Folksonomies have become a powerful tool to describe, discover, search, and navigateudonline resources (e.g., pictures, videos, blogs) on the Social Web. Unlike taxonomies andudontologies, which impose a hierarchical categorisation on content, folksonomies directlyudallow end users to freely create and choose the categories (in this case, tags) that bestuddescribe a piece of information. However, the freedom aafforded to users comes at a cost:udas tags are defined informally, the retrieval of information becomes more challenging.udDifferent solutions have been proposed to help users discover content in this highly dynamicudsetting. However, they have proved to be effective only for users who have already heavilyudused the system (active users) and who are interested in popular items (i.e., items taggedudby many other users).udIn this thesis we explore principles to help both active users and more importantly new orudinactive users (cold starters) to find content they are interested in even when this contentudfalls into the long tail of medium-to-low popularity items (cold start items). We investigateudthe tagging behaviour of users on content and show how the similarities between users andudtags can be used to produce better recommendations. We then analyse how users createudnew content on social tagging websites and show how preferences of only a small portionudof active users (leaders), responsible for the vast majority of the tagged content, can beudused to improve the recommender system's scalability. We also investigate the growth ofudthe number of users, items and tags in the system over time. We then show how thisudinformation can be used to decide whether the benefits of an update of the data structuresudmodelling the system outweigh the corresponding cost.udIn this work we formalize the ideas introduced above and we describe their implementation.udTo demonstrate the improvements of our proposal in recommendation efficacy andudefficiency, we report the results of an extensive evaluation conducted on three differentudsocial tagging websites: CiteULike, Bibsonomy and MovieLens. Our results demonstrateudthat our approach achieves higher accuracy than state-of-the-art systems for cold startudusers and for users searching for cold start items. Moreover, while accuracy of our techniqueudis comparable to other techniques for active users, the computational cost that itudrequires is much smaller. In other words our approach is more scalable and thus moreudsuitable for large and quickly growing settings.
展开▼
