Abstract
In recent years, recommendation systems have seen significant evolution in the field of knowledge engineering. Usually, the recommendation systems match users' preferences based on the star ratings provided by the users for various products. However, simply relying on users' ratings about an item can produce biased opinions, as a user's textual feedback may differ from the item rating provided by the user. In this paper, we propose SocialRec, a hybrid context-aware recommendation framework that utilizes a rating inference approach to incorporate users' textual reviews into traditional collaborative filtering methods for personalized recommendations of various items. We apply text-mining algorithms on a large-scale user-item feedback dataset to compute the sentiment scores. We propose a greedy heuristic to produce ranking of items based on users' social similarities and matching preferences. To address challenges resulting from cold start and data sparseness, SocialRec introduces pre-computation models based on Hub-Average (HA) inference. Rigorous evaluations of SocialRec (on large-scale datasets) demonstrate high accuracy, especially in comparison with previous related frameworks.