Association rules are used in recommender systems to develop a model that enhances the profiles of users, as well as to address the cold start problem. Our approach proposes a model which is implemented in a system for recommending scientific papers called Collaborative Topic Regression (CTR). Collaborative Topic Regression consists of two matrices, U and V, where U represents the relationship between users and paper topics, while V represents the relationship between papers and the paper topics. The CTR model is focused on matrix V, by adopting it, and as the result, it is influenced by the paper’s textual content, leaving the content of matrix U essentially unchanged. The depicted model extracts association rules from matrix U, and then enriches it, with the information gleaned from the mining process. The outcomes are based on both, out-of-matrix prediction and in-matrix prediction. Our approach improved the quality of the results by up to 20% for out-of-matrix prediction in the best-case scenario. Unfortunately, the same cannot be said for in-matrix prediction, which will be further investigated in the future.