Urban Maps of Social Activity

Recreational queries from users looking for ideas of what to do or see and where to go are very common in desktop, mobile and, increasingly, contextual search scenarios. Such queries typically contain what, where, when components as the user seeks future activities, whether in real-time or future trip planning. Often, the user will have additional constraints and requirements for what they are seeking, such as suitability for kids, budget, for a romantic date, etc. Currently, simple recreational queries (e.g. ``restaurants in Mountain View’'' are served by static local results, or the Points of Interest thumbnail carousel (e.g. ``things to see in San Francisco''). More complex recreational queries, such as ``romantic places to eat in San Francisco friday night'', or ``educational places to visit with kids nearby'' require the user to click on the 10 blue links to read through articles from sites such as TripAdvisor and WikiTravel to satisfy their needs. Employing location based social network data, we construct urban maps of social activity for answering recreational queries using a model based on social, geographical, and temporal information. We demonstrate the feasibility of our approach using a data set of 1B check-ins.