Skip to Main Content
Previous email prediction algorithms generate individual predictions based on the past groupings of recipients or the contents of past emails. Our work builds on this research by (a) introducing new algorithms for extending and combining previous techniques and generating hierarchical recipient predictions and (b) comparing the previous algorithms with each other and the new algorithms. We used standard metrics and developed new metrics to measure three kinds of user effort: scanning predictions, selecting predictions, and manually entering recipients. The new metrics are based on a new abstract model of recipient prediction that applies to existing schemes and the new ones developed by us. Our evaluations, based on the Enron mail database and the Gmail user-interface for recipient prediction, show that (a) content is less effective than groups, (b) the combination of content and groups is less effective than groups alone, and (c) hierarchical recipient prediction reduces user effort.