct representation of data semantics as object types in an object database as oposed to a "slotted frames" representation. On the other hand, the flexible emergent nature of structure must be considered in the authoring activity and in this sense premature typing and organizing is counterproductive. Reflecting on these apparently contradictory issues and the past proposals to handle the problem, we extend the strongly typed data model of a prototype hypermedia system, WorldView, to support semi-automatic object submission and type metamorphosis. Weak types are also necessary for some constructs, so they coexist with strong types, but these are enforced. We emphasize the benefits available to the presentation layer of keeping a uniform object oriented structure. In particular we implement a dynamic linking capability that uses queries to retrieve the objects related to some object attribute and suggest other improvements. We stress that most object oriented hypermedia systems are frame-based, especially in what concerns user-defined and emergent structure.

We describe some preliminary exploratory work in the use of the GRAIL [2] description logic and GALEN terminology server architecture [24] to support classification of art images, based on our experiences of using GRAIL to support medical applications using clinical terminology [24].