Abstract
Many collaborative recommender systems leverage social correlation theoriesto improve suggestion performance. However, they focus on explicit relationsbetween users and they leave out other types of information that can contributeto determine users' global reputation; e.g., public recognition of reviewers'quality. We are interested in understanding if and when these additional typesof feedback improve Top-N recommendation. For this purpose, we propose amulti-faceted trust model to integrate local trust, represented by sociallinks, with various types of global trust evidence provided by social networks.We aim at identifying general classes of data in order to make our modelapplicable to different case studies. Then, we test the model by applying it toa variant of User-to-User Collaborative filtering (U2UCF) which supports thefusion of rating similarity, local trust derived from social relations, andmulti-faceted reputation for rating prediction. We test our model on twodatasets: the Yelp one publishes generic friend relations between users butprovides different types of trust feedback, including user profileendorsements. The LibraryThing dataset offers fewer types of feedback but itprovides more selective friend relations aimed at content sharing. The resultsof our experiments show that, on the Yelp dataset, our model outperforms bothU2UCF and state-of-the-art trust-based recommenders that only use ratingsimilarity and social relations. Differently, in the LibraryThing dataset, thecombination of social relations and rating similarity achieves the bestresults. The lesson we learn is that multi-faceted trust can be a valuable typeof information for recommendation. However, before using it in an applicationdomain, an analysis of the type and amount of available trust evidence has tobe done to assess its real impact on recommendation performance.