A vector-space retrieval system for contextual awareness

This paper introduces a retrieval system based on context rather than content. The system uses a set of vector spaces to represent the different contextual characteristics (position, time, sound environment, etc.). In this model both the current context and the items in the corpus are represented by vectors. We then use vector similarities to identify the relevant items given the context. In order to take into account the user´s own perception of the context we also use learning techniques based on the user feedback. This paper also presents two applications we are currently developing, which make use of this retrieval system: a contextual adaptive User Interface selecting the right configuration profiles matching the current context as acquired by a computer, and an augmented reality application for iPhone suggesting to the user activities and places that could be of interest to him given its current environment.