Recommender systems have played a significant role in alleviating the “information overload” problem. Existing Collaborative Filtering approaches face the data sparsity problem and transparency problem, and the content-based approaches suffer the problem of insufficient attributes. In this paper, we show that abundant adjective features embedded in user reviews can be used to characterize movies as well as users’ taste. We extend the standard TF-IDF term weighting scheme by introducing cluster frequency (CLF) to automatically extract high quality adjective features from user reviews for recommendation. We also develop a movie recommendation framework incorporating adjective features to generated highly accurate rating prediction and high quality recommendation explanation. The results of experiments performed on a real world dataset show that our proposed method outperforms the state-of-the-art techniques.