Leveraging Semantic Similarity for Folksonomy-Based Recommendation

To recommend interesting resources such as webpages or pictures that are available through social tagging sites, recommender systems must be able to assess such resources' similarity to user profiles. Here, the authors analyze the role semantic similarity plays in calculating the resemblance between user profiles and published resources in folksonomies. Experiments carried out using data from two social sites show that associating semantics with tags results in more accurate similarities among elements in tagging systems and, consequently, enhances recommendations.