Strands Labs

Social Recommendation

In today's high-bandwidth, digital world our attention has become a scarce and precious resource. Information about events, products, friends, entertainment, finances, recreation, and so much more continually competes for our attention. More often than not, we consult others to recommend items which might interest us to filter the torrent. Strands' Social Recommendation technologies make it possible to ask friends, experts or even the whole world to recommend whatever we might be seeking.

To find items using search technology we have to think enough about what we want to find to describe it with keywords. Strands social recommendation technologies in effect perform that thinking for us by building models of how people interact with information consisting of hierarchically structured links between items. When a user requests recommendations, the system consults the models to find a set of possible recommendations. The recommender then filters this initial set of possible recommendations using knowledge of the user's preferences, previous recommendations, and the context of the recommendation request to produce a final set of personalized recommendations.

Several key features distinguish Strands social recommendation technologies: The recommendation system inherently is content and platform agnostic, content awareness is captured in the models for the relationships between items which may be provided explicitly or automatically learned over time. Real-time recommendations are available for new users with no training and increasingly personalized as the system accumulates knowledge about the user. Similarly, new items inserted into the models are immediately recommendable based on approximate relationships with previously existing items and then increasingly recommended based on their relationships with other items learned from user behavior. Finally, Strands' recommenders offer fast response times and robust scalability.