Uma arquitetura para computação sensível ao contexto aplicada a processos de software

Abstract

The present work undertakes the problem of recognizing activities of an actor engaged in a software process. The main goal is to identify the actions that the actor performs with minimal intervention of him/herself. The literature indicates that a way to recognize actions and activities is through the use of sensors attached to the environment. Still, when targeting a specific scope, a very useful approach is to explicitly define and describe a context, e.g. through specific ontologies, as was the choise of this work. An architecture integrating ontologies and sensors is defined. The ontologies describe actions, events and devices with diferent granularities, acting as tools for the activity recognition problem. A specific implementation of the architecture is detailed along with the method used to derive the ontologies. Finally, the results are analyzed and evaluated vis-à-vis existing architectures. The main contribution of this research is a flexible architecture, allowing new sensors to be added in an incremental manner, just extending the ontologies, that are detached from the code.