During the last decade, several recommendation systems have been proposed that help people to tackle information overload of digital content by effectively presenting content adapted to user's tastes and needs. However, these personalization technologies are far from perfect and much research is needed to improve the quality of recommendations and, particularly, user satisfaction. In this paper we analyze and extend two relatively recent approaches for improving the effectiveness of recommendation systems: context-aware recommenders, which mainly focus on incorporating contextual information to the recommendation process; and semantically-enhanced recommenders, which focus on incorporating domain semantics. Although these approaches are compatible, how to properly combine them to maximize their strengths is still an unexplored research issue. The objective of this work is to provide the basis for this research. Concretely, we propose and evaluate an improved content-based model that exploits semantics and contextual information in an integrated way.