Tag Recommendation in Social Bookmarking System

Abstract

Social Bookmarking System is a web-based resource sharing system that allows users to upload, share and organize their resources. The resources supported by this system are specifically bookmarks and publications. The system has changed the organization of bookmarks from an individual activity limited to a desktop to collective attempt over the web. User can annotate his resource with free form tags that leads to large communities of users to collaboratively create accessible repositories of web resources. Tagging process has its own challenges like ambiguity, redundancy or misspelled tags. Sometimes user tends to avoid ’Tagging’ as they have to describe tags at their own for the resource. These problems result into noisy or very sparse tag space and dilute the purpose of tagging. The effective solution to these problems is Tag Recommendation Sys­tem, that automatically suggest appropriate set of tags while annotating resource. Here we study various approaches attempted for tag recommendation to figure out possible tag resources and methodologies to improve the performance of Tag Recommendation System. We have modeled the tag recommendation task as multi-label classification problem. Our system suggests set of tags for new resource being posted by user in Bibsonomy which is a social bookmarking and publication sharing system. We have used Na¨ıve Bayes classifier along with different representations of dataset like boolean, bag-of-words and continuous value using TF & TFIDF representation. For the sake of result improvement we have also incorporated feature selection technique to choose good representative attributes from dataset for describing each document for which tags are to be recommended.