Contextual mediation to support ubiquitous computing

The circumstances affecting the use of devices may be termed the context of use. This includes location, social situations, tasks and the characteristics of the device itself. The ubiquitous computing vision predicts that the context of use of computing devices will become much wider and more variable than today. A default presentation of data may not always be ideally suited to the context of its use. The users expectations and experience of data presented will vary widely with context. In more restrictive situations the volume of data which may be available is likely to overwhelm the network, display capability, or user. Where the data can be transformed, or presented in part, then there may be a mediation process to find the best presentation for the context of use. In order to manage this contextual mediation in an automatic and acceptable manner a process which is sensitive to the effects on the user is required. This thesis examines the description of context and data and uses this to enable a general specification to direct contextual mediation. Structured data is described in terms of its semantic content and attributes of alternative representations of semantic elements. Description of context as both states, such as speed and the user's activity; and consumable resources, including network bandwidth and screen space, are presented. Context is used to enable specifications which reflect needs and limitations due to context in a rich and highly flexible manner. Preferences are considered for both semantic and syntactic properties of data. In pursuit of this general solution, we have developed a test-bed application related to the use of map data. Map data is an interesting exemplar as location-correlated data are clearly useful within a ubiquitous computing scenario; and vector map data offers a level of structure which is now emerging in many other classes of data. Results are presented showing that our techniques can indeed both describe and enable operation within a range of contextual constraints. In order to support our claim for generality, a case study of the techniques application to web data is also presented