We're trying to focus it on specific data sets that users have more control over. We're not using the "keyword" approach like a Kontera or Vibrant allows you to link a lot, but creates a lot of crap. In Seat Geek's case we started with a simple database of the remaining NFL games; there's a limit on what can get linked so it doesn't get out of hand. We have another data set of product names that link product names to retailers and create affiliate links for people who do reviews or have users that talk about a certain product category. The goal is to focus on uniform data sets that can keep the links specific and relevant.