Using Semantic Technology to Improve Recommender Systems Based on Slope One

Slope One is a family of algorithms proposed for collaborative filtering and has been widely deployed on websites’ recommender systems. Compared to SVD, LSI, Similarity Fusion, or some other commonly used algorithms, Slope One often gives better performance in usability, realizability, and efficiency. However, its prediction accuracy is sometimes lower than other expensive methods, because it is a collaborative filtering model only based on average rating difference and cannot meet some special or individual requirements. The user’s and item’s features are also not well considered. In this paper, we propose a new approach for enhancing Slope One using semantic technologies. We explore the implicit relationships between items based on the Linked Data and some measures for computing the semantic distances. The relevance information can be utilized to adjust the weighting when computing the prediction ratings. The approach is easy to be implemented and does not increase the complexity of Slope One hardly. A preliminary experiment is conducted and shows that our approach outperforms the traditional weighted Slope One scheme.