Characterizing and supporting cross-device search tasks

Web searchers frequently transition from desktop computers and laptops to mobile devices, and vice versa. Little is known about the nature of cross-device search tasks, yet they represent an important opportunity for search engines to help their users, especially those on the target (post-switch) device. For example, the search engine could save the current session and re-instate it post switch, or it could capitalize on down-time between devices to proactively re-trieve content on behalf of the searcher. In this paper, we present a log-based study to define and characterize cross-device search be-havior and predict the resumption of cross-device tasks. Using data from a large commercial search engine, we show that there are dis-cernible and noteworthy patterns of search behavior associated with device transitions. We also develop learned models for predicting task resumption on the target device using behavioral, topical, geo-spatial, and temporal features. Our findings show that our models can attain strong prediction accuracy and have direct implications for the development of tools to help people search more effectively in a multi-device world.