Data Recovery from Distributed Personal Repositories

We present an approach to personal disaster recovery, e.g. after a hard-disk crash, based not on an explicitly ex-ante defined recovery plan with a rigid backup regime, but rather on naturally accumulated and distributed sources of personal data, such as e-mails and their attachments. We aim to restore as much data as possible, and to provide means to organise it in a meaningful folder structure. Employing information retrieval techniques, we semi-automatically establish the context of and relations between the data objects along several different dimensions, to identify relations and groups. Different views at multiple levels of granularity then allow an interactive organisation into folders.