Resumen de Contribution to improve mobility uses through context-awareness
Elena Yndurain Gil
Dey, in his paper “Towards a Better Understanding of Context and Context-Awareness”, argues that context-awareness is important in applications in which the user’s context changes rapidly, such as in mobile environments for ubiquitous computing. In his paper, Dey defines context as “any information that can be used to characterize the situation of an entity”. In mobile environments, the entity is the mobile device itself. The device is both pervasive and person-centric; it can continuously capture information about its users and their context through its sensors. The use of context has gained importance in ubiquitous computing since the 1990s, and the technique has recently been used in mobile devices to improve their uses and applications. For mobile context-awareness to become a reality, further research is required, particularly in the field of context prediction, which can expand the possibilities of context-awareness applications by expanding the applications’ situation awareness. In this PhD dissertation, we focus on the use of data obtained through mobile device sensors and user behavior to derive and predict context to improve mobility for both the users’ experience and for the applications’ functionality. We contribute to context-aware mobile computing by showing how mobile devices can automatically learn from the user’s context and can adapt to improve the mobile experience. We begin our work with a state-?of-?the-?art analysis of “context-awareness” proposals for mobile systems and applications and of the current tools used to infer context from the existing environmental variables. In this dissertation, we analyze the existing gaps in mobile environments and propose solutions to resolve these issues. We first define “context-awareness” and propose an architecture to predict context from a mobility perspective. Numerous definitions of context, context-awareness and architectures exist, but few focus exclusively on mobility. Moreover, all of the definitions are oriented towards context inference rather than towards a prediction of future context. We develop a model that captures, processes and unifies variables from heterogeneous sources for use by a machine-learning algorithm that infers and predicts the context. We also test and benchmark several machine-learning algorithms in our architecture so that we can recommend those algorithms that we consider most appropriate for inferring context in mobility environments. We propose the combination of on-?line prediction algorithms and classifier algorithms to enhance context derivation with future context prediction. We evaluate our proposal utilizing real data from the Reality Mining project, which captures data from the daily mobile usage of c.100 Nokia smart phones during an academic year. We conclude with an example of how to apply our proposed architecture and model, and we demonstrate its enrichment of the search experience with a mobile device by including a “context-awareness” module in mobile search engines. We use Bing as the search engine for all of our search examples. ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
