Personalized recommendation systems play a major role in discovering relevant items for users within a digital service platform. Matrix factorization based collaborative filtering models such as Alternating Least Squares (ALS) have shown good results in predicting user-item interactions. Nevertheless, traditional collaborative filtering algorithms depend only on past user ratings and do not provide any information about the semantic aspects of user interests. This work proposes a hybrid movie recommendation algorithm that combines ALS collaborative filtering and semantic user profiling. The proposed approach generates a list of candidates based on collaborative filtering and constructs semantic user profiles based on positive movies’ ratings using their genres and tags provided by the user. Movie semantic profiles are computed based on feature extraction using TF-IDF and dimensionality reduction. Candidate movies are compared with semantic profiles of the users based on cosine similarity and the hybrid method ranks items based on collaborative and semantic similarities. Experimentation shows the effectiveness of adding semantic aspects of user interests in recommendations. According to the results of experiments conducted using Precision@10, Recall@10 and NDCG@10 measures semantic profiling improves recommendation results without losing scalability of candidate-based re-ranking approach.