摘要：Finding a suitable repository to deposit research data is a diffcult task for researchers since the landscape
consists of thousands of repositories and automated tool support is limited. Machine-actionable DMPs can
improve the situation since they contain relevant context information in a structured and machinefriendly
way and therefore enable automated support in repository recommendation.
This work describes the current practice of repository selection and the available support today. We
outline the opportunities and challenges of using machine-actionable DMPs to improve repository
recommendation. By linking the use case of repository recommendation to the ten principles for machineactionable
DMPs, we show how this vision can be realized. A flterable and searchable repository registry
that provides rich metadata for each indexed repository record is a key element in the architecture
described. At the example of repository registries we show that by mapping machine-actionable DMP
content and data policy elements to their flter criteria and querying their APIs a ranked list of repositories
can be suggested.
