Matching users’ information needs and relevant documents is the basic goal of information retrieval systems. However, relevant documents do not necessarily contain the same terms as the ones in users’ queries. In this paper, we use semantics to better express users’ queries. Furthermore, we distinguish between two types of concepts: those extracted from a set of pseudo relevance documents, and those extracted from a semantic resource such as an ontology. With this distinction in mind we propose a Semantic Mixed query Expansion and Reformulation Approach (SMERA) that uses these two types of concepts to improve web queries. This approach considers several challenges such as the selective choice of expansion terms, the treatment of named entities, and the reformulation of the query in a user-friendly way. We evaluate SMERA on four standard web collections from INEX and TREC evaluation campaigns. Our experiments show that SMERA improves the performance of an information retrieval system compared to non-modified original queries. In addition, our approach provides a statistically significant improvement in precision over a competitive query expansion method while generating concept-based queries that are more comprehensive and easy to interpret.