Semiotic Dynamics in Online Social Communities

A distributed classification paradigm known as collaborative tagging has been successfully deployed in large-scale web applications designed to organize and share diverse online resources. Communities of web users categorize resources by associating metadata with them, in the form of freely chosen text labels, or tags. Here we regard tags as basic dynamical entities and study the semiotic dynamics underlying collaborative tagging. We collect data from a popular system, focusing on tagging data associated with a given resource, and report our experimental findings. Remarkably, we observe a universal power-law behavior for the dynamics of tag accumulation. On studying the frequency distribution of tags, we find a generalized Zipf’s behavior and quantitatively describe the observed distributions in terms of a previously introduced YuleSimon process with heavy-tailed memory. 1 Collaborative Tagging A new paradigm has been quickly gaining ground in online information systems: Collaborative Tagging [1, 2]. In web-based applications like del.icio.us1, Flickr2, CiteULike3, Connotea4, users enrich resources (photos, scientific references, web pages and more) with semantically meaningful metadata in the form of text labels or “tags”. Tags are freely chosen and independently associated with resources. They are also used to navigate the resources entered into the system. The relationships between resources and tags that dynamically arise from the uncoordinated tagging activity of users are automatically made available to the entire the user community (see Fig. 1) so that the tagging process develops social aspects and complex interactions. Indeed, despite http://del.icio.us/ http://flickr.com/ http://citeulike.org/ http://connotea.org/