Using Wikipedia to alleviate data sparsity issues in Recommender Systems

This paper proposes that Wikipedia can effectively be used in order to lessen the negative effects of data sparsity on the accuracy of recommendations produced by Recommender Systems, provided that domain resources available for recommendation can successfully be mapped to Wikipedia articles. Under the assumption that hyperlinks between Wikipedia articles convey latent semantic relationships between the concepts they represent, we argue that by representing domain resources as a set of interconnected Wikipedia articles the volume of information available to a recommender algorithm increases, enabling it to improve its performance. The approach is evaluated using two real world datasets, giving positive results.