I'm trying to envision the metadata Gawker’s CMS must collect in order to enable this recommendation feature: