Measuring Similarity in Large-Scale Folksonomies

Social (or folksonomic) tagging has become a very popular way to describe content within Web 2.0 websites. Unlike\ud taxonomies, which overimpose a hierarchical categorisation of content, folksonomies enable end-users to freely create and choose the categories (in this case, tags) that best\ud describe some content. However, as tags are informally de-\ud fined, continually changing, and ungoverned, social tagging\ud has often been criticised for lowering, rather than increasing, the efficiency of searching, due to the number of synonyms, homonyms, polysemy, as well as the heterogeneity of\ud users and the noise they introduce. To address this issue, a\ud variety of approaches have been proposed that recommend\ud users what tags to use, both when labelling and when looking for resources. As we illustrate in this paper, real world\ud folksonomies are characterized by power law distributions\ud of tags, over which commonly used similarity metrics, including the Jaccard coefficient and the cosine similarity, fail\ud to compute. We thus propose a novel metric, specifically\ud developed to capture similarity in large-scale folksonomies,\ud that is based on a mutual reinforcement principle: that is,\ud two tags are deemed similar if they have been associated to\ud similar resources, and vice-versa two resources are deemed\ud similar if they have been labelled by similar tags. We offer an efficient realisation of this similarity metric, and assess its quality experimentally, by comparing it against cosine similarity, on three large-scale datasets, namely Bibsonomy, MovieLens and CiteULike