Users increasingly rely on their mobile devices to search, locate and discover places and activities around them while on the go. Their decision process is driven by the information displayed on their devices and their current context (e.g. time of day, day of week, weather, traffic, driving/walking etc.). The role of context is particularly important in mobile search because mobile users usually take an action immediately after a local search session (e.g., visit a restaurant, a grocery store, etc.).

Unfortunately, ranking functions today are static ignoring most contextual signals. Given a query and a location, the same set of search results is always returned to the user. In this project, we develop data driven techniques to learn ranking functions that do not only depend on the query and the location of the user but also on the context of the individual user (i.e., explicit preferences, past clicks etc.) and the surrounding environment (i.e., time of day, day of week, weather, traffic, community trends etc.).

BingNow! is the first prototype of our scalable, context-aware search engine that is transparent to the existing search infrastructure. BingNow! currently incorporates temporal, spatial, and user preference signals to create a dynamic local search experience across users, locations and time windows. However, the underlying infrastructure is flexible enough to support any additional contextual signals.

Figure 1. BingNow! WP7 Prototype: the same set of businesses is ranked differently for the same query and location, based on time of day and day of week.