期刊名称：International Journal of Data Mining & Knowledge Management Process
印刷版ISSN：2231-007X
电子版ISSN：2230-9608
出版年度：2014
卷号：4
期号：1
页码：19
DOI：10.5121/ijdkp.2014.4102
出版社：Academy & Industry Research Collaboration Center (AIRCC)
摘要：Social bookmarking system is a web-based resource sharing system that allows users to upload, share andorganize their resources i.e. bookmarks and publications. The system has shifted the paradigm ofbookmarking from an individual activity limited to desktop to a collective activity on the web. It alsofacilitates user to annotate his resource with free form tags that leads to large communities of users tocollaboratively create accessible repositories of web resources. Tagging process has its own challengeslike ambiguity, redundancy or misspelled tags and sometimes user tends to avoid it as he has to describetag at his own. The resultant tag space is noisy or very sparse and dilutes the purpose of tagging. Theeffective solution is Tag Recommendation System that automatically suggests appropriate set of tags touser while annotating resource. In this paper, we propose a framework that does not depend on tagginghistory of the resource or user and thereby capable of suggesting tags to the resources which are beingsubmitted to the system first time. We model tag recommendation task as multi-label text classificationproblem and use Naive Bayes classifier as the base learner of the multilabel classifier. We experiment withBoolean, bag-of-words and term frequency-inverse document frequency (TFIDF) representation of theresources and fit appropriate distribution to the data based on the representation used. Impact of featureselection on the effectiveness of the tag recommendation is also studied. Effectiveness of the proposedframework is evaluated through precision, recall and f-measure metrics.
关键词：Tag recommender; multilabel classification; social bookmaking
