Tag Clustering with Self Organizing Maps

Free registration required

Executive Summary

Most of the users today navigate and organize collections of resources through user-generated tags. But, due to lack of explicit semantics and differing use of tags between users, there value is still limited. One way of solving these problems is clustering techniques, which find groups of related tags. This paper shows how a Self-Organizing Map (SOM) can be used to cluster tagged bookmarks. Self organizing maps are artificial neural networks. These maps are designed to map high-dimensional data points to nodes in a low-dimensional grid, the output layer. In most of the cases, y the grid is two-dimensional and can be viewed as a graphical map. The similar data points are placed close together in a map or at the same point on the map. The paper presents an iterative method for determining the optimal number of clusters. This method is also tested for its feasibility. The paper also discusses ways of using SOM to intuitively classify new bookmarks into a set of clusters. Clustering also helps to automate the process of extraction of formal vocabularies from unstructured folksonomies. A self organizing map has been suggested here to cluster tagged bookmarks taken from the website 'Delicious'. This website is used by people to save bookmarks to the website and associate tags with each bookmark.

  • Format: PDF
  • Size: 810.3 KB