An intelligent semantic-based tag cleaner for folksonomies

The collaborative tagging provided by folksonomy systems is an un-controlled process for users, and the personal and arbitrary tag assignments lead to great tag noises. To solve the problem, authors make contributions as follows: (a) demonstrate that tags assigned to web resources are highly noisy due to the diverse un-controlled present styles of tags; (b) present a two-stage method to clean syntactic and semantic tag noises by taking semantic as the relevance measurement for tags; (c) conduct extensive experiments using dataset collected from del.icio.us. The ratio of the noise tags discovered by our method is up to 40%, and the experiment results show that the proposed method in either semantic approach is highly effective.