Programming Social Applications

Relationships Versus Entities

As we explore the concept of user relationships on the Web, we realize that a user’s social graph extends far beyond her connections to other people and groups. A user may interact with many more common interest sources on the Web, moving far beyond a person-to-person relationship cluster. This is where the concept of entity relationships comes into play.

Entities consist of any links to a root user’s behaviors, such as searches she conducts, websites she visits, or online purchases she makes. We can map a user to these entities with the intention of grouping her into additional targeted clusters. In doing so, we extend the traditional model of the social graph into a rich, interwoven online personality, thereby helping companies and applications personalize a user’s web experience through highly funneled advertising or specially targeted products.

Let’s break down the entity and relationship links into a practical example. Suppose we host an application on a social network that provides content and ads targeted to each user. Suppose also that this social network provides web search capabilities, tracks the searches that a user makes, and has a social graph providing links to her family and friends. If we look at the user’s search history, we see that she searches for fishing and camping equipment often. We now know that our user has some interest in fishing and camping. If we then correlate that hobby with her relationships to find any linked users who ...