Probabilistic approaches to tag recommendation in a social bookmarking network

Abstract

Tagging has become increasingly popular with the explosion of user-created content on the web. A 'tag' can be defined as a group of keywords that makes organizing, browsing and searching for content more efficient. Users apply tags to a variety of web-based, shareable content including photos, videos, news articles, bookmarks, friends, etc. Tag suggestions for blog posts or web-pages have changed the focus of the tagging process from generation to recognition, thus making it less time and effort intensive. We propose tag recommendation algorithms for personalized agents, that recommend tags for bookmarks stored in a popular social bookmarking website, Del.ici.ous [6]. Our tag recommender agents learn to classify the tags according to their semantic similarity based on collaborative tagging by the users. Hence this approach can be used to facilitate folksonomy formation for the social network. In this paper, we first empirically verify our hypothesis that web pages with similar content are tagged with similar tags. We compare both Content-based and Collaborative approaches to recommend tags to the users. We analyze the performance of two probabilistic approaches to recommend tags from users with similar tagging behavior. © Springer-Verlag Berlin Heidelberg 2012.