Leveraging Context for File Search and Organization.

Ideally, a user's system would automatically organize, locate, and manage their data. Unfortunately, this ideal is currently elusive, as evidenced by the considerable effort users expend in organizing and finding their data, and the plethora of tools available helping them to do so. Context---the set of facts and features surrounding data---is a promising aspect to aid in this information management problem. For instance, a user's system can observe his or her actions to determine which data is pertinent to other data, reordering and extending results during searches. This dissertation describes and evaluates such a scheme, which uses causally-related inputs to outputs to drive its contextual index. A field study with a prototype of this architecture indicates that the system increases user-perceived satisfaction in search. Unfortunately, while this context proves useful, it is only maintained locally. A user's social neighborhood, however, can provide significant advantages for the collaborative sharing of context. Challenges arise, as naively adding context sharing to a system complicates it from the user's point of view: it becomes a new entity to manage. As well, any such mechanism must also respect personal boundaries on disclosure. This work advocates leveraging acts of sharing to imbue context access rights. With this system, context is encapsulated in frames of reference, and these frames follow objects as they are shared. Frames render a separation of concerns: a recipient holding one particular frame cannot see the context residing in a frame not disclosed. Frames can be split and merged as contexts and circumstances change. To support the ad-hoc, spontaneous collaborations that occur naturally between individuals, the system eschews vestiges of centralized control and its eventual-consistency model aids mobile and disconnected users. A user study suggests this framing scheme is palatable to users and a prototype demonstrates contextual exchange induces negligible performance overheads. This research aims to help ameliorate the information management problem and to facilitate the formation of ad-hoc, spontaneous collaborative groups.Ph.D.Computer Science & EngineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/63735/1/shahsam_1.pd