Convergence of User-Generated Semantic Metadata in Social Networking Applications

Generating metadata that is useful to other users is one of the primary advantages of social media applications. This activity occurs in many Web 2.0 applications, including Twitter and its use of hashtags, Foursquare and its location data, and Delicious and its tags of bookmarks. A recurring problem with this user-generated metadata, however, is that users often fail to converge on one single term in metadata used to represent the same object. Thus multiple tags in Delicious point to the same URL though they represent the same semantic concept, or the same geolocation in Foursquare has multiple user-generated names. This failure to converge on common terminology when creating user-generated metadata hampers later use of that metadata, for example when a search of a tag returns only a partial set of relevant search results. This paper describes two approaches to handling the variances that will unavoidably exist between semantically similar pieces of metadata, using the People & Projects social networking application as a testbed for these approaches and its tagging metadata as a sample implementation of this methodology. Tag usage data from People & Projects is presented to illustrate the problem and evaluate the effectiveness of the solution that was implemented.