In a typical database application, there are many different users with a great variety of skills, needs and perceptions. The problem of supporting this plethora of user views in a dynamic, data intensive environment is the topic of this dissertation.

In traditional record-based systems, all information is represented by an idealized data structure and a set of operations on that structure. User views are defined by simple variations in this structure, such as permuting field names, selecting a subset of the data, or creating links between records. Semantic database models support more complex, "natural" structures. It is often claimed that relativism is supported because semantic schemas can be correctly interpreted (by users) in different ways. The object-oriented paradigm, with its simple and elegant structural semantics, provides both simplicity and richness. Unfortunately, current object-oriented systems only provide a single object interface (or protocol). This dissertation presents PolyView; an object-oriented data model capable of simultaneously supporting many points of view. In PolyView, objects encapsulate a single structure and any number of object interfaces (view instance descriptions). PolyView, therefore, supports distributed mappings from user views to the underlying database structure.

Algorithms are presented for generic methods which retrieve and update information through user views. PolyView "colors" queries (messages) by attaching a view identity to them. As messages are propagated through the schema, each receiving object uses the color to determine how the message is to be processed. The color is used to select the user's protocol and allows different user's queries to be processed through apparently different database structures. Because objects act independently, PolyView is a data-driven system; messages are processed without any centralized control or shared memory.

Finally, PolyView provides a set of view transformations which allow view administrators to build object interfaces. Since views are supported by both global and localized mechanisms, there are transformations which operate at each of these levels. There are three major categories of transformations presented in this thesis: those which customize the schema as a whole, transformations for changing the structure of the IS-A hierarchy and transformations for customizing attributes.

Recently some researchers in the areas of database data modelling and knowledge representations in artificial intelligence have recognized that they share many common goals. In this survey paper we show the relationship between database and artificial intelligence research. We show that there has been a tendency for data models to incorporate more modelling techniques developed for knowledge representations in artificial intelligence as the desire to incorporate more application oriented semantics, user friendliness, and flexibility has increased. Increasing the semantics of the representation is the key to capturing the "reality" of the database environment, increasing user friendliness, and facilitating the support of multiple, possibly conflicting, user views of the information contained in a database.

We present an object-oriented semantic database model which, similar to other object-oriented systems, combines the virtues of four concepts: the functional data model, a property inheritance hierarchy, abstract data types and message-driven computation. The main emphasis is on the last of these four concepts. We describe generic procedures that permit queries to be processed in a purely message-driven manner. A database is represented as a network of nodes and directed arcs, in which each node is a logical processing element, capable of communicating with other nodes by exchanging messages. This eliminates the need for shared memory and for centralized control during query processing. Hence, the model is suitable for implementation on a multiprocessor computer architecture, consisting of large numbers of loosely coupled processing elements.

Semantic database models utilize several fundamental forms of groupings to increase their expressive power. In this paper we consider four of the most common of these constructs; basic set groupings, is-a related groupings, power set groupings, and Cartesian aggregation groupings. For each, we define a number of useful restrictions that control its structure and composition. This permits each grouping to capture more subtle distinctions of the concepts or situations in the application environment. The resulting set of restrictions forms a framework which increases the expressive power of semantic models and specifies various set-related integrity constraints.