A vector space approach to tag cloud similarity ranking

Abstract

One of the most exciting recent developments in web science is social tagging that enables users to easily annotate web content using free form keywords. Well known examples include Delicious, Flickr, and YouTube which respectively allow users to tag web pages, images, and videos. A tag cloud represents an aggregation of tags to characterize some entity of interest, and it has many potential applications particularly in the context of multimedia information retrieval and recommendation. In this paper, we present a novel method that computes the similarity between tag clouds through effectively incorporating tag similarity information. The considered problem has several unique characteristics mainly due to the informal nature of tag descriptions as well as the frequent tag updates, making it difficult to apply existing approaches in the information retrieval literature. Experimental results on Delicious data show that the proposed scheme can effectively utilize the tag similarity to improve the performance of tag cloud similarity ranking.