International Workshop on Ubiquitous Data Management
(UDM2005)

 Mediating between available information objects and individual information needs is a central issue within the Information Society. In the simplest case this is an information request answered by a search engine based on the analysis of the content of individual information objects within an information collection that may be a digital library, but also the World Wide Web. However, neither the information search activity nor the information objects within the collection are isolated entities. They are both equipped with a multifaceted context. This talk discusses complementing ways to make such context explicit and to use it to improve the information mediation process.
 In general “the context of something consists of the ideas, situations, events, or information that relate to it and make it possible to understand it fully”. When information (or knowledge) is represented, context is generally left implicit. This strategy enables concise representation and communication of information under the assumption that relevant parts of the context are shared or given as background knowledge. If context is to be explicitly represented, as it is, for example, done in knowledge representation, it can be used for reasoning and decision making. In order to enable an effective use of context, it is important that the “right” aspects of the context are to be modeled and that the relationship between the context representation and the representation of the “something” in the context is clearly defined. This also holds true, when using context in intelligent information access support.
 The most obvious and common way to represent context for information objects are metadata like bibliographic metadata that embed an information object in information mainly about its creation process. This is, however, capturing only part of the original context an information object was created in. Many information objects, like e.g. propaganda films, can only be fully understood, if more information about their original context is available, like e.g. the time, the political situation or the history of the creator. This original context is in most cases different from the context, in which the information object is perceived and inspected (perceived context). This perceived context can, for example, be represented by classification information and by the enclosing information collection, which also contribute to the interpretation of an information object. Furthermore, hypertext documents are embedded into a context of linked documents and can only be fully interpreted if they are understood as part of such information object networks. Some retrieval methods already exploit this type of context. However, further explicit context representation is required to create awareness for the original context. This can for example be done by enriching information objects with annotations. Annotations can even be used to embed an information object in the context of a discourse. If the different facets of the context of an information object, the information context, is made explicit and analyzed, it can be used in the mediation process basing the information access process on a richer picture of the information object.
 The second entity in the information mediation process, the information request or more general the information access activity, is part of a user task that provides a context to this activity. This context, the user context, is determined by the task, but also by user interests, skills and other user characteristics, as well as results of preceding activities. Part of this context is captured by personalization approaches that model e.g. user interests and use it for individual information filtering. Extended user (context) modeling makes further dimensions of the context, like e.g. the relationship a user is involved in or the current user task, explicit requiring new matchmaking method for exploiting them in the information mediation process.
 Both, the user context as well as the information context, are captured by some type of metadata building upon a metadata schema like e.g. the Dublin Core (DC). The interpretation of metadata by humans as well as by machines relies on a shared understanding of the concept underlying the metadata schema. Thus, one abstraction level higher, Ontologies provide context to the interpretation of metadata. This context can be made explicit by referring to formally defined ontologies that define the concepts used in the metadata schema e.g. on the basis of more basic terms and predicates. Making this context explicit it is possible to combine metadata based on different metadata schemata by tracing them back to a shared context (the more basic concepts).

 The success of the Web has made popular the idea that from now on every information could be made available to anybody, anytime, and anywhere. While this has proven to be a very useful goal, another paradigm is taking off thanks to the development of mobile computing. This paradigm reads as the opposite to the previous one, as it aims at making information available to a limited group of users (those in the geographical vicinity of the information sources), mainly during some appropriate periods (e.g., shopping hours, weekdays versus week-ends), and only within some geographical space determined by local information sources. Location-based services is the key motto for this paradigm in current research on ubiquitous computing. This talk overviews the architectural framework that characterizes location-based services, and analyzes the main semantic issues that arise in this specific context.