Recommender systems (RSs) suggest a list of items to users by using collaborative or content-based filtering. Collaborative filtering approaches build models from the user's past behaviors (items previously purchased or selected and/or numerical ratings given to those items) as well as similar decisions made by other users, while content-based filtering approaches utilize attributes of the items to recommend additional items with similar properties. Although RS is applied in many real systems, it has several problems that need to be solved, e.g., cold-start (new users or new items) problem, data sparse problem, and especially data scarcity problem since most of the users are not willing to provide their opinions on the items. In this work, we present a semantic approach to recommender systems, especially for alleviating the sparsity and scarcity problems where most of the current recommendation systems face with. We create a semantic model to generate similarity data given an original data set, thus, the prediction model has more data to learn. Experimental results show that the proposed approach works well, especially for sparse data sets.