Business knowledge embedded in legacy information systems is a valuable asset that must be recovered and preserved when these systems are modernized. Event logs register the execution of business activities supported by existing information systems, thus they entail a key artifact to be used for recovering the actual business processes. There exists a wide variety of techniques to discover business processes by reversing event logs. Unfortunately, event logs are typically represented with particular notations such as Mining XML (MXML) rather than the recent software modernization standard Knowledge Discovery Metamodel (KDM). Process mining techniques consequently cannot be effectively reused within software modernization projects. This paper proposes an automatic technique to transform MXML event logs into event models to be integrated into KDM repositories. Its main implication is the exploitation of valuable event logs by well-proven software modernization techniques. The model transformation has been validated through a case study involving several benchmark event logs.