Open Data portals provide tens of thousands of datasets, which are described by metadata. This metadata is typically available in one or two languages only and, if translations exist, they are usually added manually. To build an inclusive data infrastructure, metadata should be available in as many languages as possible. The paper presents an approach for automatic translation of metadata within Open Data portals that are based on Semantic Web technologies and use the metadata standard DCAT-AP. Based on this approach, new functionalities are possible, such as enabling users to search for datasets in their native language. The approach was implemented for and tested within a practical application in a production environment.