Capturing, defining, and modeling the essence of context are challenging, compelling, and prominent issues for interdisciplinary research and discussion. The roots of its emergence lie in the inconsistencies and ambivalent definitions across and within different research specializations (e.g., philosophy, psychology, pragmatics, linguistics, computer science, and artificial intelligence). Within the area of computer science, the advent of mobile context-aware computing has stimulated broad and contrasting interpretations due to the shift from traditional static desktop computing to heterogeneous mobile environments. This transition poses many challenging, complex, and largely unanswered research issues relating to contextual interactions and usability. To address those issues, many researchers strongly encourage a multidisciplinary approach. The primary aim of this article is to review and unify theories of context within linguistics, computer science, and psychology. Summary models within each discipline are used to propose an outline and detailed multidisciplinary model of context involving (a) the HUMAN--COMPUTER INTERACTION, 2005, Volume 20, pp. 403--446 Copyright 2005, Lawrence Erlbaum Associates, Inc.

Context-aware technology has stimulated rigorous research into novel ways to support people in a wide range of tasks and situations. However, the effectiveness of these technologies will ultimately be dependent on the extent to which contextual interactions are understood and accounted for in their design. This study involved an investigation of contextual interactions required for route navigation. The purpose was to illustrate the heterogeneous nature of humans in interaction with their environmental context. Participants were interviewed to determine how each interacts with or use objects/information in the environment in which to navigate/orientate. Results revealed that people vary individually and collectively. Usability implications for the design of navigational context-aware applications are identified and discussed.

Abstract. While performing knowledge driven tasks, people at organizations spend time searching the required support information. In order to effectively support knowledge driven tasks, information should be timely and proactively provided to its consumers according to its usage context and patterns. In this paper we define the concepts of action and interaction contexts. More specifically, we define action context as the interplay of actor, role, task and time related features. We focus on actors’ capabilities for handling several action contexts. Interaction contexts are defined as the supporting components of actors ’ interactions related to business processes. Actors and business processes are approached as a network of action and interaction contexts managed by an “operating system”. We relate the defined concepts and propose a business process oriented and context-based modeling approach which aims to provide a richer, efficient and flexible semantics, capable of better representing the complexity of actual business process execution and facilitating a proactive and timely knowledge support of business actors. We illustrate some of these ideas with a working example. 1.

The web log data embed much of web users' browsing behavior. From the web logs, one can discover patterns that predict the users' future requests based on their current behavior. These web data are very complex due to their large size and sequential nature. In the past, researchers have proposed different methods to predict what pages will be visited next based on their present visit patterns. In this paper, we extend this work to discover patterns that can predict when these web page accesses will occur. Our method is based on a novel extension of association rule classification method. We extend the traditional association rules by including the temporal information explicitly in each rule, and reason about the confidence of each prediction in terms of its temporal region. We compare two different methods for temporal event prediction, demonstrate the effectiveness of our methods empirically on realistic web logs, and explore the tradeoff between prediction accuracy and data mining time for our models.