The Structure of Collaborative Tagging Systems Scott A. Golder and Bernardo A. Huberman Information Dynamics Lab, HP Labs { scott.golder , bernardo.huberman }@hp.com ABSTRACT Collaborative tagging describes the process by which many users add metadata in the form of keywords to shared content. Recently, collaborative tagging has grown in popularity on the web, on sites that allow users to tag bookmarks, photographs and other content. In this paper we analyze the structure of collaborative tagging systems as well as their dynamical aspects. Specifically, we discovered regularities in user activity, tag frequencies, kinds of tags used, bursts of popularity in bookmarking and a remarkable stability in the relative proportions of tags within a given url. We also present a dynamical model of collaborative tagging that predicts these stable patterns and relates them to imitation and shared knowledge. KEYWORD LIST Collaborative tagging, folksonomy, Del.icio.us, bookmarks, web, sharing. 1. INTRODUCTION Marking content with descriptive terms, also called keywords or tags, is a common way of organizing content for future navigation, filtering or search. Though organizing electronic content this way is not new, a collaborative form of this process, which has been given the name “tagging” by its proponents, is gaining popularity on the web. Document repositories or digital libraries often allow documents in their collections to be organized by assigned keywords. However, traditionally such categorizing or indexing is either performed by an authority, such as a librarian, or else derived from the material provided by the authors of the documents (Rowley 1995). In contrast, collaborative tagging is the practice of allowing anyone – especially consumers – to freely attach keywords or tags to content. Collaborative tagging is most useful when there is nobody in the “librarian” role or there is simply too much content for a single authority to classify; both of these traits are true of the web, where collaborative tagging has grown popular. This kind of collaborative tagging offers an interesting alternative to current efforts at semantic web ontologies (Shirky 2005) which have been a focus of research by a number of groups (e.g. Doan, Madhavan, Domingos & Halevy 2002). A number of now-prominent web sites feature collaborative tagging. Typically, such sites allow users to publicly tag and share content, so that they can not only categorize information for themselves, they can browse the information categorized by others. There is therefore at once both personal and public aspects to collaborative tagging systems. In some sites, collaborative tagging is also known as “folksonomy,” short for “folk taxonomy;” however, there is some debate whether this term is accurate (Mathes 2004), and so we avoid using it here. Del.icio.us, the site on which we performed our analysis, allows for the collaborative tagging of shared website bookmarks. Yahoo’s MyWeb does this as well, and CiteULike and Connotea do the same for references to academic publications. Some services allow users to tag, but only content they own, for example, Flickr for photographs and Technorati for weblog posts. Though these two sites do not, strictly speaking, support collaborative tagging, we mention them to illustrate the growth of tagging in a variety of media. In this paper we analyze the structure of collaborative tagging systems as well as their dynamical aspects. Specifically, through the study of the collaborative tagging system Delicious, we are able to discover regularities in user activity, tag frequencies, kinds of tags used and bursts of popularity in bookmarking, as well as a remarkable stability in the relative proportions of tags within a given url. We also present a dynamical model of collaborative tagging that predicts these stable patterns and relates them to imitation and shared knowledge. We conclude with a discussion of potential uses of the data that users of these systems collaboratively generate. 2. TAGGING AND TAXONOMY Proponents of collaborative tagging, typically in the weblogging community, often contrast tagging-based systems from taxonomies. While the latter are hierarchical and exclusive, the former are non-hierarchical and inclusive. Familiar taxonomies include the Linnaean system of classifying living things, the Dewey Decimal classification for libraries, and computer file systems for organizing electronic files. In such systems, each animal, book, file and so on, is in one unambiguous category which is in turn within a yet more general one. For example, lions and tigers fall in the genus panthera, and domestic cats in the genus felis, but panthera and felis are both part of family felidae, of which lions, tigers and domestic cats are all part. Similarly, books on Africa’s geography are in the Dewey Decimal system category 916 and books on South America’s in 918, but both are subsumed by the 900 category, covering all topics in geography. In contrast, tagging is neither exclusive nor hierarchical and therefore can in some circumstances have an advantage over hierarchical taxonomies. For example, consider a hypothetical researcher who downloads an article about cat species native to Africa. If the researcher wanted to organize all her downloaded articles in a hierarchy of folders, there are several hypothetical options, of which we consider four: 1. c:\articles\cats all articles on cats 2. c:\articles\africa all articles on Africa 3. c:\articles\africa\cats all articles on African cats 4. c:\articles\cats\africa all articles on cats from AfricaEach choice reflects a decision about the relative importance of each characteristic. Folder names and levels are in themselves informative, in that, like tags, they describe the information held within them (Jones et al. 2005). Folders like 1. and 2. make central the fact that the folders are