Social Recommender Systems
In this contribution we review and discuss limits and chances of social recommender systems. After classifying and positioning social recommender systems in the basic landscape of recommender systems in general via a short review and comparison, we present related work in this more specialized area. After having laid out the basic conceptual grounds, we then contrast an earlier study with a recent study in order to investigate the limits of applicability of social recommenders. The earlier study replaces rating-similarity-based neighbourhoods in collaborative filtering with subgraphs of the user’s social network (social filtering) and investigates the performance of the resulting classifier in a taste related domain. The other study which is discussed in more detail investigates the applicability of the method to recommendations of more factual, content-oriented items: posts in discussion boards. While the former study showed that the social filtering approach works very well in taste related domains, the second study shows that a mere transplantation of the idea to a more factual domain and a situation with sparse social network data does perform less satisfactorially.