Multi-relational data mining is one of rapidly developing subfields of data mining. Multi-relational frequent pattern discovery approaches directly look for frequent patterns that involve multiple relations from a relational database. While the state-of-the-art of multi-relational frequent pattern discovery approaches is based on the inductive logical programming techniques, we propose an approach to semantically condensed multi-relational frequent pattern discovery based on conjunctive query containment in terms of the theory and technique of relational database. With the novelty of the groundwork, the proposed approach deals with two kinds of semantically redundant problems. In theory and experiments, it shows that our approach improve the understandability, function, efficiency and scalability of the state-of-the-art of multi-relational frequent pattern discovery approaches.