Improving the exploration of tag spaces using automated tag clustering

Abstract

Due to the increasing popularity of tagging, it is important to overcome challenges resulting from the free nature of tagging, such as the use of synonyms, homonyms, syntactic variations, etc. The Semantic Tag Clustering Search (STCS) framework deals with these challenges by detecting syntactic variations of tags and by clustering semantically related tags. We evaluate our framework using Flickr data from 2009 and compare the STCS framework to two previously introduced tag clustering techniques. We conclude that our framework performs significantly better in terms of cluster precision compared to one method and has a better average precision compared to the other method. © 2011 Springer-Verlag.