The Complex Dynamics of Collaborative Tagging Problem Tagging
The Complex Dynamics of Collaborative Tagging
Problem • Tagging distributions tend to stabilize into powerlaw distributions. • empirically determine as to why this pattern occurs and try to determine patterns prior to a stable distribution. • The central question involved is when and how a coherent classification system might emerge from distributed tagging?
Solution • Analyse tagging behaviour and their emperical results show that it does follow a powerlaw distribution. • Dist(Ti, Tj) = N(Ti, Tj) / sqrt( N(Ti) * N(Tj) ) • Based on these similarities they construct a tag-tag correlation graph or network. • Size of Node = frequency of the tag. • Distance = inverse of similarity between tags.
Criticism • The information value of a tag suffers over time. • Still human intervention needed. • (Adversarial Nature) Tag Spam. • There is no correspondence between keywords and concepts and it is not possible to do so with few keywords. • Probably a combo of explicit (people) and implicit (software) ranking methods to get better tags and more useful search results.
Relation to what we learned in class • Bottom-up approach to taxonomies/ontologies. • It is a flat hierarchy. • User Regulation – difficult. Because the user still uses keywords. • The long tail phenomena.
- Slides: 6