Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe
AKSOE: Arbeitskreis Physik sozio-ökonomischer Systeme
AKSOE 11: Social-, Information-, and Production Networks I
AKSOE 11.2: Vortrag
Mittwoch, 27. Februar 2008, 13:30–14:00, EW 203
Networks of tag co-occurrence and measures of relatedness in social tagging systems — •Ciro Cattuto1,2, Dominik Benz3, Andreas Hotho3, Gerd Stumme3, and Andrea Baldassarri4 — 1Centro Studi e Ricerche “Enrico Fermi”, Compendio Viminale, 00184 Roma, Italy — 2Complex Networks Lagrange Laboratory (CNLL), ISI Foundation, 10133 Torino, Italy — 3Knowledge & Data Engineering Group, University of Kassel, 34121 Kassel, Germany — 4Phys. Dept., Università di Roma “La Sapienza”, P.le A. Moro 2, 00185 Roma, Italy
Social tagging systems allow web users to organize and share resources by associating them with free-form keywords (tags). The popularity of these systems has surged to a point where their study is important both for scientific and technological reasons. Their underlying data structures are hypergraphs (known as folksonomies) collaboratively built by the unsupervised activity of users: understanding their structure and evolution poses promising challenges in different fields of research. Crucial concepts are those of tag (node) similarity and tag (node) relatedness. We show that a bridge can be developed between statistical measures of tag relatedness in the folksonomy and standard notions of taxonomic distance in formal representations of knowledge. We use data from the social bookmarking system del.icio.us to analyze three distributional measures of tag relatedness (tag co-occurrence, cosine similarity and FolkRank, an adaptation of PageRank to folksonomies) and provide a solid semantic grounding of our findings by mapping the nodes of the folksonomy hypergraph into a large taxonomic database of English, and applying there standard measures of semantic similarity.